This Newsletter is called AI Supremacy and so part of my mission to consider various possibilities in the future of AI. But what if the future of Space-technology1 and AI are linked? And of recent increasing relevance, are the politics of Elon Musk even compatible with the stories of the future he is known for? Is the aging Tycoon still investible? 2 I’m looking for diverse opinions on this:

Leave a comment

Lately I’ve been thinking about the merger of SpaceX and Tesla, which seems like it’s 95% going to happen soon. There are apparently “committees” at both Tesla and SpaceX looking into the merger. In recent years I’ve become a lot more interested in space-technology and in recent months more interested in National Defense with the Iran and Ukraine conflicts both seemingly escalating. When it came out that Elon Musk ordered Starlink to be throttled (speed restriction) for Ukraine, it got me a bit concerned.

Musk said on SpaceX's earnings call that the two companies have formed committees of independent directors to explore the possibility of a deal.

On the SpaceX Earnings call he basically said Tuesday that he would like to see the consolidation of his business empire by formally merging his two largest companies, SpaceX and Tesla. We know that consolidation is quickly approaching, either in 2027 or 2028 itself.

Elon Musk already has unprecedented control over SpaceX shares and owns about one fifth of Tesla. A combined entity is an incredible concentration risk, where Tesla continues to build inside of China. If SpaceX and Tesla combined it would be in the area of a corporation with a market cap of around $3.5 Trillion dollars. To put things in perspective that’s about equal to the GDP of France.

What is the End-Game?

Elon Musk has a few big projects that could benefit from having a more diversified company with Tesla:

1. Scale Starlink

2. Dramatically increase launch cadence of Starship

3. Scaling own AI chips: Terafab

4. Winning the race to Orbital Compute (Datacenters in space)

5. Scaling Energy to compete with China

6. Winning the robotics and Physical AI race

7. Contributing to founding a sustainable Lunar Base (and colony)

8. Colonizing Mars, and contributing to making civilization multi-planetary to increase our chances for survival

With a bigger company, and one that he has more full control over, Elon Musk would be able to do a huge amount of M&A to ramp up SpaceX into a fairly large monopoly on the frontier of the Space economy, robotics, AI and AI compute infrastructure.

Elon Musk’s Control

• Tesla: Musk holds approximately 15% to 20% of equity (around 20% including vested options).

• SpaceX: Musk holds approximately 40% to 42% of economic equity, but maintains over 80% voting power due to SpaceX’s super-voting Class B share structure.

So if you do the math, Musk would alone control in the area around 35% to 42% of total economic equity in the combined company with far higher voting rights.

Elon Musk could do significant M&A in semiconductors, aerospace, space-connectivity, and even National Defense, acquiring someone like Andruil, advanced manufacturing companies (e.g., Hadrian) and Physical AI software brain companies for humanoids. The stock share leverage of a combined entity of SpaceX and Tesla would give the company a lot of optionality and Musk would be able to dominate the future of Space.

How do you call a company a monopoly when it’s at the frontier of many rather high-risk and speculative emerging technology projects? SpaceX could also be scaled into National Defense entity that becomes weaponized for a kind of peculiar world domination for an oligarch class. Starlink has been incredible implicated in politics and geopolitics. Whatever the legacy of Elon Musk, this entity is likely going to be his last chance.

Tesla brings manufacturing scale, Energy, batteries, EVs, and robotics to the table. SpaceX brings launch dominance in rockets, a struggling xAI business that got stronger with Cursor3, Starlink global connectivity4, aerospace engineering with more shared projects like AI chips via Terafab. Is the combined entity better positioned to achieve what Elon Musk hopes to achieve? It would appear so. The combined entity would be better positioned for the large capex required and aggressive M&A that would be needed.

Tesla in 2026 will spend around $20 Billion in Capex, while SpaceX will likely do around $65 Billion. That’s $85 Billion of Capex just in 2026. Tesla has around $16 Billion of debt, SpaceX have closer to $40 Billion of debt. There’s no guarantee this would work out. SpaceX becoming profitable by being a Neo Cloud is not a great sign for its long-term financial sustainability.

A merger of Tesla with SpaceX was always part of the plan obviously for Elon Musk:

“Because I’m a conflicted party, I can’t actually be in the discussions,” he said, but he added that “I think it would be a good idea.”

Elon Musk sees himself as a visionary and a combined entity gives him a longer career to explore that vision. If the AI bubble pops there will be a lot of interesting AI startups, Physical AI, Inference AI chip and semiconductor companies on sale, that SpaceX could acquire.

In harsher times, SpaceX could cherry pick bleeding AI, aerospace and space-technology startups while attracting more talent to its side with more financial leverage.

Science Fiction Dreamers

Elon Musk is part of a group of Silicon Valley tycoons who believe the era of energy abundance, superintelligence and multi-planet human capability is right around the corner. They believe having more agents and robots than people is a good idea, that will lead to a better life where off-world infrastructure enobles us on Earth5.

Leave a comment

SpaceX, Tesla and all of Musk’s companies already share talent, have circular vendor financing and investments, share major infrastructure projects, that can be bundled into Musk’s evolving grand vision for the future of humanity.

For instance one of the new promises is that SpaceX is positioning orbital data centers (powered by abundant solar energy in space) as a major future revenue source and one of the keys to dominate the future of AI and compute. Given how poorly Tesla handled its robotaxi promises it will be interesting to see if the combined entity can fulfil any, some, or all of these grand projects, dreams and promises.

Now that SpaceX went public with the largest initial public offering on record - the combined entity would have a fairly diversified business model that could super-charge even grander ambitions. Albeit perhaps, with a limited real-world actual impact on consumers and civilization.

Bundling and merging Tesla into SpaceX also enables the company to hide some of its less than stellar aspects into yet another hype cycle. Tesla’s Q3 Earnings were not good where operating income plunged 56% year-over-year to $718 million as operating expenses jumped 39%. Deliveries and automotive revenue were both down.

SpaceX will require stupendous R&D, gigantic Capex6 and nearly flawless execution to pan out in accordance with the words of Elon Musk. That’s highly improbable. But doing massive and aggressive M&A might be the only way to keep the story alive, and a combined entity can do that much more efficiently than keeping the companies separate.

SpaceX began trading on the NASDAQ exchange in New York City at $135 a share, valuing SpaceX at nearly $1.8 trillion. The stock is now trading around $146. Tesla still somehow has a market cap of $1.5 trillion and while it’s not great for shareholders to combine the entities, it will happen. Elon Musk is the most consequential paper Trillionaire on the planet.

Production of Optimus humanoid robots has started on high-volume lines in Fremont and SpaceX’s AI business is hopefully just getting started. Starship being fully reusable and ramping up cadence opens up more possibilities for Space, aerospace and National Defense supremacy.

The merger will face scrutiny by US regulators but doing it (rushing it) while Trump is still POTUS is the goal. SpaceX also in recent months said they will build a massive new $100 billion spaceport named Starbase, in Louisiana in Vermilion Parish along the Louisiana Gulf Coast.

Tesla has a Forward P/E of 182, and SpaceX has an equally aspirational market valuation. Musk’s only hope is to do a massive Space-tech roll-up and spend big on compute and hope it pans out. But if you are investing in the combined entity, you are essentially investing in Elon Musk and the power of American capital, against all odds, including China. Musk’s politics is also a significant business risk, as recently he has committed roughly $366 million to the 2026 midterm election, with much of it aimed at preserving or expanding Republican power in Washington. But at 55, some of Musk’s magic and perhaps his sanity has declined. A projection of an almost childlike masculinity7 and transcendental con.

Those are big rockets you have sir. I hear you want to build a lot of robots. But the majority of Elon Musk’s vision remains highly speculative, with timelines that rarely match up with reality. Tesla makes 85% of its revenue from selling electric vehicles. It’s a car manufacturer pretending to be something more even after all these years. SpaceX makes the vast majority of its revenue from Starlink and selling compute as a Neo Cloud. It could take the combined entity easily a decade to take a meaningful step closer to its vision. As Tesla Robotaxis have demonstrated, that vision can also come a decade late, or not at all.

Some call Musk a genius, a visionary, a technological fraud or a corrupt tycoon, but the future is self-interested. SpaceX will still have to fight for its survival. The combined Tesla and SpaceX entity could have some serious geopolitical, China8 and Succession issues and challenges ahead9.

Share

9

Where Elon Musk owning too many SpaceX shares means that its biggest problem will be Elon Musk, and also what happens to the company after he is gone.

---

Scrap Notes

President Donald Trump awarded Elon Musk the National Medal of Science on October 8, 2026, during the “Science: A New Golden Age Summit”. The National Medal of Science is the highest honor the United States bestows in science and engineering. Elon Musk has been a major donor where during the 2024 presidential election cycle, Elon Musk spent roughly $250 million to $290 million specifically backing Donald Trump's campaign including using his ownership of Twitter/X for Trump’s campaigns. On X, Musk posted or reposted hundreds of messages in support of Trump to his 200+ million followers, generating tens of billions of views.

---

Foreign Election Interference

Ahead of Germany's federal elections, Musk openly endorsed the far-right Alternative für Deutschland (AfD), declaring that "only the AfD can save Germany".

European officials explicitly labeled Musk's posts and platform changes as a form of foreign political interference. His abuses of power include interference in the political processes of Brazil, France, United Kingdom and other countries.

Elon Musk’s Starlink Ukraine and Department of Government Efficiency (DODGE) interference are significant dents in his global reputation and beyond the scope of this article. His increasing geopolitical controversies, challenges facing his core business ventures and polarizing conduct add to the already considerable concentration risk of whatever SpaceX becomes.

Last Thoughts

Whether you are a fan or a critic of Musk, he’s putting oligarchs front and center with regards to the power that will come with technological progress in the 21st human century of history. SpaceX is as much an experiment in Moon shots as it is a diversifying company with perceived value mostly based on Elon Musk’s own reputation.

Good Morning,

Yesterday Anthropic announced weeks before their IPO, Claude Duet - a professional partner you can trust.

It’s basically their answer to Grok Bot, the topic of our hands-on guide (see more guides here) today by (AI Blew My Mind). Duet is a new app where Claude works as a proactive agent. It comes to you with what needs your attention, completes tasks you ask of it, keeps working when you step away, and checks with you before it acts on your behalf.

Read about Duet

For instance for businesses, entrepreneurs, managers, executives or small business owners imagine: Duet can delivers a daily briefing and handles ​tasks ranging from scheduling meetings to building ⁠financial forecasts and web apps. Duet is for professionals, and thus feels polished and specialized.

The Personal Productivity Agents

The target customer audience for Claude Duet is most similar with the audience of Cursor’s Grok Bot. See the promo video.

Join the Duet Waitlist

Professionals of all kinds have to juggle many decisions a day, and Anthropic designed Claude Duet to be the most useful. Duet is built around Anthropic’s Opus 5.5 model ​and is initially available to US subscribers on ​Pro and Max plans, sharing existing usage limits. Most of our upcoming guides (this month) have to do with these Personal productivity agents.

For me Duet looks the most promising for small business owners and entrepreneurs. As to be expected, Duet ⁠integrates with over 6,000 connectors in the Claude Marketplace, including Gmail and Slack, ​and can make purchases via Stripe's ​Link ⁠digital wallet and Shopify's Shop pay. Claude Duet is in Research Preview and will roll-out slowly. It was announced on October 21st, 2026.

AI Blew My Mind

helps you turn decades of experiences into AI systems that work 24/7 for you, so you don’t have to. She’s for me the perfect ambassador to break down how to use these productivity boosting tools.

AI blew my mind

Turn your decades of experience into AI systems that work 24/7 so you don’t have to. Every week, I help you rethink how you operate and build workflows that run for you. (So good, readers say they cancel their day when a new issue drops.)

By Daria Cupareanu

• Jev AI Blew My Mind: 15 Ways to Use It Without Code

• Claude Dashboards Tutorial: Turn Messy Business Data Into Dashboards That Update Themselves

• How to use NotebookLM with Claude Code: 5 demos + 50 use cases with prompts

I asked Daria to give my readers an easy to absorb guide to Grok Bot. The more I look into these personal agents, the more they seem like productivity focused tools.

Cursor made Grok Bot’s growth is part of the reason SpaceX just released surprisingly decent Earnings. Grok Bot as of early October, 2026 has in the area of 800,000 paid weekly users. Cursor was acquired by SpaceX for $60 Billion. I was always very bullish about the talent at Cursor. Watching how Anthropic Duet and Grok Bot evolve is going to be interesting if you believe AI is going to boost personal productivity in meaningful ways at work, for entrepreneurship and for where a lot of Americans actually work, in small businesses across a range of diverse industries.

So, let’s get into the guide.

Grok Bot Tutorial: How to Build an AI Team That Works Without You 24/7

A step-by-step Grok Bot guide for non-technical people, so you can hand your repetitive work to AI employees that keep going while you’re offline.

Read more

Good Morning,

I wanted to review some of the recent AI News in our continued theme for the month: State of AI in infographics. Good News, Bad News and well a lot of macro concern. There’s been significant model innovation on the open-source front lately.

📚 I’m running a limited time discount in case you want to support my work and financially anchor my ability to continue doing this, you can upgrade here. I am grateful for the support.

Subscribe now

So my standards of coverage are not to hype the good news without talking about the difficulties and complex macro and social realities. The AI market needs to face its reality too and fundamentals. I will try to curate a bunch of data points to help you get a more holistic picture of the reality of BigTech, BigAI and financing risks, as well as the experience of consumers and the market.

Share

Nvidia, Reflection AI and Meta just released New Open-Source Models

Reflection’s Beam scores second among US open weights models: Beam scores 30, behind only Meta's Muse Spark Open (39) and ahead of yesterday’s NVIDIA Nemotron 3.5 Super VL (26) release. The numbers represent scores on AA. It’s less clear if this will protect from token marketshare gains for Chinese open-weight models:

Does Beam, Muse Spark Open and Nemotron 3.5 Super VL Protect?

WHAT DO YOU THINK? (Vote)

Reflection’s Beam is a nice addition.

Muse partnered with Apple and claims that Muse Spark Open is optimized to run on Mac Studio with M5 Ultra, and together they’re the best way to run a powerful model on your own desktop.

The AI Financing Breakdown of Autumn, 2026

The gap between what it costs to build AI infrastructure and what companies are actually paying to use it is widening, even as, a massive gap is opening up between AI infrastructure spending and expected returns. The accelerating demand for compute is leading to sharper negative FCF Earnings by BigTech. According to Gartner and others, more than $4 trillion in AI infrastructure spending could be at risk of write-downs by the end of 2032 if companies maintain their current investment pace.

Neo Cloud volatility and coverage centered around Oracle, CoreWeave and Softbank are starting to get more attention. As we had mentioned earlier, Gartner predicts that more than 80% of specialized so-called Neo Clouds will likely cease operating independently by 2030. There’s an infographic for that! But not somehow the full realization yet in the market. By the end of 2026 and heading into 2027, it will get much harder to secure financing. That is, Global AI-related debt issuance plunged to $23 billion in September 2026, down sharply from $113 billion in June, per the Financial Times. A major financing crunch for AI could be right around the corner.

The Demand for Compute and Mismatch is going to get more Extreme

While AI sentiment continues to worsen with datacenter moratoriums, Banks are also actively pulling back to avoid concentration risk in 2026, with JPMorgan CEO Jamie Dimon confirming his bank is tightening up on credit. Approximately $18 billion in bank loans for an Oracle data center campus recently slid into stressed territory, while Nvidia-backed AI cloud provider Firmus shelved an IPO that sought a valuation north of $30 billion. Nscale luckily has Anthropic contract rumors. Anthropic and Nscale IPOs will be ones to watch.

In a nutshell, there is a massive gap between AI spending and AI revenue, one that won't close for many, many years (if it ever does).

In a world where Anthropic is a Neo Cloud customer of SpaceX. Plainly? Anthropic is currently one of the “largest anchor customers” for SpaceX’s newly expanded "Neocloud" AI compute division, utilizing the full capacity of SpaceX’s Colossus 1 data center in Memphis.

Anthropic’s Balance Sheet Worse than we Thought

Anthropic brought in just $16.3 billion in revenue for the first half of 2026 (surpassing its full-year 2025 revenue of $4.6 billion). However, the company reported a massive net loss of $80.7 billion for the first half of 2026. But don’t worry, that’s other people’s money.

The general release of their S-1 to the public is soon, likely next week or 15 days before their IPO. The company logged a $42 billion net loss in 2025, which basically doubled.

SpaceX Squeezes Cursor made Grok Bot Growth and Neo Cloud Revenue

By the Numbers:

• SpaceX reported Q3 2026 revenue of $14.58 billion and net income of $4.4 billion.

• Super Intelligence (SI) / AI Infrastructure Segment makes up 52% of the revenue for SpaceX, which is officially no longer a rocket launch company, but something else entirely. A Neo Cloud with decent Grok Bot growth. The Cursor acquisition is paying off already.

• Consumer Starlink subscribers grew by 1.6 million during the quarter to reach 13.6 million total users growing at a 69% YOY clip. Still one of the main drivers of the overall business.

However in Q3 SpaceX is also spending big on Capex with a lot of debt. Q3 2026 Capex: $19.6 billion (a significant step up from $6.9 billion in Q3 2025 and $18.37 billion in Q2 2026). They have $38.8 billion in debt, with 78% inside ownership (mostly Elon Musk).

• SpaceX’s launch business generated just $1.77 billion in revenue in the third quarter. They are now nearly totally dependent on the AI and Starlink narrative story.

SpaceX will do Capex as if its a Hyperscaler

• Elon Musk guided to $35B to $40B of capex for the fourth quarter (Q4) of 2026, putting it in line with Microsoft's capex spend in the most recent quarter.

• The dream of orbital compute and Terafab aren’t going to come cheap.

AI is Inflationary Pushing the Affordability crisis to new highs in the U.S.

AI inflation is best seen in skyrocketing electricity prices in the U.S. The average price of US residential electricity has risen +11% since January 2025, to 19.6 cents per kilowatt-hour in August 2026, near an all-time high.

Average prices during the first 8 months of 2026 were +40% higher than in the same period of 2019.

The impacts of AI pushing down wage growth relative to its inflationary impacts are becoming more apparent in late 2026:

The US affordability crisis is worsening:

• Food prices have surged +34.5% since December 2019, now at their highest level on record.

• Rent and grocery prices have risen +33.3% and +30.0%, respectively.

• Furthermore, motor vehicle maintenance and repair prices have soared +55.2% over the same period.

• Wages of course, have not. With many who lose their jobs unable to find new ones in a timely fashion.

AI is Increasing the cost of Healthcare in the U.S.

Since 2000, hospital prices have risen 295% — triple the rate of overall inflation. Not many understand how AI is involved in this. While artificial intelligence was promised to streamline healthcare and lower overhead, its immediate financial impact is actually doing the reverse, fueling a cycle that drives up costs for patients.

Credit Markets enter Distressed Territory around AI

There’s bene a major spike in credit default swap (CDS) volume—scaling roughly 4.5x to 5x since mid-summer. Spreads are also considerably higher than the Investment Grade index.

U.S. Housing Market Flashing Warning Signals

U.S. Homes are now the most unaffordable in history. The convergence of high home prices, elevated mortgage rates, and stagnant purchasing power has pushed the U.S. housing market into a historic affordability squeeze. It’s hard to fathom. In basic consumer terms, the average home sale price hovers near $486,000, while median household incomes have failed to keep pace. AI’s impact on wages deserves our special attention given how inflationary it appears to be.

Meta’s Muse Story Too Good to be True

Meta's AI companion Muse has surpassed 10 million U.S. downloads in its first 40 days. However retention and conversion from downloads to real users doesn’t look so good. Meta is also paying to try to make its incredibly expensive 14-month investment in Muse worth it:

Meta has allocated 39% of its total advertising spend to market Muse in October, up from 23% last month.

Meta is also paying big in Ads to boost the retention of Muse. Muse is seeing apparent better retention rates than its AI competitors like ChatGPT, Claude, and Grok. But when you dig into the retention data, it’s not great. Muse only has a 23% retention rate after 14 days since the download (SenseTower).

The scale at which Meta is boosting Meta AI and Muse website impressions is also outrageous, Meta AI grew by 172.29% in website traffic during Q3 2026 compared to Q2 2025. But is any of it sustainable? The mood among consumers is far different than it was in late 2022 for ChatGPT. Muse also has much more competition in Grok Bot, Instinct, Dots, etc…

China’s Great Firewall of AI?

On October 16, 2026, China's Ministry of Commerce (MOFCOM) and the Ministry of Science and Technology published a joint draft document proposing significant updates to the country’s export control catalog. The policy update places 23 AI-related technologies under formal export restrictions, expanding government oversight over overseas transfers of model weights, source code, agentic frameworks, and specialized compute clusters.

The distinction could allow Chinese AI developers such as DeepSeek ​and Alibaba to continue expanding internationally while giving Beijing greater oversight of transfers of sensitive ⁠AI technology abroad.

U.S. Fears China, Bans DeepSeek

The U.S. Department of Homeland Security (DHS) issued a formal order directing all federal civilian agencies to remove DeepSeek software, applications, and third-party tools incorporating its models from government IT systems. The order, originally issued on July 24, applies to federal civilian agencies, including the Treasury, State and Justice departments, but it excludes the Department of Defense.

Amazon and Google Announce Important CUDA like Software updates for their ASICs

Amazon announced TorchNeuron as a free, open-source tool aimed at bridging PyTorch code directly to its custom Trainium and Inferentia AI chips. Rather than requiring developers to manually adapt custom CUDA operators or rewrite software stacks for AWS silicon, TorchNeuron enables execution via simple configuration changes.

Google introduced TorchTPU, targeting seamless PyTorch integration across its Tensor Processing Unit (TPU) infrastructure. TorchTPU acts as a translation and compilation layer that hooks PyTorch execution directly into Google’s custom compiler and TPU hardware backends. They could foster more widespread adoption of TPUs and Trainium chips.

But will they truly hurt CUDA’s moat? PyTorch is the popular open source toolkit that AI app developers and startups use to build, train, and run AI models. In 2022, Meta turned PyTorch over to the Linux Foundation, a nonprofit that manages many important open source projects. Google says it worked directly with Meta, Hugging Face, Inferact (maker of vLLM), RadixArk (maker of SGLang), and others to ensure that TorchTPU’s features and code libraries work well with these popular AI development tools, just as the original PyTorch does. Google’s TPU AI chip business has the most to gain.

4 Years Later No Productivity Boost from Generative AI

Anthropic S1 Leak Details Spook Confidence just 3 weeks from IPO

OpenAI acquires biotech startup

Orchestra Bio is a startup building AI agents to help pharmaceutical companies manage drug development. Orchestra Bio's team and agent technology are being integrated directly into Rosalind, OpenAI’s life-sciences platform. Part of OpenAI’s recent marketing insists it can transform the future of science. However I must say, OpenAI’s M&A strategy and history is not sparkling. OpenAI’s talent for copycat moves continues as the deal draws direct comparisons to Anthropic's acquisition of Coefficient Bio, which closed for a reported approximate $400 million in stock.

Cracks are appearing in America’s debt-laden $10tn AI infrastructure boom as project delays

When too much Capex, is still not enough.

Geopolitics, rising bond yields and macro stories are disrupting the AI story. The Trump Administration has broken the Capex thesis. Let’s try and go into a bit more detail on that angle.

AI Projected Shortfall in Compute is going to be Massive

According to JPMorgan, global AI compute demand is expected to reach roughly 530 GW by 2031, while available supply is projected to reach just 250 GW. That leaves us with a staggering 280 GW shortfall, even after the industry spends hundreds of billions of dollars expanding capacity.

The software, silicon (GPU) and model cycles are moving faster and faster relative to Physical AI Infra, datacenter, energy and HBM bottlenecks.

This shortage forces a radical over-spend where rising bond yields leads to an eventual crash. At least, that’s my base case thesis.

The Big Ask, Circular Financing and Borrowed Money

The AI boom is increasingly being built on borrowed money. As you might have anticipated, AI-related bond issuance has surged to $419 billion so far in 2026, 90% higher than all of 2025. We might be near a tipping point likely in 2027.

The Infinite Demand Thesis is Beginning to Crash

Corporate token spending growth collapsed from 42% in July to 0.2% in September, per Ramp data. Meanwhile, return on invested capital at the five biggest hyperscalers fell from 38% to 20% as capex outruns earnings.

There are structural signs that ROIC is actually declining.

The Pressure of Growth that doesn’t translate to ROIC is Mounting by the Quarter (2026 squeeze begins)

Cost of tokens rapidly declining while most of adoption are from agentic swarms that aren’t translating into real Enterprise adoption that’s sustainable.

Elon Musk’s Ukraine Abuse of Power

Elon Musk ordered the slowdown of a version of SpaceX's Starlink that Ukraine has relied on in its war against Russia, The Wall Street Journal reported.

Musk and President Donald Trump discussed the move to throttle Starlink "soon after" the change first became public last week. Both are aging poorly in a world where they will soon lose their power. How do you rectify the bloated SpaceX market cap with such an abuse of power in a conflict zone? Paying for Starlink now has a moral dilemma.

The move made both the Trump Administration and Elon Musk brands less popular further decreasing AI sentiment among already bearish consumers. The move led to widespread confusion and an internal scramble across Washington with U.S. Defense Secretary Pete Hegseth stated the Pentagon issued no formal directives to limit the network, noting that SpaceX makes independent decisions regarding its terms of service. With boths sides denying responsibility where Musk has considerable business in Russia allied China.

Supply-Demand Shortage for BigTech will only Increase

Iran Conflict Becoming Regional Sends Brent Crude Higher This Week

Brent crude oil surges above $110, its highest level since May 19, after the US carried out strikes on Iran's Konarak Naval Base, wounding two Iranian soldiers.

Dark Semi Nights

In Mid October, 2026 Semiconductor Stocks have now traded red for 8 straight days, their longest losing streak in 16 years.

When Capital Can’t move or solve the AI Demand Blight

Morgan Stanley now sees $1.6 trillion in total 2027 capex. More is never enough when it comes to AI. Where many major firms are cutting back on closed-model token spend per employee in mid to late 2026.

Donald Trump Administration Accelerate U.S. Debt and Interest Payments

Over the last 10 years, US Federal Government Tax Revenue has increased 65% to $5.4 Trillion while Government Spending has increased 92% to $7.4 Trillion. The result: a more than doubling of the US National Debt from $19.8 trillion to $40.4 trillion. Most Americans aren’t even aware of this.

What Would Collapse of Social Order Even Look like?

France can’t afford its own deficit. Now it’s also picking up a €1.2 billion riot damages bill. Debt at 119% of GDP. Rates already surging. No currency of its own to inflate away the problem.

What will be the social cost of an AI Inflation led affordability crisis? Just ask France. With mounting corruption in the U.S. in Government, the crisis begins to spill over to the AI trade as consumers shun AI products, slop and work pressure.

AI isn’t just Inflationary, it’s going to be Sexist

That is, Women may be at greater risk of AI job losses. The history of technological progress suggests that, as AI spreads, the employment impact will hit women more than men. Recent data and labor analyses—including reports from the International Labour Organization (ILO)—highlight that women face a statistically higher exposure to generative AI job disruption than men. So sad to say that women face higher risks from generative AI in 88% of analyzed countries.

Datacenter Delays and Moratorium Compound Losses

At least 47 major data center projects have been DELAYED, withdrawn or blocked so far in 2026. Grassroots resistance has escalated into a national (U.S.) movement. More than 1,000 U.S. communities have sought restrictions, zoning bans, or temporary moratoriums on AI infrastructure.

Furthermore, Industry analysts (such as those at Bernstein) have cautioned that only roughly 35% of the total planned U.S. data center pipeline is currently considered credible or realistic given the harsh realities of supply chain bottlenecks, transformer shortages, and liquid-cooling constraints. Huge wastages of capital at the model (duplication) and capex (Infra) levels.

French Riots Spread to Milan as Bond Yields Spike

France's 10-Year Bond Yield just hit its highest closing level in more than 22 years. French riots over affordability (even pre austerity) have now spread to Italy, Spain, Portugal and even Belgium. In Spain it’s housing, and in Portugal it’s over rising rents.

Cursor Paying Back xAI for Acquisition with Grok Bot Traction

Rental Rates on even Older Nvidia GPUs are Rising Fast

This is boosting the cost of compute and inflating the cost of Capital Expenditures. Even as HBM shortages continue. High Bandwidth Memory (HBM) shortages (SK Hynix, Samsung, Micron) are not only still ongoing, but they remain one of the primary binding bottlenecks across the entire AI hardware supply chain.

Bond Yields Rage as Protests Rock Europe

10-Year Real Yield hits highest level since the Global Financial Crisis . ⏰

Neo Cloud Debt to Equity Sends Warning Signals

SpaceX Revenue Diversifies Substantially

Token Prices Fall as Agent Swarms Take Over

Over the last year, the effective price per one million (1 Mil) AI tokens has fallen by more than 50%. The AI slop internet is about to get an even more automated quality.

Why Open-Weight Models are Winning

Tokens are flowing to different models depending on task, cost, and efficiency. Harnesses and routers mean open-source models have more pricing power for a majority of tasks.

Is China Winning in Open-Weight AI Supremacy?

The 2027 to 2028 period might be the key years.

Europe is Over-Spending on Physical AI and Spuculative AI Startups

Europe in its desperation to keep up with the U.S. and China are now spending in a Venture Capital sense too much, too late, in the AI cycle. It’s what they should have been doing in 2024 and 2025. For instance:

1. Recursive 🇬🇧: $650M. AI that improves itself by analysing its own performance.

2. Focused Energy 🇩🇪: $240M. Fusion power using lasers to compress fuel.

3. EUCLYD 🇳🇱: $232M. Energy-efficient inference chips for AI data centers.

4. OLIX 🇬🇧: $218M. Next-generation AI chips.

5. Gravis Robotics 🇨🇭: $200M. Autonomous systems for heavy construction machinery.

6. Humanoid 🇬🇧: $152M. Industrial humanoid robots for manufacturing and logistics.

7. HYNAERO SAS 🇫🇷: $136M. A next-generation amphibious firefighting aircraft.

8. Quobly 🇫🇷: $134M. Silicon-based quantum computers.

9. Universal Quantum 🇬🇧: $100M. Modular quantum computing systems.

10. Breye Therapeutics ApS 🇩🇰: $77M. A Copenhagen biotech.

11. eleQtron 🇩🇪: $66M. Quantum computing.

12. RougeTx 🇳🇱: $58M. The first treatment for HHT, a rare blood vessel disorder.

13. Biorce 🇪🇸: $51M. AI to accelerate clinical trials.

14. Isembard 🇬🇧: $50M. AI-powered factories for aerospace and defence manufacturing.

15. Sensible Biotechnologies 🇬🇧: $47M. mRNA produced by engineered cells.

16. constellr 🇩🇪: $43M. Thermal satellite imaging.

17. Ore Energy 🇳🇱: $42M. Iron-air battery storage.

18. Adaptyv 🇨🇭: $40M. Protein testing for drug and biotech developers.

19. Mantle8 🇫🇷: $36M. Exploration for natural hydrogen reservoirs.

20. Duna 🇳🇱: $35M. An AI business identity platform.

While Europe has some leverage in emerging technologies like Physical AI and Quantum, it’s general AI positioning is not good.

AI’s Impact on Academia and Science is Alarming

It’s not just about Mathematicians. According to the Economist, Scientists adopting AI appear to be finding success. Yet, while a boon for individuals, a study shows AI use shrunk the range of topics studied and cut scientists’ engagement with each other’s work.

More AI Races Emerge in the pressure for AI Supremacy

Energy Infrastructure and Russian Oil Crisis

Ukrainian strikes have knocked out more than 45% of Russia's oil refining capacity in recent escalating weeks. On the other hand, Russia’s strategy focuses on destroying electrical generation, transmission nodes, thermal plants, and natural gas storage to disrupt civilian life, degrade war industries, and induce severe winter heating crises. The Ukraine and Middle East conflicts appear to be escalating.

Open-Weight Models are Changing the Future

Wikipedia was written in volunteer hours. This AI age commons is written in GPU-hours and data. With Nvidia acquiring HuggingFace, your marketing is my marketing. In effect, Meta, NVIDIA, Alibaba, DeepSeek, Moonshot, Mistral and about ~30 other labs have shared close to $1B (!) worth of training compute on the Hugging Face hub. All of it free to download. At the end of 2023 that number was under $40M.

Will the Personal Agent Fad even Last?

We are about to be bombarded by AI devices and AI personal agents. Forbes just dropped its first-ever Agentic 20. Who do they feature? So you think you are a Personal agent adopter? Try these:

• -> Instinct - personal assistant living in iMessage. $10B valuation, 14 people, $1B in annual transactions.

• -> Meta Muse - the finance-fixing agent. 5M downloads in 3 weeks.

• -> Fo by Wajo AI - books your plumber, humans step in when it gets stuck.

• -> Pion by Andon Labs - agents running a real store, a café, and 4 radio stations.

• -> Known - AI matchmaker. 60% of intros accepted.

• -> Town - AI coworker for small teams. $50-200/user, 9x growth since July.

• -> Govra - AI tax auditors for governments. Already piloting with California.

• -> Dots by OpenAI - always-on agents, connect to 4,000+ apps.

• -> Hermes by Nous Research - open-source agents Fortune 500s run on their own servers.

• -> Atarraya - AI running shrimp farms in shipping containers. $22M raised.

• -> Viktor - the “AI employee” you hand a revenue target. $50M ARR in under 7 months.

• -> TeamTangoAI - answering calls for funeral homes.

• -> Grok Bot by SpaceXAI - teams of agents working in parallel. 400K weekly users.

• -> Link by Stripe - payment layer for agents. Agentic purchases up 38x in a month.

• -> Browserbase - the browser agents actually use. 2,000+ years of browsing automated.

• -> Exa - the search engine built for agents. $250M raised, $2.2B valuation.

• -> XBOW - AI red teamers. 14,000+ exploits found in real apps.

• -> Element451 - AI admissions agents. 400+ colleges on the platform.

• -> Salient - AI agents for loan servicing. $40M ARR.

• -> Stuut - agents chasing invoices. $3B moved through the platform.

Weird AI Devices set to take over Holidays Soon

Amazon last week introduced Bee Frontier and Bee Eclipse, two wacky looking AI devices.

• Bee is “personal AI” that remembers what matters from your conversations and turns it into useful insights and reminders.

• Bee Frontier is Amazon’s so-called next-generation “AI wearable”, and Bee Eclipse is a magnetic phone attachment.

• Of course you can also connect Bee to Alexa+ to ask Alexa for hands-free summaries and to-dos, or to Amazon.com for personalized shopping.

• Bee is private—audio is processed in real time then deleted, and customers have control of their data.

Bee Frontier is $74.99. Bee Eclipse is $99.99. Frontier comes in Onyx, Lavender, and Sage. While Eclipse comes in Silver, Midnight, Honey, and Blush colors. Apple is of course releasing its own suite of Apple Home Collection products. AI devices are essentially always-on by default, for the privacy conscious be warned.

According to OpenRouter Data

Open-weight models took 72% of tokens and 27% of spend last week, in the 2nd week of October, 2026. If they continue to take more token marketshare could begin to erode some of OpenAI and Anthropic’s margins. OpenAI weight gaining token marketshare over closed-weight models is one of the biggest stories of 2026.

Anthropic’s Characterization of RSI

While OpenAI used the “AGI” marketing trick, it appears “RSI” is the prefered acronym for Anthropic. Anthropic employees using Claude Code and the model itself doing more of Anthropic’s R&D work is supposed to be some sort of ramp to RSI. Real superintelligence would imply RSI generalized to everything and independently developing outside the scope of human oversight and organization. An AI singularity for which there is no empirical evidence for.

Nvidia’s Cadence of Nemotron is showing Rapid Improvement

Nemotron 3.5 Super VL achieves twice the intelligence of its predecessor, Nemotron 3 Super (13), and does so using ~60% fewer output tokens per task. It launches alongside Nemotron 3.5 Bolt 74B-A9B 24GB, the first model in its new Bolt family, which scores 21 on the Intelligence Index by Artificial Analysis. Nvidia routinely ships smaller revisions or specialized checkpoints every 2 to 3 months between major updates that took about 5.5 months (170 days). Together that’s the cadence of a potential frontier model maker of the future for open-source.

Just a few days later, DeepSeek released DeepSeek-V4.1-Flash-1017, a minor yet high-performance update to its DeepSeek-V4.1-Flash model family.

New Kind of Visual Model called Glimpse

I’m bullish on one new approach to multi-modal LLMs. For former Google DeepMind researcher and Elorian CEO Andrew Dai this is a step towards more visual reasoning. Mid-reasoning, it can crop into a region, pull specific frames from a video, or run code to measure or compare what it’s looking at, a mechanism Elorian calls a “visual scratch space.” Think of someone pinching to zoom on a phone photo to double-check what they thought they saw. It was announced about five days ago.

Read More about It

The Rising Productivity Gap

McKinsey's 2026 state of AI survey, published in August, found that 80% of respondents say AI has improved their personal productivity, while 37% say it has contributed to EBIT, essentially unchanged from last year.

A recent HBR article offers an interesting perspective on that gap: many companies are using AI to automate individual tasks while leaving the broader processes around them largely unchanged. You can hire more forward deployed engineers (FDEs) but there are structural issues to this adoption story. You need to completely redesign workflows and roles to see max gains. That is, nearly 75% of companies generating the strongest returns from AI are fundamentally redesigning their workflows, compared with roughly 20% of other organizations. This means the impact of AI is a culture story so much as it’s a tech story. Still that’s not the same as real labor productivity data in society at large.

There was so much more I had wanted to cover, but all of these are just snapshots in the end.

I’ve run out of space and time, three days of work and curation. Hope it illuminated.

Mike

Good Morning,

Today we move ahead in our 🗺️AI guides series. You can always check out more of our applied AI guides here.

By the way, if you run an AI Newsletter on Substack and write AI guides and want to be featured in a guest post that can drive some more traffic than usual, you can DM me on LinkedIn here.

Async Personal Assistants - Everywhere!

Since the launch of Meta’s Muse, there’s been a significant uptick in so-called personal agents. ChatGPT has so many users, so it will be interesting to see what they can do with dots. Using AI to run asynchronously in its own cloud environment is simply the new normal in 2026 both for work and personal tasks.

ChatGPT in recent months have added so many more options with Codex, ChatGPT Work and other new additions. OpenAI claims Dots are always-on AI agents integrated into the ChatGPT ecosystem. I asked to give us a simple but detailed guide to OpenAI’s AI agents called Dots.

📚 I’m running a limited time discount in case you want to support my work and financially anchor my ability to continue doing this, you can upgrade here. I am grateful for the support.

Subscribe now

The two themes on AI Supremacy in October, 2026 are (State of AI type) AI Infographics and AI applied how-to guides. In case these approaches interest you.

LearnAIWithMe

A great publication for building with AI, not just to learn how to prompt.

LearnAIWithMe

If you want to build with AI, not just learn how to prompt, this is the right place.

By Gencay

is trending on the technology rising leaderboard of Substack due to his helpful AI guides. Gencay Işık is a former mechanical engineer and AI consultant. As you may know I’m a huge fan of supporting international and foreign writers in the space and his guides are very detailed and easy to follow.

is an AI educator, engineer, and content creator who specializes in teaching people how to build and monetize practical AI systems, agents, and automation workflows. He is a former mechanical engineer and data scientist who transitioned full-time into the AI space. His ability to reverse-engineer AI products and find innovative ways to monetize AI is really fun too - Here are a few that stood out to me:

• Claude Code for Beginners: The Complete Course in One Post

• I Built 10 NotebookLMs From MIT, Hormozi and Buffett

• I Gave OpenAI Dots $600 to Trade

• I Built a Claude Skill That Makes $700K/Year

OpenAI introduced dots on their lats OpenAI DevDay about two weeks ago. This is the guide by of the Newsletter LearnAIWithMe.

“Unlike Codex or ChatGPT, Dots are meant to operate independent of any specific hardware or interface, pursuing user-defined goals continuously in the background with minimal oversight.”

OpenAI Dots

I tested it for two weeks. Here is the setup, the Slack connection, and 4 use cases worth copying.

OpenAI released Dots on September 29, and I’ve been testing it ever since.

OpenAI introduced Dots on September 29, powered by GPT-6 Astra.

Now let’s jump straight in.

Dots

To reach Dots, you need a ChatGPT Pro plan.

Pro comes in three usage tiers: $100, $200, and $500.

All three include Dots. I’m on the first one.

The rollout is gradual, so it can take a few days to show up in your account. If you can’t find it, open chatgpt.com/dots.

How to Create OpenAI Dots?

After you upgrade your account, click on “Your dot” and then “Continue”.

Step 1. Open Your dot and click Continue.

Next, it asks you to connect your apps to make it more helpful.

Step 2. Dots asks to connect your apps. You can skip it.

Since I don’t have any annoying meetings on my calendar, I clicked on “Continue without connecting additional plugins.”

Next, it started setting up my Dot and asked me to wait a couple of minutes.

Step 3. Setting up your Dot takes a couple of minutes.

I feel like I bought a new Apple product and am setting it up. At this step, it asked me to choose where my Dot can work.

It even has its own cloud computer, so it can keep working while I’m away from my laptop.

But I also wanted it to read my files, so I selected it like this below.

Step 4. Choose where your Dot can work and what it can read.

It wants me to customize it, and it answers me in both Turkish and English because I talk to it in two languages.

Your Dot answers in the languages you already use with it.

I clicked on “Customize your dot”.

Step 5. Click Customize your dot.

I named my Dot after my favorite TV show character, with a little adjustment.

(Burhan Altintop + Alfred the butler = Alfred Altintop)

And I picked the following “characters” and “pets”, and clicked on “Save”.

Pick a character and a pet, then save. Mine is Alfred Altintop.

When I clicked his name:

Click your Dot’s name to open its page.

This page opens.

Here I can call him, connect with him via Slack, and see his tasks.

Your Dot’s page. Call it, connect on Slack, see its tasks.

Being able to call it is a really powerful feature.

Dots are very similar to OpenClaw, Hermes, or Grok bots, but being able to call your Dot is a huge differentiator.

Here is the conversation.

Dots on Slack

Your Dot doesn’t live only inside ChatGPT. You can connect it to Slack.

To do that, I clicked the Slack icon here:

Step 1. Click the Slack button on your Dot’s profile.

Next, it asks for my approval. I clicked on “Add to Slack”.

Step 2. Pick your workspace and click Add to Slack.

After a few seconds, it connects. Then click Open Slack.

Step 3. Connected. Click Open Slack.

And here it is, ready on Slack.

Alfred’s first message on Slack.

You can now talk to it on Slack. Add it to a channel, mention it, and it answers in the thread, like this.

Mention your Dot in a channel, and it replies in the thread.

You can check your Slack connection from here.

Alfred’s profile. Slack sits right next to Call.

It’s the same Dot with the same memory. Moving from ChatGPT to Slack doesn’t reset anything.

One gotcha. Adding your Dot to a channel doesn’t mean it watches that channel. To make it monitor, tell it which channels to check, what to watch for, and when to ping you. Then ask it to confirm the recurring task.

My favorite idea from OpenAI’s own docs.

Keep routine progress in ChatGPT, and send only the decisions to Slack.

(If your Slack belongs to a company workspace, your admin has to turn Dots on first.)

Dots Use Cases

I reviewed X to see what people actually do with Dots. Here are four that show something real.

1- Your Dot runs Codex for you

Now let’s say you feel lazy, but you still have things to get done.

First thing, open your computer and give your Dot access to it. (Open your Dot’s profile, find your computer under Computers, and click Allow access.)

Then you can leave the desk.

OpenAI shows exactly this in the video below.

A message goes out from the phone, and since the computer is still on, the Dot gets Codex working on it.

Video by @ChatGPT on X.

In this example, one message asks for three things, and the Dot splits it into three Codex tasks and writes the instructions itself.

A few minutes later, the phone buzzes, telling you all PRs are ready to review without even sitting at a computer.

2- Your Dot runs a one-person company

Now let’s say you want to run a whole company with Dots.

First thing, connect your apps and set your approval rules up front.

(What your Dot can do on its own, and what has to wait for you.)

Then you can stay a team of one.

@beamnxw drew exactly this in the video below.

One founder sits on top, and one Dot runs the work around them.

Video by @beamnxw on X.

In this example, one Dot runs five workstreams for sales, operations, content, research, and product.

In the end, it puts you in the loop, so you approve everything, which is the most important thing.

3- A trading bot with Sonnet, Dots and Jev

Now let’s say you want a trading bot, but you don’t want one AI deciding alone.

First thing, give each model one job. (One writes the agent, one brings the data, one checks it.)

Then you can let three models do the work.

@Sprytixl shared his setup in the video below

Video by @Sprytixl on X.

In this example, this guy built a trading bot using Jev, Sonnet 5.5, and OpenAI Dots, and he says it backtested 1,840 trades over four years in one hour.

Sonnet 5.5 writes the agent, Dots bring the data and 240 test cases, and Jev fences it and ships it. The build took 55 minutes and $4.92, and 221 of the 240 tests passed.

But one thing to tell you. I did the same thing in this post, and OpenAI Dots can’t make trading decisions. I believe that’s why he added Sonnet 5.5 to the mix.

4- Your Dot checks your inbox

Now let’s say your inbox is a mess, but you don’t want to open it.

First thing, connect your email to your Dot. (In your Dot’s profile, go to Customize and then Plugins.)

Then you can just ask.

Thibault Sottiaux from OpenAI showed exactly this in the video below.

His tip is to ask your Dot to “create a pet and set it as your avatar,” so his Dot even has its own face.

Video by @thsottiaux on X.

In this example, one question asks how many unread emails he has, and the Dot answers 39,939, with 21,498 in the inbox.

01-thsottiaux-dot-pet.mp4

Right after, it offers to sort out what actually needs his attention first, without him even opening the inbox.

Final Thoughts

I’ve been using Alfred for two weeks now.

If I had to compare it with OpenClaw or Hermes Agents, which I think are the closest alternatives, being able to call it and let it use your computer whenever you want is a big plus.

But is it worth the $100 subscription? If you don’t use it for anything else, I don’t think so.

Thanks for reading.

Read more about the author of this guide here.

can think outside of the box on how to use AI to improve your career circumstances and monetize AI with unusual clarity and entertaining simplicity.

• I Built a Claude Trading Bot That Copies Millionaires

• Anthropic Pays $300K for a Forward Deployed Engineer. I Built the Training as a Claude Skill.

• I Built a Claude Code Mod That Kept Claude Working Hours Without Hitting the Limit

Thanks for reading. Kindly share the guide if you know someone who might find it useful.

Share

Good Morning,

I’m running a limited time discount in case you want to support my work and financially anchor my ability to continue doing this, you can upgrade here.

Grateful to have the support.

Subscribe now

Nathan Benaich and Air Street Capital released the 9th edition of their incredible State of AI 2026 report. I’ve been covering this every year since I started AI Supremacy because it’s a great snapshot and reminder of where we are at with some unique takes.

Explore the Report

Many of the charts today will come from this source. It’s easier to read these infographics on a computer.

Check out our first batch of explorations of the State of AI 2026 infographics here. Today we’ll go a bit further into the report.

Is Meta’s Muse too good to be true?

This week Meta released its Muse Wonder model and claimed that Muse has 5 million downloads after just 22 days. While Muse Wonder is also a speedy model, the speed of who is blocking the Muse personal agent is equally fascinating. So far:

• Amazon

• JP Morgan Chase

• Multiple Telecom Carriers

• eBay

• Automated Cloudflare Blocks at scale (thousands of websites).

Anthropic has become More Distillation Attack Proof

Anthropic shared a blog about its recent advances in anti-distillation protocols. About to go IPO in just weeks, Anthropic says early evidence suggests that these methods have reduced the amount of data harvested by distillers by around 90 to 99%. This appears to have led to a slowing down of Chinese Open-weight Model releases in Autumn, 2026.

“Preventing distillation is important both for our national security and for ensuring that powerful models, especially publicly available ones, are released with adequate safeguards.”

AI is Getting Cheaper Faster than Any Other Tech in History

This according to the accelerationist leaning Epoch.AI.

Gemini 4 Argon is Now Generally Available

Google has finally given the world access to its latest frontier model, after a seven month delay since its last one. According to Google, Argon excels at deep work in complex, real-world tasks across domains including coding, finance, legal, cybersecurity, and science. Muse Wonder Full and Gemini 4 Argon land in a very similar scoring tier, roughly on par with OpenAI’s GPT-6 Astra, and roughly well behind Anthropic Opus 5.5.

General Enterprise Utility

Google says Argon is especially adept for Enterprise customers:

• To solve major engineering problems like large codebase refactoring and optimizations

• To optimize business workflows across domains like legal, finance, and science

• To find and patch critical cybersecurity vulnerabilities

Frontier Models Over Time

Explore Frontier Models

Is Generative AI really reaching its Peak Impact in the 2027 to 2030 period?

There’s a certain degree of uncertainty with regards to the impact of Generative AI on major industries and the macro economy as compute and models ramp up. I think for healthcare, education and biotechnology we’ll have to wait some more time to see.

Will Europe ever Keep up with the U.S. in Compute?

Europe’s ability to keep up and even adopt and adapt at the level of the United States and China is in some doubt. Europe has neither the capital advantages of America or the talent and energy abundance of China.

BigTech Incumbents Slow to Release Models

Read more

This month full disclosure I’m having a family health issue, so I have less time for research and writing than my usual.

I’m running a limited time discount in case you want to support my work and financially anchor my ability to continue doing this, you can upgrade here. Due to these personal circumstances, the publication may include more guest contributor posts in October than usual. Grateful to have the support.

Subscribe now

Audio Version 🎧 19 min:44

Significant Macro Headwinds, October, 2026

• Recently changes in the macro landscape have triggered renewed concerns over an AI bubble, monetization deficits, margin debt with dangerously rising yields compounded by circular vendor financing, circular financing among BigTech, and significant concentration worries that now nearly totally clouds the AI infra roll-out picture.

• As I’m not a macro analyst, I asked (see LinkedIn), who is a Dubai-based macroeconomic analyst, financial writer, and investment strategist to take a look at the prospects of an AI bubble given all of this recent pressure.

• I hope you are in the mood for reading and understanding about how the big picture could impact the AI boom, Capex and the AI bull market.

MacroVisor

Global Macro & Financial Market insights. About section.

MacroVisor

Global Macro & Financial Market Insights

“MacroVisor takes a top-down view of markets, tracking the trend and cycle shifts that often show up in the macro data before they are fully priced. Our research is rigorous but accessible, written for investors who want to understand the story behind the numbers.”

I’m a huge fan of Ayesha’s Notes on the Substack app that help me keep up with the macro picture. More from the author:

Related Articles

• What Higher Rates Mean

• US Q3, 2026 Earnings Preview: Record Profits Meet 5% Yields

• Why a Diesel Export Ban Could Actually Raise Gasoline Prices

Meanwhile the head market strategist at Panmure Liberum projects that the AI bubble will likely burst in 2027 or 2028, noting that heavy capital is flowing into resource-intensive frontier models. His thesis around local AI and small language models prevailing is super interesting. He’s joins a growing chorus of bears on the AI bubble and was recently on Bloomberg to explain his thesis.

MacroVisor (not a sponsor) is an independent financial research and strategy firm that specializes in global macroeconomics, thematic investing, and cross-asset class strategies in the UAE. This is based on Ayesha Tariq’s expertise of over 20 years working in the field. With higher bond yields globally the accounting for AI Infrastructure begins to look a lot more risky.

Rising Global Bond Yields

Driven by persistent inflationary pressures and seemingly heavy government debt issuance, bonds are rising at an alarming rate in the second half of 2026.

If this post is useful to you or someone you know, kindly share it.

Share

Borrowed Time: How Rising Rates Reach the AI Build-Out

By (Early October, 2026).

Big Tech is borrowing and leasing its way through the AI boom. Higher rates, the midterms and the war will show who can afford to keep going.

Read more

If this info is valuable to you or someone you might know, feel free to share it.

Share

Good Morning,

Editor’s note: This month full disclosure I’m having a family health issue, so I have less time for research and writing than my usual. I’m running a limited time discount in case you want to support my work and financially anchor my ability to continue doing this, you can upgrade here. Due to these personal circumstances, the publication may include more guest contributor posts in October than usual. Grateful to have the support.

Subscribe now

Nathan Benaich and Air Street Capital released the 9th edition of their incredible State of AI 2026 report. I’ve been covering this every year since I started AI Supremacy because it’s a great snapshot and reminder of where we are at with some unique takes.

Explore the Report

Many of the charts today will come from this source.

Claude Haiku 5.5

Five weeks from the Anthropic IPO, Anthropic released their latest iteration of Haiu 5.5. Claude Haiku 5.5 is of course their cheapest, fastest, and most capable small model where they claim it costs around 75% less to run than Claude Haiku 4.5. Haiku 5.5 is considered to be the world’s frontier model in its class.

Only one Frontier Model in Reality

Nathan in the State of AI report insinuates that Google and OpenAI share the “frontier lab” placement with Anthropic, but I do not believe this is the case. Anthropic going IPO for me is the sole frontier model builder now in the world weeks to months ahead of all other AI labs.

I can understand The State of AI’s thesis of three labs sharing the frontier though:

The State of AI report is backwards looking and is written weeks before the present.

In Pioneering Computer Use Haiku 5.5 shows the Frontier is just Anthropic

Full disclosure Nathan Benaich is a British VC, so their report will have an AI accelerationist bias and may not report on social, public and civic issues that are important for the macro context. However for academic and industry synthesis, it’s a highly respected report:

Download the PDF

Air Street Press have a Newsletter, have made decent picks in the development of Generative AI and AI startups in general. Nathan also has a 33 minutes video review of the State of AI 2026 report you can check out on his most recent article on Substack. I have no affiliation with them, some of their best bets include:

Venture Capitalists Claim AI has moved beyond Benchmarks

• Benchmarks built to last years they say, are now saturating in months. It’s not entirely clear how to evaluate this.

Drama between OpenAI and Mathematics Community Continues

“We urge mathematicians to discontinue their work with OpenAI and to return to a vision of science that centers human understanding.”

The backlash to AI is only just beginning.

Terry Tao via the Association for Human Mathematics Communications Working Group. Read the full statement.

Read the blog here. This is a response to OpenAI exaggerating its capabilities when on October 6th, 2026, OpenAI published a GitHub repository (openai/math) containing 722 mathematical manuscripts (grouped into 372 subject families) tackling open research problems and long-standing mathematical conjectures.

AI’s role in the future of science is open for debate. OpenAI stated that out of roughly 4,000 posed problems, each generated paper consumed an average of about 3 hours of "ChatGPT Pro thinking compute".

OpenAI has been Caught Over-Stating its Revenue

A report by the Financial Times indicated that OpenAI’s annualized revenue as of late September 2026 stood at approximately $50 billion, which was roughly $20 billion lower than the $70 billion figure that had previously circulated in financial media. So we know that A significant portion of OpenAI's enterprise revenue comes through cloud distribution partners like Microsoft, Amazon, and Google.

According to FT reporting, the discrepancy arose from attempts by OpenAI's own investors to produce a direct comparison with Anthropic's annualized revenue. Annualized revenue run rate (ARR) ​is sometimes misleading ​sales metric, ⁠which often involves multiplying one month's revenue by 12. It’s not clear what led OpenAI to exaggerate its financial progress to profitability.

Company Behind Hermes

Nous Research may become a serious player in the personal agent B2C space for individual developers and AI first adopters. They have raised $90 Million in a Series B, but also have an Enterprise Hermes on the way. This is an incredible sign for the capability and future of open-source agents. The round was led by Robot Ventures, with participation from other backers including Nvidia, Union Square Ventures, Menlo Ventures, Samsung, and 1789 Capital, where Donald Trump Jr. is a partner.

Nous Research’s will begin a more organized push into the enterprise sector with “Hermes for Businesses,” allowing companies to deploy customized AI agents that can handle multi-step workflows while keeping their data private and secure. Nous was formed in 2023. There’s been a sudden influx of funding for personal assistants since the popularity of Meta’s Muse. Including Underdog (via Conway Research), Hark and Tab.

Muse has 2 Million DAUs

These are estimates from Similarweb, when Meta only had 3 millions weekly users just a few days earlier according to The Information based off of internal Meta data. It’s hard to tell how well Muse app downloads are translating into MAUs, let along weekly or daily active users. The Muse product appears in 2026 far less sticky than Chatbots were in 2023.

State of AI 2026 Report

Hard to believe that this is the 9th edition of State of AI by Nathan Benaich. View the full report.

Will Agents be more popular than Chatbots?

OpenAI’s study of agent adoption finds the fastest growth among non-developers, including people in legal, sales, recruiting, and marketing. Obviously OpenAI has been pushing Codex as its 1.2 billion weekly users of ChatGPT have more options. Outside of individual plans, there are around 9 million corporate/business seats across ChatGPT Team, Enterprise, and Edu offerings. The adoption of Codex however is key to its ability for sustained Enterprise revenue.

Is AI accelerating AI?

AI is already helping build better AI. According to Anthropic’s internal index, Claude led 26% of measured model R&D work in August, up from under 1% in February. How fast will internal model RSI teams accelerate new more powerful models?

Researchers set the tasks and supervise execution, and Anthropic reports that this is speeding up development.

Did Physical AI just Reach its GPT-2 Moment?

Skild’s S1 performed unfamiliar tasks from video demonstrations without updating its weights. With 100,000 hours of pre-training, it scored 66% on the company’s cumulative per-step success measure, against 9% for a language-prompted baseline trained on the same data and compute.

I have written about this trend already here.

If 2026 is around GPT-2 level, when do we arrive at GPT-3 or GPT-4 for robotics and humanoids with general purpose utility.

AI-assisted drug discovery is finally reaching late-stage clinical development

AI-assisted drug discovery is finally reaching late-stage clinical development. Generate:Biomedicines’ GB-0895 for severe asthma and Insilico’s rentosertib for idiopathic pulmonary fibrosis are recruiting in Phase 3. Enveda’s ENV-294 is in placebo-controlled Phase 2a trials for eczema and asthma. Human trials are of course the ultimate test and will determine whether these medicines are safe and effective.

Do this give AI in drug development a more viable path? Will Anthropic continue going further into biology, computational and synthetic biology and biotechnology? It appears they are interested in doing so. While projects like Biohub are expanding, even as China moves ahead in some aspects of biotech and new drugs.

The Inference Explosion

As the demand for compute accelerates but the cost per token is declining rapidly, inference has become the significant intersection of revenue and commercial capture. Meanwhile the cost of Capex and datacenters has risen tremendously due to HBM costs and other supply bottlenecks. This creates a severe environmental press and bottleneck for how fast and meaningful compute has become while AI sentiment continues to degrade both from ordinary American consumers and in politics. As models become more capable with declining prices, Chinese open-weight models have gained market. This is probably the biggest single that has occured in 2026 in AI.

Can Anthropic and/or OpenAI continue their 3x Revenue Growth?

With Anthropic going IPO in five days, they have shown incredible ARR Growth. We will soon have more details about via their S-1 that will go public a few weeks before the date. As for the first period of 2026, OpenAI and Anthropic’s combined reported annualized revenue run rates reached an astonishing $105B by late summer, up from about $30B at the start of the year. That’s more than 3x.

Research firm New Constructs wrote in a note on Tuesday that Anthropic’s IPO would be “the most ridiculous IPO of 2026,” posing a historic risk to financial markets. With a $2 Trillion valuation, we will soon find out what the market thinks.

I don’t believe personally that Google or OpenAI can be characterized as frontier model builders any longer and that Chinese open-weight models should take more token marketshare in 2027. "The Big Short" investor Michael Burry highlighted the valuation gap between Anthropic and established S&P 500 companies, noting that Anthropic's valuation equals the combined value of 78 S&P 500 companies, including top household brands and established industrial firms. A lot is riding on the IPOs of Anthropic and OpenAI (presuming OpenAI does go public in the first half of 2027).

Anthropic’s premium high-spend token share should continue as the leading frontier and Enterprise AI company in 2027, but I don’t believe OpenAI can keep up. OpenAI had to hit pause on Astra 6.1 due to alignment and cybersecurity issues. That is totally unprecedented for a company losing as much money as OpenAI continues to do. This is not what super intelligence looks like! I never trust any post lumping these two companies together.

The SaaSApocalyps Scare - real or imagined?

The success of Claude Cowork and its business plugins across legal, finance, marketing and more brought whiplashing uncertainty to software markets. Aided by some fairly deceptive macro (Citrini was later sold to Semianalysis along with its x following) coverage. Nearly $285B was wiped from software stocks in two weeks during February’s “SaaSpocalypse.” This thesis that AI will be disruptive to SaaS companies continues to persist, when while at the same time I will note that SaaS companies are some of the best at executing new high-growth products related to LLMs.

The Problem of Local Consent for the AI Buildout

With a significant rise in State wide datacenter moratoriums, AI is becoming a political issue. More AI requires financing, construction, and local consent. But these are not infinite where in 2026 AI sentiment sourced considerably for most people. This as a European plan is seeking up to €10B in public funding and at least €20B privately. While IPOs like Nscale will test “Sovereign compute” in places like Norway. Most funding of European AI, robotics and datacenter initiatives actually comes from the U.S.

RESEARCH

Read more

Hello Everyone,

This week full disclosure I’m having a family health issue, so I have less time for research and writing than my usual. I’m r…

Read more

There’s something that’s not quite right with what’s going on in the world. Listen to me carefully today, if you care about the future of the world. I think we are misinterpreting what AI risk actually is. It’s not just about the broken culture in companies like OpenAI, but the incentives and cultures of the powers behind it and the Venture Capitalist interests and incentives who have, in part, taken over key areas of tech and AI policy in much of the direction of this U.S. Government, including at the Pentagon. Let me try and explain briefly. It’s a very serious situation.

The full-scale invasion of Ukraine by Russia began in late February, 2022. That’s the same year that Generative AI took off four and a half years ago. I’m noticing worrying signs of escalation both in the Ukraine conflict but in how the U.S. is preparing to use AI to militarize (and weaponize) the technology in the future. The overlap of armed conflicts with the advent of drone swarms and more sophisticated military robotics means the U.S. is preparing for the real possibility of world war III.

The Ukraine conflict has become a mini-laboratory for the U.S. to study asymmetric warfare including the use of frontline robotics. So suddenly in 2026, there’s a huge spending spree by the U.S. and European VC community on robotics and Physical AI systems. Nowhere is this more apparent than the activities of people like Eric Schmidt while the Palantir related private company Anduril is ready to arm Taiwan. After years of setting the U.S. on a collision path with China, Schmidt runs Project Eagle (formerly called White Stork) this is basically a stealth initiative founded in late 2023 that focuses on building low-cost, AI-enabled kamikaze and interceptor drones.

Everything from Space to deep-sea mining and everything beneath the ocean will become contested zones between the West and the Eastern Block. In the next ten years, the U.S. will try to build an unimaginable weaponization of AI that will make the world a far more dangerous place, even as the Ukraine conflict appears to be escalating.

“We Must Win”

I try my best on this publication to cover AI news from many angles, from industry news to startups, to guides on how to use AI. But my biggest fear is that we’d reach this point where global conflict escalation and increasingly AI mediated technologies appears to be converging in an unregulated and accelerated way. There are signs we are near a tipping-point for Military AI in the United States of America.

I believe that between 2027 and 2037, in this immediate next decade we’ll be reaching the most dangerous period for military AI to cause massive harm to human civilization. I’m not a National Defense analyst but with Russia and the Ukraine targeting civilian infrastructure and fears that Russia will also begin offensives or escallate on the European (NATO) front are a cause for concern. That is, just when AI appears to becoming more powerful in things like cybersecurity (a military application), global conflict appears to risk serious escalation. I believe it will soon become impossible to maintain human-in-the-loop systems and safeguards in some, and a growing number of multiple and many new aspects of warfare.

Subscribe now

Read more

Hey Everyone,

So what’s going on at the AI frontier this week? I wanted to check out and round-up some of the AI news this week that caught my attention.

Frontier Models

As we anticipated, Google announced their Gemini 4 Argon model this week. Alphabet says its built to sustain deep reasoning across complex, long-horizon workflows. It’s last flagship model was way back in February and it’s difficult to understand what took them seven months to launch another one. We’ll know more on post October 1st after cybersecurity testing, but the first signs look promising.

It appears to truly put Astra GPT-6 behind us. Astra OpenAI’s frontier model was only seemingly frontier for 19 days, before Opus 5.5 was announced.

Meanwhile in the span of 2026 Google saw another huge exodus of top AI researchers (e.g. DiscoveryLoop) and talent. It’s also not clear if Google will have an answer for the personal assistant wars of Muse, Instinct, Grok Bot and Dots. Have you tried them?

• Muse

• Instinct

• Grok Bot

• Dots

The other companies most likely to make a Muse competitor for the B2C space are ByteDance and Google.

Personal Agents Accelerating Token Spend

Read more

Good Morning,

Dermot McGrath is an Irish entrepreneur and advisor based in Shanghai with over a decade in China. He recently founded a research and strategy advisory firm focused on AI, robotics, energy and advanced manufacturing.

ZenGen Labs Newsletter

Dermot and his team report on the ground in Shanghai and beyond, on how technology actually gets identified, built, and scaled.

ZenGen Labs

Operator-led analysis on deep tech, global markets, and execution across complex ecosystems.

By Dermot McGrath

Recent Articles by the author

• From Turbines to Tokens

• Space as a Service: China's Alternative to the SpaceX Model

• Building Trust at Grid Scale

• The Printer Maker That Built a World Model

For more deep dives like this, consider supporting AI Supremacy and our numerous guest contributors. We’ve lowered our subscription price for a limited time.

Subscribe now

Consider sharing this article if you know someone interested in the future of Physical AI, robotics, embodied AI and China’s role in its future.

Share

The State of Chinese Physical AI

Five years ago a humanoid robot was a clumsy research tool that cost about the same as a new house. Today, for the price of a used car, you can order a humanoid that does kung fu and parkour, from a company in Hangzhou. That company is called Unitree (宇树) and it went public on the Shanghai stock exchange in August 2026 at a valuation of $50 billion.

Morgan Stanley estimates that by 2050 there could be more than 1 billion humanoids on the planet deployed in a range of scenarios including: * running the night shift in factories * stocking the shelves of your local supermarket * running warehouse fulfillment * cooking, cleaning and looking after the elderly at home.

To get anywhere near that future, someone has to build these bots. If current trends continue, most of them will be built in China.

Humanoids are one part of a much larger field now known as physical AI. These machines can perceive what is happening around them, reason about what to do, then act, all in the physical world. It spans self-driving vehicles, industrial robots, drones, quadrupeds, dexterous robotic hands, as well as the AI models that control them and the data systems used to train them. The Chinese term for this industry is 具身智能 (”embodied intelligence”, also called “embodied AI”).

China already leads in many of these fields. It is the world’s largest market for industrial robots, and DJI leads the global commercial-drone market, according to market research firm IMARC. China now operates the world’s largest combined robotaxi fleet, spread across companies including Baidu (百度), Pony.ai and WeRide. Crucially, its factories produce many of the motors, sensors, batteries, and precision components these machines rely on.

Humanoids get most attention, and most of the capital, because they make the loftiest promises. A robotaxi is built to drive, an industrial arm to perform tasks at a workstation. A humanoid, however, promises a general-purpose machine that can move from the factory floor to the warehouse, the shop and the home, carrying bricks, threading needles and everything in between.

But it takes more than just brawn. Despite the promise of a machine that can do almost any job, most humanoids still rely on human remote control or preset programs, according to Chinese technology publication TMTPost.

The companies that make them are raising obscene amounts of money anyway.

Half a century ago China was among the poorest countries on earth. Today its companies ship more humanoid robots than the rest of the world combined, supply many of the components other robot makers need, and are spending billions to teach these machines to think.

In this piece, we will examine how we got here, the current state of play from the body to the brain, where the money is going and what the future may hold.

First, how did China get so good at building these machines?

The Workshop of the World

When Deng Xiaoping, then China’s leader, began opening China in 1978, the path out of poverty was well trodden: industrialize. Leverage abundant labor and low-cost land to build factories and make the things the world needed. It had nearly a billion people and a long coastline where factories could ship to the world. Over the following decades it attracted foreign companies and accumulated infrastructure, manufacturing capacity and supply chains at extraordinary scale.

Read more

It’s impossible to cover AI and not talk about Meta Muse. This is not an endorsement, just some of my observations and information. Just yesterday Meta has announced Meta Enterprise Platform.

This at a time when AI has literally disrupted media and creatives1.

AI and the Fall of the Creative Class

Read It

Young people might be deskilled in how to code and how to write simultaneously, within just a micro generation. That’s not by the way, the empowering AI we were promised.

Will Agentic Swarms Disrupt Core Skills of Human Literacy and Intelligence?

ChatGPT Era is official Over

Whether we like it or not, agentic swarms are here. In terms of AI Supremacy on the consumer (B2C) front, the last four years have been the Chatbot era where ChatGPT of OpenAI was the dominant leader2 (late 2022 to late 2026). This weekend + all Monday I’ve been going over Meta’s Muse product and I can say with some confidence that the ChatGPT era is now really over. We’re about to embark for likely the next few years into a journey of Personal Super-Intelligence Agents (PSIA). Meta is SOTA in PSIA, let me explain:

Subscribe now

Muse’s Rapid Ascent and a new Personal AI Race

I predict Muse will have around 1 million users by November, 2026. In 2027 it could continue to grow quite fast. The AI industry is a bit confused about the branding, not a “companion”, but maybe more like a teammate or casual personal assistant that is supposed to help you optimize your personal life that will become increasingly capable to actually accomplishing tasks for you.

When do Personal Super-Intelligent Agents Arrive?

The new era of AI begins when two things occur in Meta’s timeline: the arrival of its Watermelon model (due in late October) and the finishing of its Prometheus (near Columbus in New Albany, Ohio) datacenter coming fully online in the next few months. By then adoption of Meta’s Muse agent will be even more impressive than its first month. So this new era arrives circa or around December, 2026. Right after Anthropic’s IPO a new race in consumer AI kicks off (that’s right after U.S. midterms). Simultaneously ChatGPT will be losing marketshare to open-source models that are taking more token marketshare especially in the last six months3. OpenAI is thus in deep trouble in 2027 for a few major reasons:

Signs of Cracks in OpenAI Dominance are starting to show

1. Pressure in Consumer AI from Meta and SpaceXAI.

2. Pressure from Chinese open-source models on cost and token marketshare.

3. Late to IPO, and markedly over-taken in ARR and growth by Anthropic in Enterprise AI.

With Google late to deliver their next frontier model, among BigTech incumbents only Meta has made a serious challenge to be a frontier player in both LLMs and consumer AI products. The Meta AI app even before Muse has gotten very decent traction the past nearly 1.5 years. Muse is currently available only in the U.S. and Canada.

Try Muse

Meta’s Scale Advantages Up-sell its First Comer Personal Super-Intelligence Move in Muse

Muse was launched on September 8, 2026 or around 3 weeks ago. By the time its Watermelon 🍉 model drops it will be about seven weeks and where more countries will likely have been rolled out to try the first iteration of Muse. Meta has two major advantages over SpaceXAI, OpenAI and others trying to be players in the next personal AI market. These are huge numbers of daily active users and a warchest of Ad revenue. Meta reports an average of 3.60 billion “family daily active people” (DAP). Meta is also the internet’s dominant digital advertizer.

Meta’s full-year total revenue is projected to be in the area of $245 billion and Muse could materially accelerate their advertising spend for small businesses at scale. Meta also has resources to spend on marketing their new AI products that most AI startups don’t have. Meta is widely expected to have significant negative FCF by the end of 2026. The compute spend on Muse will make a difficult company to value, and likely far more risky to invest in. All of this inspite of Meta’s impressive adoption rates and the performance of the app on leaderboards.

Token Spend and Machine-to-Machine Inference (M2M Internet)

The PSIA (personal super-intelligent agents) market will radically accelerate token use and only truly profitable companies will be able to compete due to the high compute (costs to subsidize), talent, operations, training and RSI costs. These systems will be designed to keep getting better and more personalized to the user including up-selling them to new more specialized AI products in the next few years and nudging them to use Meta’s digital advertising to grow their side hustles, entrepreneurial projects or existing small businesses.

Products like Muse will contribute to a new definition of what it means to be compute constrained in the late 2020s. It’s going to be so expensive, few will be able to follow.

Share

Why is Meta Muse the First Mover?

Meta’s Muse is the consumer culmination of Meta’s ambitions in the Metaverse and AI merged. According to my calculation Meta will have spent upwards of $180 Billion on this PSIA project by 2027. They will have gone into significant debt to do it.

That’s multiples more than Grok Bot of SpaceXAI or OpenAI’s Codex Bot (“o”) have to put into a persistent agent experience with dedicated memory and a cybersecurity hardened and encrypted Virtual Machine space for swarms of agents to work dedicated for you. Meta also has such a treasure trove of data to enable Muse to help us maintain and “show up” for our friends, family and personal connections. This is language Mark Zuckerberg specifically used in his multiple recent interviews and podcasts.

It’s actually fairly interesting to think of Muse as the first major project of Meta’s new Meta Superintelligence Labs founded around 15 months in.

Meta’s Muse Timeline

It only took AI to go from ChatGPT to Muse in 45 months. Not fast at all when you think about it. ChatGPT was released on November 30, 2022, and Meta's Muse was released on September 8, 2026. But for consumer B2C AI products, a lot happend in those nearly 1,400 days. For most people it hasn’t been that exciting to say the least. Meta’s Muse might have a great team, and the product might still not catch on like ChatGPT did four years ago. But how did we get here?

Meta’s Superintelligence Lab (MSL) was formed just around 14 months ago: Meta’s core Superintelligence frontier team has about 44 all-star members. They are supported by thousands of other machine learning researchers. Muse is the result of a long journey of capital, compute and talent concentration (that will be hard for Google, OpenAI or anyone else to emulate). The entire MSL division is around 3,000 AI researchers with a total of around 6,500 AI related employees in the applied AI support unit. This is what it takes to build something like Muse.

• Some Context: The early prototype of OpenClaw was first launched on November 24th, 2025. OpenAI later acqui-hired the founder on February 14th, 2026 but failed to capitalize on the opportunity. The entire time Muse was in development.

• April, 2025: The major setback for Meta's Llama series occurred with the release and aftermath of Llama 4 in April 2025, culminating in early 2026. This signalled Meta AI would need a restart.

• April 29th, 2025. Meta releases its standalone Meta AI App on April 29, 2025

• Late June, 2025. The Great talent poaching: The bulk of Meta's aggressive poaching to build its Superintelligence Labs took place during the summer of 2025, specifically kicking off in June and July.

• Late June, 2025. Mark Zuckerberg announced the creation of Meta Superintelligence Labs (MSL) on June 30, 2025, with the official public announcements and restructuring details following in early July 2025.

• Late July, 2025. Meta's blog post titled "Personal Superintelligence" (also published under the title "Personal Superintelligence for Everyone") was written and published on July 30, 2025, read it here.

• Mark Zuckerberg's supposed Public Manifesto, a 6,000-word essay titled “The Future is for Everyone” was released on August 10, 2026. Read it here.

• Meta’s Muse agent was launched on September 8, 2026

• Meta’s next-generation frontier AI model, codenamed "Watermelon," is expected to be released in late October 2026 and to be at the level of GPT-5.5. approx.

• Re Prometheus datacenter near Columbus Ohio, the marquee milestone for Prometheus to hit its initial 1-gigawatt (GW) target is scheduled for late 2026 around December, 2026. Major data halls and initial on-site gas-turbine power infrastructure (such as the Socrates sites) are turning online throughout late 2026 to reach this capacity. (Hyperion is approx 5x the size of Prometheus).

• Meta’s biggest datacenter Hyperion, will likely finish its first initial stage in 2027. The first data halls and infrastructure zones are scheduled to complete commissioning and go online, delivering an interim capacity target of roughly 1.5 GW in Richland Parish, Louisiana. Full capacity will take at least until 2030.

• September 28th, 2026 - Meta has just announced Meta Enterprise Platform.

• September 29th, 2026 - Meta announce Muse for Small Businesses.

Is Muse a ChatGPT Killer?

I’m curious if these persistent AI agents will essentially cannibalize vanilla chatbots. How about for you in the time you spend with and on AI?

What can you really do with Muse? (Use Cases)

Read more

Good Morning,

As Anthropic gets set for a $2 Trillion IPO in just over a month, it’s getting close to unleashing its version of RSI in biology, drug discovery, biochemistry and biotechnology. What could go wrong? Recursive self-improving AI or today’s version of it should not be unleashed on drug research and biotechnology.

Both Dario Amodei and his angel investor, Google DeepMind’s Demis Hassabis have made incredible claims about AI curing Cancer in the next decade. Predictions such like AI being capable of accelerating biological research by roughly 10x, compressing 50 to 100 years of biological and medical progress into just 5 to 10 years. But at what cost? Anthropic will tread dangerous ground without hardly any oversight.

I’m getting increasingly alarmed by the compute Anthropic will have at its disposal to put its world-leading models at the intersection of biotechnology in the 2030s. The beginning of this dangerous turn was when Anthropic lured Eric Kauderer-Abrams who joined Anthropic in August, 2025. He initially came on board to co-lead life-sciences R&D and partnerships, and currently serves as the company's Head of Life Sciences, overseeing its wet lab operations and AI-driven drug discovery programs. Then in June Nobel laureate John Jumper, co-creator of the protein-structure AI AlphaFold, left Google DeepMind after nine years to join the AI startup juggernaut Anthropic.

America’s leading monopoly LLM lab will be trying to do what so many other Pharma startups have failed to do, use AI to accelerate drug discovery. But it will and it has already conducted dangerous experiments along the lines of accelerating RSI in biology. It has now come out that Anthropic has set up a wet lab in the San Francisco Bay Area to conduct physical biology experiments, as the company moves beyond computer-based research in its effort to develop treatments for rare diseases, or so it claims publically.

At a time when fears of AI are gripping the public, Anthropic will push the limits of the intersection of the world’s frontier models and our understanding of pharma and biotechnology. Anthropic this week claimed Claude discovered a novel enzyme system.

Read about It

Read more

Editor’s Note: AI Evolution is a new series I’m doing that goes out at 12:00 PM Eastern Time, with a focus on AI related breakthroughs and macro developments that are key to the industry’s future. I’m not sure of the cadence rate of these articles yet but they are meant to be short and easy to read/skim about pivotal can’t miss moments in AI development and AI history useful for a wide range of readers including investors, executives, AI news fanatics and venture capitalists.

Good Afternoon,

The incredible Capex by BigTech have created massive Neo Clouds that specialize in AI Infrastructure with GPU-as-a-Service business models at the most physical level of compute. The biggest of these Neo Clouds with the top financing and biggest customers are going to become massively powerful companies.

Many of the more innovative and well funded Neo Clouds have yet to go public. On September 17th, 2026 Crusoe confirmed it raised a $3.9 Billion Series F round. Companies like Crusoe and Nscale yet to go public have a lot of momentum to catch up larger Neo Clouds like CoreWeave and Nebius.

Crusoe is now the most backed private neocloud in terms of equity, enabling it to get significant debt for expansion. They are a vertically integrated company independent from grid readiness that has overcome early reliance in customer concentration with OpenAI, now serving Microsoft, Meta, Oracle and Jane Street. They are backed not just by Peter Thiel’s Founders Fund but by several of the world’s top Sovereign Wealth Funds and major Tech corporations.

Subscribe now

Read more

Share

Good Morning

I don’t usually nerd out on non-LLM machine learning announcements. But extraordinary claims have been made. I hope you are in a curious mood.

Now that agent swarm consumer products like Meta’s Muse are going mainstream, we are on the cusp of models that will make machine to machine (M2M) processing and communications faster and more efficient. In my humble opinion that in a nutshell is what Jev is.

Try Muse (in U.S. & Canada)

So what is this Jev AI exactly?

Jev has gone Semi-Viral as new kind of Model

About a week ago on September 15th a former OpenAI employee, Diogo Almeida, announced Jev:

Jev: A Probabilistic Classifier with a Manifesto

Jev is a classifier that generalizes very well. Almeida is credited with being one of the co-inventors of reinforcement learning from human feedback (RLHF) and thus a co-inventor of some of what which became ChatGPT. Jev has potentially the ability to make some aspects of agentic AI more cost efficient while it gets its name, Jev, after Jevons paradox.

We still know very little about Jev's internal architecture, training setup, and model size. TypeSafe.ai have raised $40 million Seed and built a product called Jev that is an AI model built for fast, structured decisions rather than open-ended text generation. So you feed it state + typed questions (Choice, Score, or yes/no), and it returns calibrated decisions in 70–500 ms at almost zero cost ($0.042/M input tokens). You can read their manifesto here.

System-1 Might be useful for Agentic AI

Jev is Typesafe AI's new System-1 model that skips token-by-token generation for instant decisions and the internet the past week have found all kinds of use cases for the technology. According to Nicolas Bustamante what we might be witnessing in AI is - err, well, a textbook example of Clayton Christensen’s Innovator’s Dilemma playing out in AI. We are at that weird stage where the market is commoditizing the bottom of the ML market: think classifiers, routers, scoring and context triage. So is Jev viral AI non-sense or something useful? I think that remains to be seen.

Almeida left OpenAI two years ago to start TypeSafe AI and release a transformer based model that is not an LLM that produces probabilistic calibrated decisions. That is, Jev is what Typesafe calls a “System 1” model: it outputs decisions directly, as probabilities and confidence scores, rather than autoregressively predicting the next token. It’s certainly making debate about advances in classifiers sound more sexy.

It’s a model that is garnering significant interest from developers. That is, Jev appears most useful for software automation. It might turn out that consensus among software developers will see it as a cheaper and more robust way to incorporate intelligence into their code. In a world where agentic swarms are using more and more of the total demand for inference compute, that kind of makes sense. This a generalist model that might be useful for swarms of agents to become even faster and less costly.

Almeida claims Jev is built for the decisions inside software. Now Jev has been released in early access and its stimulating the open-source community. Typesafe’s selling point is that Jev is up to 100 times faster and 100 times cheaper than conventional LLMs for certain tasks.

Read the Blog

The Era of Hundreds of Auxiliary Specialized kinds of Models is Coming

In today’s world developers can embed Jev in products for high-volume, low-latency decisions such as classifying/routing support tickets, scoring leads or content, safety/flagging checks, verification steps, agent routing, or other “if-statement–like” logic inside automated pipelines with likely hundreds of potential use cases that have at least some business value. Jev is a sign of things to come in probabilistic AI and one one of the many new kinds of (often open-source) models that can make products better.

In an AI boom it’s interesting what becomes a potential product. Classification, scoring, routing, and intent detection are not new machine learning problems. What TypeSafe AI is doing differently is building a model specifically around these kinds of typed, probabilistic decisions, rather than taking a general-purpose LLM and prompting it to behave like a classifier.

Typesafe AI claims that at $42 per billion input tokens and free output, intelligence becomes a resource accessible to all.

Read Jev Docs

A week after launch, Jev has already been added by popular model gateway platforms such as Vercel, Cloudflare, LangChain, and Langfuse to their stacks. Jev is like a weird viral decision-making AI in September, 2026 and yesterday they announced that Jev is now available to everyone, without a waitlist, with access starting at $5 in credits, equivalent to roughly 120 million tokens. This is not a sponsor, but you can create your account here. Jev is also anticipated to be useful in agentic AI tasks like model routing.

Jev can use Far Less Tokens than Incumbent LLMs for some Tasks

Jev is very similar in spirit to zero-shot classification and seem to be able to do things a bit better and a lot more cheaply. I’m super interested in how Jev could make agentic AI faster, smarter and more efficient. In that fundamentally Jev is an AI model optimized for “machine-consumable decisions” rather than human-readable conversation and therefore likely better positioned as infrastructure for software rather than a chatbot/LLM. Jev is closer to the essence of what machine learning is but powered by the transformer architecture and perhaps made more powerful in an era where LLMs are so prominent.

Jev is described as non-autoregressive and might be based off of the work or related to the work of others. This work was done previously in part by Nandakishor Mukkunnoth and which he explains in a LinkedIn post. There have also in the past week been many spin-off versions of Jev that have gone open-source or been announced. He also shared his story on Hacker News here. Jev doesn’t reduce hallincations per se but delegates them to the user in a sense. Anything that reduces token usage is bound to go viral more easily in the current climate of AI compute.

TypeSafe AI has re-designed zero shot classification substantially since the time it was popular (2019-20) and added multiple structured decisions, probabilities, parallel inference, and a new calibration-focused training approach all in terms that are fairly familiar for NLP engineers. Six years later saying it’s a new kind of model isn’t 100% correct, but it’s certainly re-packaged in ways that are relevant to the era of Agentic AI and the challenges developers and product managers face.

Jev Reminds us that LLMs are not all We Need

TypeSafe AI attributes Jev's efficiency to its specialized architecture, parallel sampler, and calibration-focused training. It’s a generalized model for much narrower tasks designed to make structured decisions directly and in parallel. Immediately it can help AI agents decide which tool to call next and whether to retry the task, by answering those questions in under half a second for $0.042 per million input tokens. Output is free. It makes you wonder how in the near future different kinds of transformer models will combine, complement and improve what LLMs are capable of doing today. What will the decisions trees and probability matrices of the future even look like? Jev really unlocks the imagination of what will become possible.

“You can break a workflow into individual judgments, run them in parallel, and combine them however you want.”

Jev is an architecture is optimized specifically for fast, structured decision-making and will radically reduce the cost of some layers of the product for developers all over the world. Think about what that means? Jev could arrive just at the time to become an important part of the decision-making layer underneath a growing ecosystem of AI agents.

The model is named after William Stanley Jevons, the 19th-century economist whose eponymous paradox describes how the falling cost of a commodity can lead to it being used more and more. Jevons paradox has indeed been that whisper witty water-cooler response moment on repeat all during the AI boom.

Models such as Meta's facebook/bart-large-mnli have long been used for lightweight zero-shot classification. But the developer tool kits keep getting refined.

“The problem is we are optimising for human language. We have been super good at human language for four years, but it’s not useful for automation because computers speak a different language.” - Diego Almeida, TypeSafe AI

Jev is most likely built upon an open-weight LLM, i.e. most likely a Chinese model. Almeida says Jev is trained exclusively on synthetic data using a technique he calls “reinforcement learning from calibrated decisions.” TypeSafe AI reports Jev at around 68% (accuracy) on its own workflow evaluation - which is to say we have no idea how accurate these results are on real-world tasks. Jev’s general accuracy is very much now being debated and time will tell.

News Models are Boosting what Developers can do

I think Jev is super interesting especially for developers who are builders working on real-time systems, games, agent loops, or anything that currently avoids LLMs because of latency or cost. I’ve seen an impressive amount of use cases online in just the past week being mentioned.

RLCD stands for Reinforcement Learning for Calibrated Decisions, the training method TypeSafe AI says it developed for Jev. In essence this makes classifiers more useful. Think about it RLDC here refers to training methods in which a computer program not only learns to make good decisions, but also to realistically assess how confident it is in doing so. The goal is a system whose self-confidence matches its actual hit rate.

A standard RL agent might assign a 98% probability or preference to an action that only succeeds 60% of the time. Here they are trying to correct this and for narrower tasks it could make sense. If Jev’s accuracy and confidence improves it could end up being a rather useful model. The online speculation around Jev’s architecture (this blog by Archer Hume) is also fairly fascinating.

“What I like about Jev: for years, large generalist models sucked almost all the oxygen out of AI.

But there are huge opportunities in much more specialized and customized models, built for specific tasks and languages and as a result, orders of magnitude cheaper, faster and more optimized. There are 3 million of them publicly available on HF” - Co-Founder of Hugging Face.

I believe Jev is just a beginning for new kinds of non-LLM models that are made for more specialized workflows, tasks and use cases. Many of these specialized models are trained on open-weight Chinese models and you can see how open-source AI is stimulating complementary machine learning and models to complement and make up for some of the limitations intrinsic to language based models, or LLMS.

Not only are LLMs compounding with compute but more specialized models are doing important work in industry domains where LLMs are not optimized to solve their key domain problems.

Subscribe now

Meta’s Muse has better First Two Weeks than ChatGPT?

According to Apptopia, Muse’s app on mobile devices has been downloaded more times during its first 12 days on the market than ChatGPT was in the 12 days after its debut. This is with regards to where Muse is now available, in the U.S. and Canada.

• In this subset of the data, Muse has now seen 1.8 million downloads to ChatGPT’s 1.3 million.

• Overall, Muse has seen 2.8 million total installs globally in its first 12 days.

App Figures made the calculation before that Muse had overtaken ChatGPT as the most viral new AI product in history (see chart below). Apptopia’s data also indicates that Muse’s U.S. daily active users are now higher than they were for ChatGPT at the same point after its launch.

To be fair we have to note that Muse is available across both iOS and Android, while ChatGPT launched on iOS only at the very beginning. Meta also has multiple apps with billions of users (to upsell Muse) and a larger marketing spend available to them.

Conflicted reports on X about the Jev Announcement

Thread about Potential New Use Cases for Jev

Will Jev survive as a Real Product?

Pre GPT ML Engineers are Confused

The Usual Copy-Cat Cadence from a Viral Announcement

Jev asked us to think differently about Old Problems

Confidence scores for business development outreach?

On Using AI as a Tool

Could this unlock AI on the Edge and on Mobile?

Jev can Generate the Best Option

Could Jev Accelerate Automated Advertising?

How about HR and Recruiting?

Could Jev accelerate Computer Use and take it to the next level?

Subscribe now

How about Design?

And classifying Research papers?

Share

As of today September 22nd, Typesafe AI’s Jev has basically been overwhelmed with signups, they’ve literally had to pause them. Suffice to say it was a successful launch so far.

Read more

Good Evening,

Visions of AI is a new feature format I’m experimenting with that will amount to a short profile on an AI or emerging tech startup. The cadence of this style of article is unknown as of yet, but there are a lot of fascinating startups I want to discuss and share about.

This is designed to be light evening reading to go out at a time-slot of 8 pm EST.

In late 2026 there’s been a very sudden spike in massive Physical AI and embodied AI funding rounds, especially in the software around the robotic brain, and humanoid robotics, a form factor of bipedal (human-like) robots. While it remains to be seen how significant or enduring this trend is, it follows a flurry of Chinese robotics companies going public. China is widely seen as the robotics leader of the world, especially in current sales, hardware and manufacturing.

About three weeks ago a Physical AI startup called Skild AI released their S1, an in-context learning model for robotics. Skild AI has hit over $100 million1 in revenue already in 2026. The company claims that S1 new is a foundation model that learns from one example (one-shot learning). They said that it can be taught 10-minute long tasks that it has never seen before, from one video prompt without any fine-tuning.

Share

Embodied AI with a Real Revenue Ramp (Skild)

After doing some digging while it’s a marketing idea (being pushed by the industry), I do believe S1 could qualify as the GPT-3 moment for robotics learning, at least in its 2027 iterations. In the history of AI, that would correspond to around 2020. Generative Pre-trained Transformer 3 (GPT-3) was a landmark 175-billion parameter LLM developed by OpenAI where the research paper introducing it was published on May 20th, 2020, about 3.5 years before the ChatGPT moment itself.

This is tremendously exciting for robotics and Physical AI enthusiasts.

Subscribe now

Read more

Good Morning,

It’s been a while since we’ve brought you an AI Guide, see the rest here.

The consumer AI landscape around personal agents is going to be heating up in 2027. I always suspected Anysphere’s Cursor product was going to be a problem for OpenAI. On August 14th, SpaceX completed its acquisition of Cursor for $60 Billion. With Meta pushing products like Muse in B2C, OpenAI’s ChatGPT and its features might be tested on all sides. Even as Chinese model builders test OpenAI’s models on the token share angle. With a crisis of competition and alignment OpenAI has decided to delay its IPO till sometime next year.

Suddenly in the later quarter of 2026 SpaceX AI appears to be gaining some traction with Grok Bot. Grok Bot is an autonomous AI agent platform developed in collaboration with Cursor. It is included directly as part of paid Cursor subscription plans (such as Cursor Pro and Teams).

Subscribe now

Grok Bot and Muse Could Erode ChatGPT Usage in the next 18 Months

Grok Bot appears to be taking a niche that would have gone to OpenAI’s ChatGPT Work or ChatGPT Operator. While it’s too soon to say in September, 2026, it’s quite possible Meta Muse and SpaceX AI’s Grok Bot might be products that can take some customers and their time that would have been spent in ChatGPT previously.

I asked for a no nonsense guide on Grok Bot. Jeff is a software professional and writer of The AI-Augmented Engineer to teach people how to use AI to write better code, ship faster, and level up their careers. He writes weekly deep guides on using tools like Claude Code, Cursor, Codex, and more to build awesome apps.

Meta's standalone personal AI agent after Billions spent on compute and talent is called Muse and it launched on September 8, 2026, and early indicators show rapid adoption too. Will OpenAI build a “Codex Bot” to compete with Grok Bot? How will they take on Muse? Is ChatGPT Work going to be competitive with Claude Code? OpenAI have too many products and outside of ChatGPT and Codex, not many are well differentiated or have a defined brand. Anthropic’s Claude Cowork is somewhat similar to Grok Bot in that it’s a bit more accessible. Overall since ChatGPT there really hasn’t been a killer B2C AI product, outside of Google Gemini. Now with Gemini 3.8 with Extended Thinking. In a world of multi-agent collaboration, the way we use AI might shift a lot in the next 18 months.

Share

Autonomous Teammate Systems

For me, Meta’s Muse and SpaceX’s Grok Bot make chatbots a bit less relevant in that these are about AI as a utility, helper and assistant in accomplishing tasks. They feature agentic capabilities that have:

• Isolated Cloud Machines

• Tools with asynchronous & background execution

• Human-in-the-loop safeguards

• Proactive goal orientated assistants

Grok Bot is aided by X user integration and so even if the Grok chatbot was a relative failure compared to ChatGPT, Gemini and Claude - Grok Bot could accelerate SpaceX AI’s Enterprise adoption and Cursor’s real world utility. It remains to be seen what Anthropic and OpenAI come up with in this vertical. Grok Bot was first introduced about a month ago on August 11th that sounds a lot like Meta’s Muse, branded as an AI teammates you can give real work to.

Grok Bot also seems more like a primarily B2C product but with strong B2B support. We’re still seemingly a far way away from the kind of AI companion that is truly capable even four years after ChatGPT first launched which is astounding to me.

Check out the easy to read and detailed guide by Jeff below.

The AI-Augmented Engineer

Teaching software engineers how to accelerate their careers with AI. Deep, technical guides on Claude Code, Codex, Cursor, and more.

By Jeff Morhous

How to use Grok Bot as a personal assistant

Read more

My Friends,

The datacenter State capital of the U.S., Texas, is going dark. It paused Datacenter approvals way back last month on August 3rd and is now putting the brakes on them. BigAI featuring OpenAI and Anthropic were already compute constrained, and now suddenly as if to rescue us (on X), a CEO who has the biggest AI IPO in history in 7 weeks is mysteriously calling for a frontier model slowdown, with a manifesto blog entitled We Must Pace the Frontier.

Ah, I think the American people already decided that one for you. So are we all doomed, or something? The AI news aura got a bit wacky over the weekend and I’m curious to see how markets respond this week.

P(doom) is the probability of existentially catastrophic outcomes (so-called "doomsday scenarios") as a result of artificial intelligence and its uncertain future impacts. If you turn AI trust & saftey into a marketing stunt, what exactly should we expect from you in the near future?

"No other human activity poses this level of danger." - Jacob Coxon

As more Anthropic employees leave the company commenting on Coxon’s Twitter/X remarks, we have to cover the story a bit deeper. Now there has been another announcement by a former Anthropic employee, Joe Benton, the former manager of Anthropic’s Scalable Oversight.

• Humanity are building machine that are much smarter than any human, and we may not survive this, he tweeted. While he’s going to work at METR a company Anthropic and OpenAI are both nomating to oversee some of their work as frontier evaluators. METR stands for “Model Evaluation and Threat Research”, claiming to be an independent non-profit research institute that evaluates frontier AI models for complex, long-horizon, and potentially dangerous agentic capabilities. What an incredible coincidence!

Share

Suddenly two of the lesser rival AI firms (with lesser models and less ARR), led by tycoons Sam Altman of OpenAI and Elon Musk of SpaceX AI, have both said they agree with Amodei frontier slowdown thesis. Anthropic’s anti-China CEO cited the OpenAI-Hugging Face incident (OAI-HF) and the risks of recursive-self-improvement (RSI) improving beyond our ability to control. So how long till the end-times exactly1? X-Risk chatter is distracting and masking from other macro events that threaten the AI boom.

“One of the most worrying risks linked to frontier AI is extreme power concentration.” - François Chollet

The AI slowdown appears to be a marketing theme that suddenly emerged on X this weekend. The scripted sequence of events was supposed to introduce the public and AI researchers to why a frontier slowdown or pacing of model releases is a good idea and how their chosen “evaluators” could safeguard model releases in the future. Approach with extreme caution and skepticism.

A “Scary Good” Slowdown into a $2 Trillion IPO

In 2026 it’s hard to imagine how super-intelligence leads to our extinction as a species. But don’t argue, P(doom) is trending and I’m guessing Anthropic wouldn’t have it any other way. Anthropic is telling investors that they will be profitable for a second straight quarter with gross margins are above 80% before accounting ​for revenue shared ​with distribution partners, including Amazon and covering the cost of training.

Being compute constrained, while open-weight models take token marketshare while the general public show a marked decline of sentiment towards AI is not a great macro position for BigAI labs to be in in 2026.

There are so many AI risks to democracy, capitalisms and society2 not being talked about when x-risk takes the center of the stage.

What would AI at the end of civilization actually look like?

• AI being used in a global nucleur war

• AI being used to create a catastrophic biological weaponized virus

• Humans weaponizing extremely powerful AI in unpredictable ways is inevitable

• Humans losing control of super-intelligent AI that has goals that conflict with human civilization is now possible in the foreseeable future

Subscribe now

American Protectionism Rising in AI

Read more

Good Morning,

I had originally wanted to do a deep dive into Meta’s Muse launch. However a couple of things stopped me in my tracks surveying the AI landscape this …

Read more

Good Morning,

Mistral appears triumphant having just raised a massive 3 Billion Euros ($3.5 billion) in funding at a 21 billion Euro valuation. It’s great on paper, it’s the largest equity fundraising round ever completed by a European technology company, three years after the company's launch. Sounds historic right? But the reality tells a different story.

Europe’s Mistral has not been able to compete favorably with either China or the U.S. in open-source AI models. Mistral has huge investors like ASML, Samsung and Nvidia. Behind this huge move is a business pivot more into AI infrastructure, datacenters and sovereign compute just as the likes of SpaceX and Google invest more in Energy behind the compute binge.

By pushing heavily into Mistral Compute and transforming into a full-stack neocloud player Mistral will now also compete with the likes of Nebius and Nscale, European Neo-Cloud leaders in Sovereign compute. This is a tell of where Europe can compete in Sovereign AI and where it is overmatched.

How optimistic should we be for Europe’s ability to remain a sovereign AI leader and provider in its own right in 2026 and moving forwards? For the first time, a non-European entity is leading a funding round of Mistral, in this behemoth of a Series D. Does this begin a coalition of middle powers like Canada, South Korea and others (e.g. Australia, Japan, Brazil) moving in the direction of the teaming up with the EU? How do countries move forward in an AI future so dominated by U.S. and China?

Sovereign AI could become an extinct dream without proper support.

In a world of increased centralization and National self-interests, I ask you:

All startups have to pivot with reality and where we are in the AI boom cycle. The 2023 founded Mistral’s strategy shift is a sign of the times.

Subscribe now

Read more