The Open Source AI China Problem Just got Worse

Model supremacy in a token-efficient macro environment plagued by HBM, energy and datacenter compute bottlenecks. The 2026 story of AI is getting geopolitical.

👋 Hey there, I’m Mike. Each week I share AI articles at the intersection of tech, business, society and the future. If you want to support the channel or gain full-access to my work, go here. Read Archives | See Substack Notes | Visit our community Chat | Visit Homepage. I’ve been tracking the U.S. vs. China dynamics of AI and its future for nearly six years.

How do you summarize the heat that is July, 2026 in the AI industry? It’s been a very bizarre and multi-layered drama. We are witnessing history.

Geopolitics and AI on the Front Burner 🔥

As you likely know, Chinese company Moonshot AI released a new version of its Kimi model called Kimi K3 that is causing a lot of Enterprise AI to switch to open-weight models. It might be one of the biggest AI moments of 2026. With the Iran war unresolved in the strait of Hormuz and the Tech heavy NASDAQ 100 in freefall, it’s becoming a geopolitical and Trump Administration catch-22. There are no clear solutions in war and AI, as it turns out.

American geopolitics have rattled markets once again. An extremely unpopular and costly war.

Generative AI models are evolving, but likely slowing the revenue growth of AI behemoths OpenAI and Anthropic. With American hyperscalers approaching negative free cash flow via incredible AI capex and datacenter investments, you have to wonder whether it’s all worth it - if Chinese models that are getting larger and more efficient and can replicate the performance while under cutting the cost. It has the potential to become an AI crisis in the stock market even as the Semiconductor boom seems to have hit a bear market correction, after the U.S. listing of South Korean HBM leader, SK Hynix. South Korea (the KOSPI) is now a leading indicator.

With Chinese DRAM maker CXMT about to go public in Shanghai, it’s a very charged China vs. U.S. setup in the future of AI. With DeepSeek’s stunning funding rounds and planned IPOs for DeepSeek, Moonshot, OpenAI and others in 2027, it’s shaping up to be quite a year next year too. DeepSeek raised around $7.4 billion in June last month. We have to assume Databricks has also been one of the beneficiaries of this pivot of Enterprise AI to routing and cheaper tokens of open-weight players in a world where Databricks announced a new round of funding that values the company at $188 billion.

Google’s Gemini 3.5 Pro has been crucially delayed at the worst possible moment. The launch of SpaceXAI’s new model Grok 4.5 was completely overwhelmed by the Kimi K3 moment. Anthropic Fable 5 confusion has given China the ultimate return of that DeepSeek moment vibe back from the dead of January, 2025. OpenAI’s own GPT 5.6 Sol release has also been nearly entirely overshadowed.

The Trump Administration quick to restrict Mythos class models has a serious problem, what if Chinese models are able to approach those same capabilities with open-weight models that were thought to be many more months behind? The Trump administration is showing signs it could ban Chinese cutting-edge models and take drastic steps to try to curtail China’s rise in AI. I thought they were not going to regulate the AI industry. So much for global free market capitalism.

The Brave New World of Token Efficiency Looks Chinese

With Microsoft, Amazon, Meta and even Google behind the Big 7 hyperscalers have lost some of their AI talking points and credibility in this cycle. The mishandling of Mythos class models by the Government and the rise of Kimi K3 type models is the perfect storm that is seeing many Enterprise companies pivot to more rational token usage that’s radically more cost efficient.

While Open-Router isn’t representative of the entire situation, it’s an interesting data sampling point of the overall macro trend: it appears like cheaper open-weight models from China are winning over marketshare.

The era of prompting has given way to the reality of routing and token efficiency.

In my opinion, some of the best Open-weight models from the U.S. are Thinking Machine’s Inkling (about a week old), and whatever Reflection is likely to put out soon. Nvidia’s Nemotron is an often cited alternative to the previous leadership by Meta. The problem is the real lack of leadership in open-source AI in the United States. For America this is a potential disaster in its AI leadership on the frontier of models, tokens and Enterprise adoption.

To make matters even more peculiar, Chinese President Xi Jinping’s most significant recent remarks on artificial intelligence were delivered via a keynote address at the World Artificial Intelligence Conference (WAIC). President Xi Jinping attended in Shanghai the opening ceremony of WAIC 2026 and delivered a keynote speech titled “Joining Hands to Build a Just and Equitable System for Global AI Governance.” China appears to be more advanced in AI governance and regulatory leadership than the U.S. trying to actively build global collaboration around the theme.

Alibaba’s own Qwen 3.8 Max (preview) will also have an open-weight component with an aggressive international token pricing plan. Kimi K3 is not aimed at hobbyist developers but at Enterprise customers in order to ramp ARR before they also IPO in about six months time. The U.S. and China competition in models, even at a time when there are few great application layer products is exaggerating the demand for compute at a time when China has both cheaper token generation and more abundance energy. At a time when the Trump Administration’s key mandate appears to be keeping the AI boom on the stock market rolling for the financial elite and business class. Meanwhile everyone from Anthropic to Moonshot AI are positioning themselves to maximize their IPO hype and revenue sales momentum.

Do Ranking Models even Matter Any Longer?

If you go by Artificial Analysis Intelligence Index, and rank by smartest model, the ranking is:

The state of affairs on peak model performance is roughly as follows:

1. Anthropic – Claude Fable 5

2. OpenAI – GPT 5.6 Sol

3. Moonshot AI – Kimi K3 (open weights*)

4. SpaceXAI – Grok 4.5

5. Zhipu (Z.ai) – GLM 5.2 (open weights)

These lists and benchmarks that they models are trained to perform on are fairly artificial and likely to change next week and certainly by next month.

Is comparing LLMs the right way to measure AI progress any longer? Artificial Analysis

The “AI Disconnect” is Getting Worse 🔎

Chinese open-weight models have been cheaper but in 2026 they are becoming way more capable. This means U.S. companies are starting to adopt them for everyday tasks to reduce their token budgets that got out of hand with Fable 5 & Mythos 5 level pricing. If every new Chinese models has the potential to debut in the top five best models by intelligence, that’s a big problem for the AI industrial complex of the U.S. The China problem in AI for the United States isn’t just there, it’s getting worse.

I have a huge respect for the founders of DeepSeek, Moonshot AI and Zhipu AI, because you sense they have real idealism and not just incredible AI talent and business smarts. The capabilities of their models show sophisticated innovations and incredible AI business acumen given their compute and funding constraints. You don’t get the same feeling from the executives at Meta, Google or Microsoft, there’s a pretty stark disconnect. Relative to BigTech’s Capex it’s starting to look really bad. The AI execution of BigTech in Generative AI models (and product) has been surprising bad.

The past week this might have been best confessed by OpenAI’s (new) head of strategic futures Dean Ball on a viral Tweet on X. Where he said that Open-weight models are inherently decelerationist and mused that an open-weight-model-dominant world could lead to full on AI communism that could lead to a dystopian hellscape.

The fear mongering around Open-source AI reached a new level in July, 2026. The China argument is getting tired.

It’s clear the closed-model capital intensive companies are concerned about these recent developments with claims of distillation and cybersecurity concerns full of anti-china sentiment and national defense arguments. While Anthropic’s Mythos class models appear on paper impressive, we don’t know their most advanced capabilities because they have been restricted from the general public. The lines between AI future strategists and lobbyists in America looks fairly blurry. Just as the rational for Capex in the Enterprise AI price-war on tokens looks fairly poor from an ROI perspective.

BigTech’s Capex Crusade is looking grim for 2027. Bank of America.

What we know about American protectionism is what they can’t compete with on the global market, they will definately try to ban and restrict at home. But the pivot of Enterprise companies to open-weight models has already been the dominant story of 2026, it might even be too late and by doing so they might set up an open-source AI race for U.S. companies as well in doing so. That’s the catch-22 of open-source AI’s peculiar persuasion on cost in 2026. This wouldn’t be a problem if the Generative AI technology was actually giving reliable ROI, but it’s a nascent technology and not many great AI products exist yet to leverage these incredible models.

The U.S. Administration's Commerce Department last year considered adding multiple Chinese AI labs to its "Entity List” and I’m sure that’s now again on the table as American companies struggle to compete in a world of a rapid token efficiency ramp where harnesses and routing matters more than ever and so does the spiraling costs of the tokens and which models to use for which tasks.

The U.S. Administration restricting their best closed-source company in Anthropic’s best models might have been a historic mistake that has allowed this painful situation to occur. The Kimi K3 moment is more dramatic for the comedy of errors just mere months from Anthropic’s own splashy AI IPO where Semianlaysis publication has analyzed Anthropic’s incredible operating profits. You have to assume the pivot to open-weight models and the Mythos hijacking by the Administration has slowed down America’s top company in AI. Semianalysis believe Anthropic is projected to reach over $1 billion in GAAP EBIT by Q3 2026 (a ~6% profit margin), making it one of the first major frontier AI labs to achieve sustained quarterly profitability. All of these recent events positions OpenAI, Meta and SpaceXAI as among the biggest losers. This even as Meta’s Muse Spark 1.1 isn’t such at terrible model.

What If?

If the Trump Administration restricted Chinese open-weight model Nvidia, Thinking Machines, Meta and Reflection (yet to release a model) could be among the big winners for U.S. based Open-weight token solutions for Enterprise customers.

The mid 2026 Macro AI narrative has become far more interesting. The levels of customization in open-source AI are being upgraded. Token efficiency and routing has become a bigger deal. AI Capex in the markets are coming under increased scrutiny.

A Story of Many DeepSeek Moments

Open-weight model AI Supremacy isn’t AI communism or a Nuclear threat, cheaper tokens benefits Jevons Paradox and how companies, developers and consumers can do with AI. The entire Generative AI model training paradigm has been a multi-string distillation hijack of the world’s language based data for the last five years. The demand for compute isn’t slowing because Enterprises choose cheaper models, it will actually accelerate it. That’s the catch-22 of the entire macro AI situation.

America is a troubled nation if capital and monopoly capitalism is their moat. An AI future will require new kinds of innovation and not just a better Venture Capital system. There’s obviously some narrative abuse going on here and business consolidation and geopolitics will sort itself out like usual. With DeepSeek raising more funds and then rushing to IPO, the DeepSeek moments keep piling up. Their flagship model DeepSeek-R2 was never released and it begs the question to China’s most advanced models in the future. Liang Wenfeng has a 78% to 84% stake in DeepSeek and they are also developing their own chip.

The Race to IPO in the Roaring 20s

A world where both Anthropic and Moonshot AI are compute capacity constrained, Google is an LLM laggard again, and Meta and SpaceXAI are still on the margins. A world where OpenAI’s IPO looks less appealing by the quarter. A world where we are tired of waiting for the IPOs of Databricks, Crusoe, Anduril, Stripe and others even as Chinese Physical AI, AI chip and Memory giants rush to go public. It’s claimed that Chinese chipmaker CXMT Corp's $8.6 billion initial public offering was more than 500 times oversubscribed by institutional investors.

I predict 2027 will feature elements of a ChatGPT moment for robotics (now also called Physical AI), as Anthropic is rumored to be in talks to acquire Physical Intelligence. Even as AI pioneer Yann LeCun warns of how incapable these humanoid robots are and will be for quite some time. Silicon Valley and Wall Street will be so desperate for new narratives to hold the AI boom stock market up, they might stretch the truth again.

This at a time when there’s a stark lack of pure-play robotics' or robotics-software companies even public on the U.S. market. China struggling to recover from an epic housing crash seems to be going first in the IPO race. As impressive as Moonshot AI, Zhipu, Minimax or the smaller Chinsee AI labs are, it’s the macro momentum of the Open-source vs. Closed-source debate that has me enthralled. In many ways it’s a battle of capital vs. talent.

“This future strikes me as a dystopian hellscape, but I’ve never met an open-weight models advocate who doesn’t ultimately concede this is where things end,” said Dean Ball or ironically, OpenAI.

With BigTech earnings on deck this week, we’re going to get a few more answers. The capex to ROI story has never looked so dire. Any failure to execute on Earnings is going to get punished by this market that has in recent weeks seen some valuation revisions on growth names where SpaceX is down 25.5% and Grok 4.5 was yet again not good enough to move the needle (the bar is high). With an expensive acquisition of Cursor, SpaceXAI also appears to be still regulated to be among the AI losers.

Moonshot AI launched Kimi K3 on July 16, a massive 2.8 trillion-parameter model that matches top U.S. systems like Claude and GPT on benchmarks, with plans for full open weights by July 27 to enable cheap self-hosting.

By 2027 we’ll know how strong Anthropic’s IPO could be with its fastest ARR growth in software tech history being challenged by macro AI headwinds and Government intervention. Will American protectionism shield it, or will the market pick the winners? The tension is going to spill over into a robotics race that will encompass Space, AI and National defense like never before.

GPT Moment for Robotics is Near

If Anthropic really does acquire Physical Intelligence, the GPT moment for robotics I’ve spoken about multiple times could be near. The Chinese IPO of Uniree is already a huge moment for the future of Physical AI. It would raise 4.2 billion yuan ($619.4 million) while it has aggressively drove down the price of entry-level quadruped robots by over 90% in recent times.

Since China is robotics native, this part of the AI race is something the United States is frantically trying to make up ground in with several huge VC funding rounds in 2026 in the space. The U.S. appears to have a lead in the world-model and software brain startups implicated in making robots more useful while China has way more humanoid (robots that have a human form) startups and experiments in drones and new kinds of hardware form-factors for robots.

Unitree should IPO later this month or roughly three months before the pivotal IPO of Anthropic. Unitree is rapidly becoming one of the key de facto standard hardware platforms for AI researchers and robotics labs globally with the humanoid robotics race likely to ramp up so quickly Tesla will have to merge with SpaceX in the near future. In mind this happens likely in early 2028. If the Generative AI boom falls from the heights (like many analysts expect it will) I think we can expect the robotics race to take its place as the next big American technology narrative.

Ultimately the China vs. American competition in AI is good for global developers, consumers and businesses for more real-world utility. Whether Generative AI models or robots will bring much real world value anytime soon is another question entirely compared to the investment. The accelerated timelines of startups with drastically higher funding rounds is notable even as the urgency to increase AI business adoption increases. The heat is on and it’s going to get worse in 2027 with a tremendous competition setting up for the 2030s in technology. China’s energy abundance and American semiconductor supremacy keeps things interesting.