The Pentagon just went all in on AI
Pentagon Awards Up to $200 Million in Contracts to xAI, OpenAI
On July 14th, 2025, the U.S. Department of Defense (DoD), through its Chief Digital and Artificial Intelligence Office, awarded contracts of up to $200 million each to four leading AI firms: Anthropic, Google, OpenAI, and xAI (Elon Musk’s company).
But What is the CDAO?
The CDAO’s mission is to accelerate DoD adoption of data, analytics, and AI from the boardroom to the battlefield. CDAO exercised its organic acquisition authority to issue the awards announced today, demonstrating that DoD acquisition can move at the speed of emerging technology and operational necessity. The head of the CDAO is currently Douglas Matty. His title is “Chief Digital and Artificial Intelligence Officer.”
This pushes the AI race into the arena of military automation and Generative AI tools and sets a very dangerous precedent. The National Defense community is seeking to push agentic AI from lab to battlefield, only weeks after giving OpenAI the same mandate.
These two-year prototype deals aim to rapidly develop and deploy agentic AI workflows, large language models, and related technologies to address both warfighting and enterprise needs.
The total ceiling for these contracts is $800 million, signaling a “commercial-first” pivot for Pentagon technology acquisition, following the recommendations of Task Force Lima’s 2024 report and the DoD’s FY-26 budget emphasis on AI and autonomy funds.
The contracts represent a significant revenue stream for AI companies, which have been looking to expand their reach into the public sector. This comes after lucrative contracts to the likes of Anduril, Scale AI and Palantir, companies that have ties to Peter Thiel (Founders Fund) and his friends in the Trump Administration. This momentum will be wildly accelerated when the Trump Administration leverages a U.S. Sovereign Wealth Fund. The concentration of Pentagon influence by BigTech’s and powerful VCs and their active participation ushers in a much different world both for the U.S., its allies and its potential adversaries. Also for an increasingly fragmented world of geopolitical trade wars and tensions. This is leading to a golden age of National defense startups being formed in the united states.
Thunderforge, Maven, Friends and War Prep with AI while gutting Science
In April 2025, budget planning documents within the Trump administration indicated the intention to reduce funding for a variety of science agencies in Fiscal Year 2026, which begins on October 1, 2025: Centers for Disease Control and Prevention: reduce funding from $9.2 billion to $5.2 billion, a 44% cut. Trump plans to gut funding for Weather science and natural disaster responses, as well for Space like NASA. These cuts will include gutting the Education Department. Globally, Nato leaders have agreed to ramp up defence spending to 5% of their countries' economic output by 2035, following months of pressure from President Trump.
Back in May, 2025 the Pentagon raised its contract ceiling for Palantir's Maven Smart System to $1.3 billion through 2029, up from $480 million, marking a $795 million vote of confidence in AI-led battlefield operations. The Maven Smart System (MSS), also known as Maven, is an artificial intelligence (AI) and machine learning (ML) tool developed by the U.S. Department of Defense to process imagery and video from drones. It combines sensor data, AI, and ML to modernize battlefield operations, including targeting, logistics planning, and predicting supply needs for deployed troops. Essentially, Maven helps analyze large amounts of data from various sources to create a common operating picture for decision-makers.
In March, 2025 it was revealed that Scale AI has partnered with the Department of Defense to use AI agents for U.S. military planning and operations.
“Thunderforge” is the DOD’s flagship program and will work with Anduril, Microsoft and others to develop and deploy AI agents.
Global military expenditure reached a record high of $2.46 trillion in 2024, marking a significant increase from previous years, with a 7.4% real-terms growth. However in 2025 we’re seeing an unprecedented rise in military spending. It’s not clear what kind of a world that will lead to when more conflicts emerge. The weaponization of AI is well underway. What will AI Supremacy in warfare and national defense even look like in the future?
Today’s Deep Dive
Thus I’ve been reading a lot more into National Defense and the impact of AI. This space is often very related to U.S. Venture Capital moves. I asked
for his take on the impact of AI at the Pentagon that is the deep dive of today.Based out of Boston, Austin Gray is Co-Founder and Chief Strategy Officer of Blue Water Autonomy, a VC-backed defense-first shipbuilder, where he leads customer engagement. Austin previously worked in product development at a Ukrainian drone factory and served in the U.S. Navy as an intelligence officer. He holds degrees from Davidson College, MIT, and Harvard.
If you want to learn more about the intersection of AI and National defense, 2025 would be the ideal year to start:
The U.S. national defense budget has increased. For fiscal year 2026, the proposed budget includes $1.01 trillion, a 13% increase compared to the enacted budget for fiscal year 2025.
Moving forward we’ll be exploring and reporting on the intersection of AI and National Defense with greater frequency on AI Supremacy Newsletter. The acceleration of datacenters is what amounts to a national project by the United States to achieve AI hegemony. In such a world, Venture Capital, Government, Energy and Space becoming more intertwined with aspects of National Defense with rising capabilities in the automation of the Military, new kinds of drones, swarms, robots, Quantum capabilities, surveillance and cybersecurity tech.
Not Those Shades of Gray
⚓ Best of Shades of Gray
Defense M&A and the Deal Flow Dinner
Startups & Sea Power
Software-Defined Warships: The Navy’s Digital Future of Necessity
Red’s Hellscape: Far More Dangerous Than Blue’s
From Powerpoint to Payload: All the Ways AI & Automation Are Disrupting the Pentagon
From Powerpoint to Payload
All the Ways AI & Automation Must Disrupt the Pentagon
’s Newsletter is one of the best publications for understanding the Venture Capital dynamics of the National Defense industry on the planet.
From the muddy trenches of Ukraine, to an aircraft carrier’s crisply air conditioned command center, to the Pentagon’s paneled conference rooms, militaries have the same problem as civilian office workers: workflow. And workflow needs technology.
Lots of the attention focuses on killer robots and fully-autonomous weapons systems. This attention is well-meaning, but it distracts from the true tech transformation underway in armies and navies around the world. AI and automation are eating the business of military power - just as software has been famously “eating the world” for quite some time - and the business of military power is booming.
I’m a defense tech founder building fully autonomous ships for the Pentagon, so I wake up thinking about the military’s tech needs every day. My product is certainly on the shiny end of ambitious, futuristic robots. However, far more often than I see robots, I see the bureaucratic reality of the Pentagon’s business systems - systems that need automation and AI even more than the front-line troops do.
This piece will break down all the ways automation and AI are transforming warfare - with just as much attention to warfare-essential workflows as to killer robots - and explore some of the new companies delivering them.
Example 1: The Colonel’s Favorite Workflow - Faster Warheads on Foreheads
Example 2: The Admiral’s Least Favorite Workflow - Military HR
So How Big is This Market?
Services, Agencies, and Budgets - The DoD Org Lens
Tech Verticals and Companies Attacking Them - The Tech Lens
Adjacent Markets - The B2B Lens
What Next?
Analogous Markets
History is Happening
The Colonel’s Favorite Workflow: The Targeting Cycle
Putting warheads on foreheads fast is a holy grail of military process. Christian Brose, Anduril’s Chief Strategy Officer, titled his book Kill Chain because commanders obsess over speeding up the kill chain. The kill chain - today it’s thought of more as a kill web because mesh networks don’t have single points of failure like a string of christmas lights - is made up of sensors, shooters, networks connecting them, and platforms carrying all three.
The perfect kill web will have many sensors spread out in front of its forces, allowing them to know what to shoot at, when, and to strike the adversary with the right weapons. Because the adversary has his own kill web, a targeting workflow that moves decision data faster than the adversary’s is essential. I wrote a fun piece about kill web vs kill web matchups, which are happening today in real life
When Russia invaded Ukraine in 2022, ubiquitous smart phones meant that the defender’s kill web was ready. “We are like a hive of bees,” said one soldier, who led a reconnaissance team that helped stop a massive column of tanks driving toward Kyiv as the war opened. Ukrainians didn’t have the armor to meet Moscow’s column head-to-head, but they had phones, drones, and satellite imagery. This tech made their home turf transparent.
To strike, they started with artillery and shoulder-mounted anti-tank rockets, and quickly were using drones to strike too. Most important: their kill chain was fast because they used software called Delta, a live map developed by NGO Aerorozvidka, which allowed them to fuse together everything the humans and drones were seeing, and strike.
By the time I got to Ukraine in 2023 drones and live maps were scaled across the force and artillery targeting times were falling from minutes to fractions of a minute. Not only were targeting timelines tightening on both sides, but the battlefield was expanding. Visiting a new drone factory one day, I realized that the drones I was seeing assembled had the range to hit Moscow. Soon, they did. Today, strikes deep into Russia are routine. The US military is now testing these same long-range strike drones.
Ukraine shows that tech, from map software to new flying platforms, expands and accelerates targeting on the battlefield. When I joined the US Navy nearly a decade ago as an intelligence officer, I had to memorize a series of boring process flow charts like these:
When I deployed on “high tech” aircraft carriers, most of the steps in these workflows were achieved via copy / paste.
But Pentagon officers have long dreamed of the process looking more like this, with satellites instantly relaying targeting data between fighter jets and warships to relentlessly strike the adversary:
These workflows are about moving data. Some are more tactical, while others more strategic. But all of them are about making decisions and all of them are ripe for automation. Warfare’s most important workflow, the targeting cycle, is now driven by software, AI, and automation. Real battlefields around the world already have the elusive “CJADC2” that the Pentagon has long sought. And it’s just getting started.
The Admiral’s Least Favorite Workflow
For all the sexiness of timely targeting, most military work is deeply bureaucratic.
When I onboard a new employee at my company, I enter her name, email, and salary information into Rippling, my HR software. Rippling sends her a welcome email with a link, and she plugs in her banking information, t-shirt size, home address, and other details. Automatically, she’s set up with IT accounts, healthcare, payroll, and company swag. Her laptop is mailed to her home before she even starts.
In the military, there is little automation for administrative work like this and - especially every 2-3 years when you join a new unit - soldiers spend weeks per year doing paperwork for each function. Creating a new IT account takes weeks, sometimes months, with many approvers in the process chain. At my first job, doing IT for the Navy, we tried fixing this process. Like those who came before us, and those since, we failed.
When I returned from Ukraine in 2023, very motivated, I signed back up for the Navy reserves. It took 12 different steps over 48 weeks to activate a military aged male - activate not enroll, since I was already in the reserves - with zero health conditions and a full top secret clearance. Today in 2025 I have still not drilled a single day.
IT and HR are the easy parts. In this piece, we’ll look at how automation and AI are starting to modernize these back office workflows. On top of these mundane opportunities, the Pentagon needs software companies to help with much more challenging internal business processes. Today, budget mandarins and acquisition tzars fight - with paper, emails, committees, and sometimes their bare hands - through layers of process to procure the tech transforming the battlefield.
How Do Military Services and Organizations Spend on AI & Automation?
My friend Maggie wrote an excellent primer explaining how the DoD is adopting AI: Follow the Money: What the Pentagon's Budget Data Tells Us About AI and Autonomy Adoption. Read the sister publication:
I have four key takeaways from that piece.
Let’s break down each takeaway with a chart.
Different parts of DoD (e.g. Army vs. Navy vs. obscure agencies) spend on AI and automation very differently
The Air Force (which includes the Space Force) is way out ahead of the other services. Additionally, all the blue means it’s procuring more permanent tech, instead of spinning on continued R&D. The Navy, on the other hand, spend slightly more than the Army on this tech, but only a tiny portion ~$80M is procurement of AI, suggesting the rest is still R&D.
Tech-Forward offices that circumvent the military services are the main buyers of AI & autonomy
The military services are supposed to fill out the man-train-equip mission for the Pentagon, providing forces for the combatant commanders deployed around the world, but “Fourth Estate” agencies end up doing lots of the heavy lift to equip troops with tech. This graphic shows the “Fourth Estate” agencies, which are more tech-forward than the services. Much of the “Other OSD” codified above is the Pentagon’s Defense Innovation Unit, which has been leading the charge procuring tech.
More of the AI spend seems focused on weapons systems and combat-centric IT and less on internal business processes
Although there are myriad opportunities to automate systems like IT and HR, combat tech seems to be getting most of the shiny new tools. Many of the large capabilities on the list above - combat vehicles on ground, sea, and air, sensing systems for the Army and Space Force, and combat-focused IT efforts like the Advanced Battle Management System - are for tip of the spear capabilities. The Marine Corps Enterprise Network may represent the only back-end tech function with a serious funding commitment. This data suggest there is still massive opportunity in automating internal business processes.
Almost all the AI adoption is still in the research and development (RDT&E) phase, not the procurement phase.
Last, very little of the Pentagon’s AI spend comes as recurring procurement. Most of it is still research, and R&D spending outstrips procurement 4-1. This last datapoint suggests how much opportunity remains ahead, as tech verticals and customer conviction mature.
Tech Verticals Show Clear Need for AI & Automation
The last section looked at who within the government spends on AI & automation, and what colors of funding they use. Now, let’s look at what they actually buy and need.
To start, the DoD’s CTO has an annual report on all their science and tech (S&T) spending, broken down by tech areas. $5B of $17B, nearly a third, goes to Trusted AI & Autonomy, with additional smaller categories such as Advanced Computing & Software, Human-Machine Interface, and Integrated Sensing and Cyber - all of these are AI-intense.
More important than S&T spending - which is only a small portion of the DoD’s budget - is transition spending. The below chart shows that 30 Trusted AI & Autonomy projects transitioned to formal DoD programs, where they have recurring budget lines and safe homes for life. This makes AI a great investment category, since other tech categories like Space and Cyber did not transition to full programs at the same rate as AI & Autonomy R&D.
Now that we’ve looked at the actual budget numbers, I’ll review 8 categories of tech where the Pentagon needs companies to deliver using AI & Automation.
1. Autonomous and Uncrewed Systems
2. Command and Control (C2) / Decision Support
3. Cyber and Information Operations
4. Logistics and Sustainment
5. Training and Simulation
6. Intelligence, Surveillance, and Reconnaissance (ISR)
7. Warfighter Support and Human-Machine Teaming
8. Administrative and Back-Office Functions
1. Autonomous and Uncrewed Systems
Air: Drones for ISR (intelligence, surveillance, reconnaissance), strike missions (e.g. Loyal Wingman concepts). Check out the Collaborative Combat Aircraft.
Sea: Unmanned surface and undersea vehicles (USVs/UUVs) for mine countermeasures, ISR, anti-submarine warfare. Check out Anduril’s LDUUV or Blue Water’s unmanned ships.
Land: Robotic combat vehicles, logistics support robots, autonomous convoys. Companies like Overland AI and Forterra are making incredibly cool product here.
Space: Autonomous satellite navigation, maneuvering, and threat response.
Analysis. Competition in this space is fierce. Many startups compete in a single product / domain lane above. Most assemble low cost systems from commercial supply chains, but good ones have some sort of moat. Neros, for example, has a market moat, on the exclusive DIU Blue List. Firestorm, on the other end, has a tech moat, with 3D printed, rapidly assembled drones, and special sauce propulsion developed in house.
The major primes, from Raytheon and Lockheed to Anduril - they compete with products across land, air, sea, and space. Many startups are building small unmanned systems that are easily assembled from commercial components.
Last year, I wrote a deep dive on the size and shape of the unmanned systems market - check it out here.
2. Command and Control (C2) / Decision Support
Battlefield AI assistants: Recommending actions based on real-time data and simulations.
Sensor fusion: Merging inputs from various platforms (satellites, drones, ground units) for a unified picture.
Mosaic warfare / JADC2: AI-driven dynamic tasking and network-centric coordination across services.
Analysis: Making quick decisions about what’s happening on chaotic battlefields is very hard. This area is ripe for automation, but since command and control often gets centralized, sales cycles and customer access in this field will be tough for new companies. The Primes will hide behind thick moats, including the need for top secret IT system access, such as in this live Defense Innovation Unit solicitation.
3. Cyber and Information Operations
Threat detection and response: AI for intrusion detection, anomaly spotting, and automated countermeasures.
Deception and influence: Automating and detecting influence campaigns or disinformation.
Resilient communications: AI-enhanced adaptive networking under jamming or attack.
Analysis: Barriers to entry in cyber are notoriously low, so this is a great area for new companies. Although there are many cyber companies - personally I struggle to differentiate them - the information operations space is less saturated. Vannevar Labs has supposedly made some strides in this arena.
This tech area doesn’t need so many new companies, as much as it needs consolidation to create smaller number of more capable companies. Overhead costs in defense markets are high, so some M&A could help.
4. Logistics and Sustainment
Predictive maintenance: Machine learning to forecast part failures before they happen.
Autonomous supply chains: Automated warehousing, routing, and delivery (e.g., autonomous ground vehicles or drones).
Fuel and resupply optimization: AI to manage energy consumption and distribution in complex operations.
Analysis: I worry about contested logistics every day. In a China-US matchup, the economics of supplying a military force on the other side of the Pacific Ocean are terrible. Aircraft and ships have huge maintenance backlogs - they need every bit of smart predictive maintenance they can get. They also need smart, automated supply chains with less power point and less paper. Companies like Gallatin AI are trying to build a Fed-Ex for defense. This is hard work, and a ripe market.
5. Training and Simulation
AI-powered adversaries: Adaptive opponents in virtual training environments.
Personalized learning: Tailoring training to individual soldier performance.
Mission rehearsal: Real-time, AI-updated simulations for pre-mission planning.
Analysis: Nothing was cooler in the Navy than simulating a massive wargame, from an aircraft carrier, from a real battle station. Without the risk of burning to death, it was like a 100% realistic video game. AI and VR can make the US military’s strong training culture even stronger, and maintain our edge over dictatorships who don’t invest in their people.
One of the BFDs of 2025 in defense tech: Anduril and Meta are taking over the Army’s most important VR training contract. Much smaller startups like Orama are making adaptive bad guys to train against in VR or helping spies and hostage negotiators in scenario based training, like Delta AI.
6. Intelligence, Surveillance, and Reconnaissance (ISR)
Automated imagery analysis: Computer vision for drone or satellite imagery.
Signal intelligence processing: AI tools to sift through electronic and communication signals.
Multi-domain threat detection: Identifying patterns and anomalies across land, sea, air, cyber, and space.
Analysis: This is a massive space. Period. Agencies like the National Geospatial Intelligence Agency and the National Security Agency have insane amounts of visual and signals data to process. They’ve been at it for a long time - but every year there is more data. This is a great market, but a more mature one.
7. Warfighter Support and Human-Machine Teaming
Exoskeletons and wearables: Augmenting soldier strength, endurance, and health monitoring.
Voice and gesture control: For hands-free interaction with autonomous systems.
Cognitive load management: AI to reduce information overload during combat.
Analysis: Consumer wearables companies like Whoop are way ahead of the paper-heavy DoD, and have started working with the military. Because it takes lots of compliance work to alter tech that touches people, this space will likely continue to lag the consumer world, but will remain an important one. Fighter pilots wear $400k helmets to help them make good quick decisions in combat. Infantry need decision aids they can carry easily, like those made by Reveal Technologies.
8. Administrative and Back-Office Functions
Contract and procurement automation: AI to detect fraud, optimize acquisitions.
HR and readiness tracking: AI-enhanced personnel and training systems.
Financial forecasting: Budget prediction and risk identification.
Analysis: This is the most thankless area, but maybe the most essential. Procurement offices across all the services are undermanned, but are still drowning in months of process and piles of paperwork to buy drones, widgets, and weapons that warfighters need. DOGE cuts have accentuated the problem here, because civilian administrative workforces have been purged, but the more challenging process simplifications have lagged. So they have the same amount of work, but fewer people. Companies like Adyton have been working on the HR side - we need more heros in this arena.
Adjacent Markets, Speed, and The Way Ahead
In addition to the tech needed by the government to modernize the military, there are important adjacent markets we shouldn’t forget about. Traditional B2B services, like consulting, legal, engineering, and business development support, for companies building all the above tech will continue to bloom. Cybersecurity consultants to help get through the dreaded CMMC are already a must for every defense startup.
More innovative B2B plays are essential though. Second Front Systems hacked a back door into the Pentagon’s compliance process so that new companies could get Authorized to Operate on DoD networks through Second Front. Nooks is working on SCIFS as a service - a SCIF is a special room where you can access Top Secret information, and where you can have access to the classified internet. Because most B2G contract opportunities are classified - my research found that only 16% of the Navy’s ship and submarines related contracts can be bid on by companies without a classified facility - companies need access to SCIFs, facilities clearance, and more.
The most important type of companies in these adjacent markets: ones that can offer speed. Today, the market is still shaped by the legacy of the cold war, the process of an entrenched bureaucracy, and not yet by the urgency of a world at war.
Speed must come on the front-end, accelerating targeting cycles and moving data via automation so warfighters can make quick shoot-or-scoot decisions. And speed must come to big data processing, with heavy use of AI, so that commanders can process massive amounts of information to know if their enemy is deceiving them or where the next attack might come. Hopefully, it will come to the back-end too, so that logistics flow, men and women who want to help can mobilize, and companies can quickly click through the compliance moat that protects legacy cold warriors.
Editor’s Note
There are increasingly emerging voices who are great sources of information around National Defense, related Venture Capital details and how AI and Agentic AI will be integrated into some of the Pentagon’s systems and related military/geopolitics news and coverage of conflict zones including:
 | A guest post by
|
Thank you for the shout out!
Hard to believe that the Pentagon and government is becoming more forward thinking than most nontech S&P500 companies. I have yet to see this type of efficiency in the private sector.