Notes from Stanford DEFCON: Defense Tech Venture Capital

A choppy week. The S&P and Nasdaq were both about flat on the week as of Thursday’s close, but this masked a choppy week for the markets. The week started out strong, with tech shares (including Nvidia and AMD) driving the S&P close to record highs, but skidded midweek over rates (Fed minutes showed most officials expect another hike by year-end and the 10-year hit its highest level in more than 20 years) and geopolitical concerns (more tanker attacks in the Strait of Hormuz drove oil prices up). Our NPM Private Market Tracker*, which shows the average price performance of the 50 largest names in our internal NPM Price data, is up ~91% YTD versus the S&P 500/Nasdaq up ~13%/~17%. (Bloomberg, Wall Street Journal, NPM Daq)

We attended the Stanford DEFCON conference this week and our high level takeaways are below.

Strategy

• The DoD needs to think farther ahead on procurement. It has been buying 10-15 years behind the technology curve and it should be procuring 3-4 years ahead of the technology curve. One example is underground capabilities. The advent of increasingly capable drones is creating risk to above-ground assets  — for example, several US aircraft at a Jordanian air base were damaged in recent Iranian strikes (Reuters 9/10/26). Ukraine and Russia are investing heavily in underground warfare, creating tunnel networks, underground command posts drone production sites, and underground medical facilities (Modern War Institute at Westpoint, 8/12/26). The US has not been as focused on these capabilities to-date, but future wars could hinge more on them.
• Mobile first, mobile always. On a similar thread, there was a lot of focus on mobility at the conference given the ubiquity of ISR (intelligence, surveillance and reconnaissance). Assets will increasingly need to be able to be moved around to avoid being sitting targets or drones. We think this is one reason why companies such as Castelion (maker of hypersonic missiles) and Radiant (maker of nuclear microreactors) are developing their products to fit inside shipping containers.
• Prepare for asymmetric warfare. The US should not only consider fighting a war against an adversary that has the same or similar weapons, as it did in the Cold War. Instead, it’s imperative to think about what adversaries will do precisely if they do not have the same weapons, and then work on solutions for that asymmetric warfare.
• Russia, China, Iran and North Korea are cooperating in the current wars in Iran and Ukraine, even if they are not explicitly fighting side by side (though thousands of North Korean soldiers have fought for Russia in the Ukraine war, BBC 9/13/26). The US and its allies, including Taiwan, Europe and others, will need to cooperate as well in response.

Autonomy

• Land autonomy is an important area for investment, as last mile logistics in theaters such as the Pacific will still be important even if getting there relies on air and naval power. Ukraine is using autonomous ground robots to deliver food, and autonomous ground robots could be important ways to deliver food, medical supplies and ammunition without risking human life.
• Land autonomy is also hard to do – we noted in our “Will Build Turn to Buy” report that there are fewer competitors in the ground autonomy space than there appear to be in the air/drone space or the maritime autonomy sector, and this could be because of the relative difficulty. There are fewer obstacles to navigate in the air, for example, whereas an autonomous vehicle on the ground needs to navigate a wide array of obstacles such as rocks, trees and ditches.
• The orchestration layer of autonomy is critical to get right. Autonomous systems operate in three basic layers: (1) the vehicle autonomy, which allows one drone, for example, to navigate without a pilot and potentially without GPS; (2) teaming/swarming, which allows several autonomous vehicles to coordinate with each other with little human input; and (3) orchestration/command and control (“C2), which allows one operator or commander to task many assets across domains. Shield’s Hivemind is focused on the first area, though Shield has advertised teaming/swarming as well. Saronic offers a software stack onboard for individual vessels and its Echelon platform enables orchestration of maritime fleets. Anduril’s Lattice enables all three areas. As more autonomous vehicles become pillars of warfighting, getting the orchestration layer right will be important.

Software versus hardware

This is a key discussion not only in defense tech but also in areas such as humanoid robots, which we have written about previously. Many companies are having to enter these new spaces doing both – the hardware is the entrée to the industry and grabs attention, but you have to develop the software to make it work. Over time, questions abound as to whether hardware will become commoditized and software will be the most lucrative part of supply chain, or whether software will only be as good as the hardware it operates. We heard both sides of the argument at the conference, and we think it’s safe to say that, for now, both are important because both are needed to prove nascent technologies. As technologies mature, this could change.

Investing

• Investors are looking for “choke points” or “pain points” in the defense supply chain as interesting areas to invest in. These include critical minerals, some aspects of semiconductor manufacturing capacity (such as photolithography, which at the moment is dominated by The Netherlands’ ASML and is very hard to replicate). Beginning on January 1, 2027 the US will ban imports of critical mineral batteries from China and the supply chain is not ready. There was discussion about the inertia of suppliers 3-4 layers below the primes that have not prepared for this and whether it would be positive or negative if the US pushed the deadline out.
• Government contracts are an important signal for companies. There are many startups that have won contests or reached certain milestones or been selected for pilot programs, but investors are looking for contract dollars to add certainty to revenue projections. This raised the issue again of appropriations and the fact that the FY2027 defense budget has not yet been appropriated.
• Dual use technologies are important. Defense tech is very en vogue now but the US military is just one customer. Investors are looking at dual use technologies for longevity beyond the military procurement cycle.
• Echoing what we wrote in our weekly a couple of weeks ago, “Will Build Turn to Buy”, many attendees we spoke to expect to see M&A in the space given the number of companies that are competing for business from one primary customer, the DoD. On this topic, we heard a view from one investor that they would rather invest in Anduril and Palantir, even if valuations appear high, over startup drone companies because those companies will likely be survivors in a shakeout (and potentially be able to acquire smaller peers at attractive valuations).

Taiwan

There was a lot of focus on Taiwan at the conference.

• While many lessons from the Ukraine war can be cross-applied to Taiwan, including iteration on weapons development, decentralized procurement, and open architecture for data, a potential conflict in Taiwan will require a different set of solutions for other challenges. For example, the weather in Taiwan is different because of humidity, the electronic signature in the area is different because it is a dense urban environment as opposed to where a lot of the fighting in the Ukraine war is taking place, and of course the tyranny of distance to get to the Pacific theater is real compared with the relatively short distances that drones need to travel in the Russia/Ukraine war.
• Nuclear energy could be a key advantage for Taiwan in a future conflict if it can develop the capability. 80% of Taiwan’s energy supply currently comes from fossil fuel imports, which could be a risk to its industrial semiconductor capacity as well as its broader economy in the event of a conflict.
• Building up a defense manufacturing base is not only good for Taiwan from a military perspective, but it can help the country expand its industrial base as well.
• With respect to autonomy, Taiwan can manufacture hardware, but at the moment it lacks software capabilities and is looking to systems such as Anduril’s Lattice and Shield AI’s Hivemind. The US can also play a role in helping it to scale production for hardware.

Based on our proprietary NPM Price data, the best performers of the large cap names in the private market thus far in 2026 have been:

Based on our proprietary NPM Price data using weighted averages for each sector, the top performing sectors YTD taking into account the top 100 largest names are Climate and Sustainability (173%), AI and Machine Learning (136%), Consumer (+108%), Healthcare/Life Sciences (+82%) and Industrials (+61%). Interestingly, Climate and Sustainability have moved to the top sector after having lagged this year. We note that this is a small sector with only five companies, so individual company performance matters a lot. Two companies have driven the performance: (1) Valar Atomics, having recently raised a round at a $6.2 billion post-money valuation compared with $2.0 billion in March 2026 (TechCrunch, Bloomberg) and (2) Helion Energy, which closed its Series G at a $15.5 billion valuation versus its Series F at $5.4 billion (Company report, ESG Today). These companies are in focus now more because they can provide abundant energy for power rather than because they can provide clean energy to meet decarbonization and climate goals, but their technology accomplishes both.

1. 01
   On the macro side, Brazilian assets rallied after Flavio Bolsonaro took a surprise lead over President Luiz Inacio Lula da Silva in the first round of elections. (Bloomberg, 10/5). French bond yields surged over concerns about the government’s debt load and interest costs, with yields of other euro-area countries Italy, Belgium and Greece following suit Bloomberg, 10/8.
2. 02
   Google and Constellation struck a 20-year agreement that would boost power at 11 existing Constellation reactors Wall Street Journal, 10/6.
3. 03
   Nvidia is rethinking its deals with AI cloud providers to provide credit support in exchange for a revenue share from chip rentals The Information, 10/6.
4. 04
   Apollo is expected to lead a $40 billion debt financing deal for SpaceX to purchase Nvidia chips Pitchbook, 10/8.
5. 05
   Lambda secured a $1 billion delayed draw term loan at a 6.78% interest rate to purchase over 30,000 Nvidia GPUs (The Information, 10/5). It is also reportedly raising up to $4 billion at a $14.5 billion pre-money valuation StrictlyVC, 10/7.
6. 06
   The US DoD pledged to issue a conditional loan of up to $1.5 billion to Wolfspeed, a chip company that filed for bankruptcy last year Bloomberg, 10/8.
7. 07
   Reflection AI announced its first open source model called Beam The Information, 10/6.
8. 08
   Anduril announced that it will build a $6.6 billion shipyard in Baltimore that will manufacture critical components for the Navy’s Virginia-class submarines Tectonic, 10/6.
9. 09
   KKR agreed to acquire Gen II, a private capital fund administrator, for $5.1 billion from Hg and General Atlantic Wall Street Journal, 10/6.
10. 10
    Amazon pledged to give up NDAs and give back $1 billion to communities as it builds out data centers The Information, 10/5.
11. 11
    The Navy is making a $150 million modification to a $50 million award for Shield AI’s X-BAR, which, along with Shield’s cash, brings total public and private funding for the program to $400 million Tectonic, 10/8.

1. 01
   AMD agreed to buy World Labs for $8.2 billion Pitchbook, 10/5.
2. 02
   Day One Data Centers, a Singapore-based hyperscale data developer, filed for an IPO Axios, 10/6.
3. 03
   Australia-based data center operator Firmus plans to allocate shares in its IPO to existing backers Nvidia and Blackstone Pitchbook, 10/6.
4. 04
   In defense tech, both RedLattice and Powerus went public via SPACs The Merge, 10/6.
5. 05
   Northstar Earth and Space, a Canadian space domain awareness company also went public via a SPAC Via Satellite, 10/5.