Why China will win the next war; does AI “think”?; and China rules AI in Africa. Plus my interview with Kyunghyun Cho, a computer science professor at New York University and executive director of frontier research at the Prescient Design team within Genentech Research & Early Development (gRED), and Shirley Ho, Group Leader of Cosmology X Data Science at the Flatiron Institute of the Simons Foundation as well as Research Professor in Physics at New York University.
From Lux Capital
If you missed it and didn’t read last week’s Lux Recommends, we (and by the royal “we,” I particularly mean my partner Grace Isford and our whole Unum Lux family) brought together hundreds of AI professionals — including engineers, scientists, researchers, and startup founders — for a New York City gathering focused on “the AI canvas.” Our goal was to shift the discussion from what’s technically possible to create to what’s worth creating. You can check out my Riskgaming podcast interview with two of the conference’s superstars here, visit our just-posted YouTube playlist of talks, or follow Grace on X for more updates.
In other news, we co-led an $80 million round into Applied Compute to support its work on what it dubs “Specific Intelligence.” Meanwhile, we are backing European space sovereignty startup EnduroSat, where we co-led a $104 million round for the Bulgaria-based compact satellite producer.
Whew, we’ve been busy – and there’s only more to come in the weeks ahead ;).
Speaking to Lux’s work in the defense space, Shahin Farshchi noted at the recent Satellite Innovation conference, “DoD is basically saying, ‘Hey, private sector, spend your equity dollars, do some preliminary work and that will help us decide what we want’.” He continued, “that puts those who have the access to capital in an advantaged position because they are now guiding the purchasers in terms of what they should ultimately solicit. That has been reinforcing the cycle of the most well-funded companies getting access to more business.”
From around the web
1. Guns, not guts
This week, our team liked Phillips Payson O’Brien’s latest for The Atlantic. Contra the Trump team (looking at you Secretary of War Hegseth), modern wars are won through industrial production capacity and logistics rather than battlefield valor. That’s why O’Brien argues the United States would likely lose a prolonged war with China: even as it has lost its manufacturing base, China produces over 50% of the world’s ships and 90% of commercial drones.
Right now, the U.S. has what appears to be the more capable military, and certainly the more battle-tested and technologically advanced one. It might inflict disproportionately higher losses on the Chinese at first. But because of its diminished production capacity, the U.S. would struggle to make up even a small part of the battlefield losses that it would inevitably suffer. China—which is as much the workshop of the world today as the United States was in World War II—could churn out replacement weaponry at an impressively quick pace.
2. Stocks and stockpiles
Over in Politico, Christopher Leonard offers a deeper dive into the United States’ materiel woes. Over the course of this year, major U.S. defense contractors faced waves of strikes as workers protested wages falling behind inflation while the companies prioritized billions in shareholder buybacks and dividends—creating a national security crisis for a Pentagon desperately in need of increased production of missiles and munitions.
Jobs in defense manufacturing are becoming less and less attractive at a time when they need to be getting far more attractive. Many workers are leaving the field or declining to enter it. A survey by the job recruiting firm Acara found that annual turnover in the defense and aerospace industry hit 13 percent in 2023, compared to an average U.S. rate of 3.8 percent. And this is happening just as the need for those skills is rising. Demand for advanced manufacturing skills in the sector is outpacing the number of trained employees, and 75 percent of defense companies are struggling to find qualified employees, the survey found.
What to do? Well, you can judge the options for yourself with our Riskgaming scenario “Hampton at the Cross-Roads.” Play a union boss, CEO, mayor, U.S. representative, admiral or a good ole’ Substacker to see how all these dynamics play out over the course of four scenes at Norfolk Naval Shipyard (NNSY), the country’s largest and oldest publicly-owned naval shipyard and repair facility; Naval Station Norfolk, America’s largest naval base; and Newport News Shipbuilding, which is constructing America’s next-generation Gerald R. Ford-class aircraft carriers.
3. Belt and network
China is not just pulling ahead in military production. According to reporting in Bloomberg, its AI company DeepSeek is dominating Africa’s emerging AI market by offering open-source models that are dramatically cheaper and require less computing power than American versions. It’s China’s Belt and Road strategy Part Deux—leveraging infrastructure dominance to secure broader, long-term influence.
Although much of the world’s attention has been focused on Western tech companies vying for lucrative corporate contracts in the US and Middle East, the meeting in Nairobi illustrates how their Chinese rivals are taking a different approach. OpenAI and its American competitors have focused almost exclusively on proprietary AI — models whose software, training data and algorithms are entirely controlled by their parent companies, with customers paying for access. Chinese firms like Huawei and Alibaba Group Holding Ltd., by contrast, are courting Africa’s startups and innovation hubs with open-source AI models — ones that can be accessed and modified for free, letting companies build products without expensive licenses.
4. Cogito, ergo AI?
Taking a step back, our scientist-in-residence Sam Arbesman flagged Robin Sloan’s “Thinking modes.” Robin laments that we’re losing the deeper meaning of “thinking” when we use it to describe AI’s text generation process, and calls for models that can genuinely comprehend complex systems holistically rather than just reasoning through language chains.
Here is one distinction among several: this process can only compound — the models can only “think” by spooling out more text — while human thinking often does the opposite: retreats into silence, because it doesn’t have words yet to say what it wants to say.
Human thinking often washes the dishes, then goes for a walk.
5. Turning down the heat
Finally, there’s be a lot of discussion around our Slack of Bill Gates’ effort to set a new climate agenda ahead of COP30. He argues that while climate change is serious, the world has made significant progress on cutting emissions and calls for a strategic shift at COP30 to prioritize human welfare alongside emissions reduction. Whether you think his manifesto is an about-face or a logical progression of his previous philanthropic efforts, the whole thing is worth a read.
This inequity is the reason our climate strategies need to prioritize human welfare. This may seem obvious—who could be against improving people’s lives?—but sometimes human welfare takes a backseat to lowering emissions, with bad consequences.
Climate actions need to prioritize improving life for people in low-income countries.
For example, a few years ago, the government of one low-income country [Sri Lanka] set out to cut emissions by banning synthetic fertilizers. Farmers’ yields plummeted, there was much less food available, and prices skyrocketed. The country was hit by a crisis because the government valued reducing emissions above other important things.
