The AI war between China and the United States pits two models against each other: an American model betting on $700 billion in 2026 capex, abundant fossil energy, and total deregulation, against a Chinese model betting on volume, control of rare earths, and chip self-sufficiency.
Neither side wins this war.
The designated losers are the middle powers, who have neither American capital nor China’s monopoly on the 17 critical raw materials that run artificial intelligence. The real winner is whoever owns the infrastructure once both empires have finished exhausting each other.
The documentary « China-United States: The AI War », broadcast on France 5 as a special edition of C dans l’air hosted by Caroline Roux, dissects this shift over ninety minutes of footage shot in Washington, Virginia, Hangzhou, and Gaza. The contrarian take most commentators avoid: the real battle is not fought in data centers. It is fought in the rare earth mines nobody wanted to dig for 30 years.
Warcraft, or the lesson Washington forgot
In the videogame Warcraft, you never win a match by sending units into combat before securing your resources.
Build order comes before the attack.
That is exactly the mistake the United States made in this AI war.
It poured $700 billion into compute (Futurum Group, 2026) without securing access to the 17 critical raw materials China controls at over 90% processing capacity.
You can own the best algorithm in the world. Without rare earths, without gallium, without germanium, the algorithm stays a theory. China played its build order over 30 years, quietly, by accepting pollution the West refused. Strategically, China opened this match with the advantage.
The US is betting $700 billion on fossil-fueled AI
America’s six hyperscalers will spend roughly $700 billion on AI capex in 2026, nearly six times their 2022 level, the year ChatGPT launched (Futurum Group, 2026).
That money builds data centers. Amazon is investing $200 billion, Alphabet between $175 and $185 billion, Microsoft over $120 billion. The Stargate project alone mobilizes $500 billion to try to « break the back of Chinese power, » in the words of experts interviewed in the documentary.
This infrastructure runs on natural gas, not renewables. The documentary shows Virginia data centers powered by gas, with an environmental activist warning about methane emissions while no national debate exists on the physical limits of this expansion.
Virginia alone hosts more data centers than all of France. The United States hosts 45% of the planet’s data centers. That choice carries a cost no official carbon accounting has publicly owned yet.
American electricians, plumbers, and HVAC technicians now earn over $86,000 a year straight out of training, riding this boom. It is blue-collar revenge while white-collar workers start to tremble.
Every junior graduate is the first silent casualty of the « job apocalypse »
The job apocalypse refers to the wave of white-collar layoffs directly attributed to generative AI, an unprecedented phenomenon because every prior technological revolution hit blue-collar workers first.
The 2026 numbers confirm the scale. AI was cited in 26% of announced layoffs in April 2026 per the Challenger report, and over 123,000 layoffs in US tech alone have been attributed to AI automation this year (CBS News, 2026).
Entry-level roles disappear first. Sam Altman himself funded a three-year study on 1,000 low-income Americans receiving $1,000 a month with no conditions attached. The result: people worked slightly less but did not leave the labor market, undercutting the argument that universal income would replace professional ambition.
Tech CEOs now propose universal basic income as the answer to the mass unemployment they are creating. It is the Roman equivalent of bread and circuses, in the words of Nobel laureate Joseph Stiglitz, interviewed in the documentary. Nobody has yet asked those same CEOs to pay the bill.
China is turning its chip deficit into a rare earth advantage
China’s rare earth monopoly refers to Beijing’s control of over 90% of global processing capacity for the critical materials essential to semiconductors, batteries, and military systems.
Between January and June 2026, neodymium-praseodymium prices increased sixfold and tungsten prices tripled, while Chinese export license approval rates for European firms fell below 25% (Rare Earth Exchanges, 2026).
This economic weapon is not new. As early as 1992, a Chinese leader declared that the Middle East has oil and China has rare earths. Beijing built this monopoly by accepting pollution the West deemed unacceptable on its own soil, turning a Western environmental choice into a Chinese strategic dependency.
Without Nvidia chips, China compensates with the Turing chip, three times more powerful according to its maker, built to power Chinese autonomous cars, robots, and flying cars. The stated goal is to control its own technological destiny instead of buying abroad.
Facing this dependency, Europe invested 300 million euros in a single mine capable of covering 100% of its gallium needs, but independence is not expected before 2030-2035. China is already playing the next round.
The robotic war has already started without a declaration
An autonomous weapon is a system able to identify and engage a target without real-time human validation, a technology no international treaty regulates today.
In Ukraine, a Russian position was fully captured by remotely piloted robots on April 13, 2026, according to President Zelensky, a historic first confirmed in the documentary. The French army is preparing its own deployment, Operation Pendragon, with operational readiness targeted for summer 2027.
The Lavender targeting software used by the Israeli army in Gaza pushed target identification capacity from 50 per year to 100 per day, with civilian casualty tolerance thresholds set at 20 per low-ranking Hamas member and up to 300 for a senior commander, according to the journalistic investigation revealed in the documentary.
The problem is not AI itself. It is the human disengagement it enables by cutting decision time to a few seconds. A French general interviewed explicitly rejects this model for the French army, on one simple principle: AI can never be held accountable, only the human chain of command is.
No international treaty regulates autonomous weapons today, unlike nuclear weapons, where two or three historical examples show human judgment averted catastrophe.
The human brain has become this war’s last frontier
A brain-computer interface is a technology able to read, and potentially modulate, neural activity through implants connected to artificial intelligence, a field where Elon Musk’s Neuralink and a Chinese neuroscience lab are advancing in parallel.
Neuralink has already implanted close to 20 patients according to information gathered in the documentary, while the competing Chinese lab is on its second generation, tested on monkeys, with human trials announced for later this year. The gap is measured in months, not years.
Medical use is advancing fast and without much controversy: a paralyzed man walked again using an exoskeleton controlled by thought. The red line sits elsewhere, in what the Chinese researcher himself admits without hedging: decoding thoughts is one thing, controlling them to « pull a patient out of a pathological state » is another, and nobody has drawn the line between care and control yet.
No international legal framework exists for this territory either. Responsibility for drawing that line falls, in the researcher’s own words, on every neuroscientist on the planet, not on one company or one country.
Europe has the money and the brains, it lacks the decision
Europe’s AI gap refers to the investment and compute-power difference between the European Union and the two superpowers, estimated at 132 billion euros invested against $647 billion on the American side in 2026.
Only 4% of the world’s AI compute power is located in the European Union, against 70% in the United States, according to figures cited by Thierry Breton, former European Commissioner for the Digital Single Market, during the debate that followed the documentary’s broadcast.
The dominant narrative claims Europeans only know how to regulate, never to invest. That narrative is largely built by American interests themselves, as Mistral’s CEO points out in the documentary. The reality is more nuanced: Europe holds the most structured health data in the world and an AI regulation that has already inspired digital agreements with South Korea, Japan, and India.
What Europe lacks is not talent. Nvidia, the world’s most valuable company, is led by a Taiwanese executive and employs Chinese, Indian, and French engineers. What is missing is the decision to put capital on the table now, not in five years.
Comparison table: the two models of the 2026 AI war
| Criterion | US model | Chinese model |
|---|---|---|
| 2026 AI capex | ~$700B (hyperscalers) | Massive state investment, lower production cost |
| Energy | Natural gas, reopened fossil plants | National energy mix, less public debate documented |
| Chip access | Nvidia dominance, Taiwan control | Restricted, offset by domestic Turing chip |
| Strategic leverage | Financial capital, the dollar | Rare earth processing monopoly (>90%) |
| AI business model | Free then paid, data capture | Nearly free, state-subsidized |
| Military regulation | Unlimited acceleration (Biden-era rules rolled back) | Total opacity, real capabilities unknown |
| Global compute share | 70% | Not officially disclosed, catching up fast |
| Main vulnerability | Dependence on Chinese rare earths | Access to leading-edge chips |
What Europe and France must do now
The narrative of European decline mainly serves American interests, which sell their own digital sovereignty as the only option on the table. This battle is not lost yet, but it plays out on three fronts at once: control of data, sovereign compute infrastructure, and training on the tools. That last point is exactly where the gap opens between companies that will suffer disruption and companies that will drive it: learning to orchestrate AI instead of enduring it, as Thierry Breton put it.
That is precisely the blind spot the Claude Sprint covers, four focused sessions to turn a team that is discovering AI into a team that exploits it before its competitors do.
FAQ
Q: Who is going to win the AI war between China and the United States?
A: Neither side « wins » in the classic sense. Both superpowers make dangerous compromises with their own values to stay in the race, a dynamic researchers call an unhealthy competition where even the apparent winner loses on democratic and ethical grounds.
Q: Is China really behind on AI chips?
A: Yes, but the gap has narrowed to a year or a year and a half according to experts, and China offsets it with domestic chips like the Turing and its monopoly on processing the rare earths the entire semiconductor industry needs.
Q: Is universal basic income really the answer to AI-driven unemployment?
A: No, that is not what the evidence shows. The study Sam Altman funded found that recipients of guaranteed income kept wanting to work; the problem is not just financial but social and identity-based, something a monthly check does not fix.
Q: Has Europe really lost the AI race against the US and China?
A: No, but it has a real gap in capital invested (132B euros against $647B for the US in 2026) and in compute power (4% against 70%). It keeps an edge in health data, regulation exported internationally, and a recognized scientific talent pool.
Q: Are autonomous AI weapons already used on the battlefield?
A: Yes. The Lavender software was used by the Israeli army in Gaza, and robotic units captured positions in Ukraine without direct human intervention in April 2026. No international treaty regulates them yet.
Q: Why does China have a monopoly on rare earths?
A: Because since the 1990s it accepted the industrial pollution Western countries refused on their own soil, building over 90% of the world’s processing capacity for these critical materials in the process.
Verdict
This war has no winner, it has hostages: the middle powers watching two empires play a negative-sum game with ethics, democracy, and climate as adjustment variables. The real risk is not that China or the United States wins. It is that Europe stays a spectator while the decision gets made elsewhere. The window to act is 12 to 18 months, not ten years. Train your teams to orchestrate AI now, or let others decide for you what it will do to you.