The timing was perfect. Just as Vladimir Putin stepped onto US soil last year for talks over Ukraine, President Donald Trump invited him to look up to the grey Alaskan skies. A B-2 stealth bomber and four F-22s swooped low overhead in formation, a vivid expression of America’s superpower military strength. When Trump visits other nations, though, he is accompanied by something he considers just as prestigious as a B-2: his entourage of tech chief executives. Silicon Valley has become his touring backing band. Team Trump considers artificial intelligence (AI) to be the ace card. Here are our best brains, capital, technology and success, Trump is saying. Beat that, Mr Xi.
Favourites come and go from Trump’s imperial court – just ask Elon Musk – but when the president visits Shenzhen next week for the Asia-Pacific Economic Cooperation summit, one of his chosen few will be missing. Founder Jensen Huang this week was forced out of his job as chief executive of AI behemoth Nvidia.
Huang stands accused of betraying his country by sleeping with the enemy and so permitting China to get its hands on arguably the world’s most powerful product, the AI processing chips his company produces.
Over 33 years, he had steered the chip maker to become the most valuable company on the planet. For intensive AI processing, Nvidia chips have outperformed rival products. And – so the wonkish narrative goes – AI holds the key to dominating the future of the planet.
“Whoever wins AI, wins,” Trump declared bluntly in September.
It is this potentially unprecedented capability that makes AI and its champions not just technological, but also political powers. Three years ago, Sir Tony Blair suggested that “how you understand, master and harness this technology revolution will define the place of this country and the shape of the world”. Blair forecast a world in which “China and the US are leading the race for AI dominance and [the] 21st century is going to be shaped by the competition between the two countries”.
Now it feels as though Blair’s forecast has taken concrete shape, with the leather-jacketed Huang’s defenestration coming amid a frenetic and formative period of American politics. Last week, Trump faced an astonishing electoral defeat in midterm elections. Next week, he sets off to see Xi Jinping in China.
President Donald Trump and President Xi Jinping will meet at the upcoming Asia-Pacific Economic Cooperation summit in Shenzhen (pictured at the White House in September 2026) – AFP via Getty Images/SAUL LOEB
At the latter, the newly weakened Trump will find that AI advances are bound into great-power politics every bit as tightly as explosive yield increases were in the postwar nuclear age. For US vs Soviet megatons in the 1950s, read training compute – the raw processing power used to build AI models – and floating-point operations (also known as Flops, these measure the number of calculations) of rival AI models today. The critical difference being that where nuclear bombs were designed, built and controlled by nation-states, today’s ever-more-powerful AI agents are the property of companies and, often, the single founder at their apex. As never before, Huang’s fall is testing the strange relationship between entrepreneur, company, governing political party, opposition and nation-state itself.
As if to ram home the point, Trump will fly to the summit from a domestic political scene roiled by popular anger about the data centres that drive American AI models. For now, these are acknowledged to lead the world. But for how long? Will the US now side with greater controls and caution or, recognising the single most important game of “winner takes all” history has ever seen, go for broke?
It is a debate that is shaping partisan politics: Democrats and AI companies such as Anthropic on one side, and Republicans, Elon Musk and the Trump White House on the other.
Unlike Trump’s Alaskan meeting with Putin, the location of his upcoming superpower summit will be to his interlocutor’s advantage. Shenzhen is China’s tech hub (it is home to BYD, Huawei and Tencent) and the summit is an opportunity for Xi to show off his own country’s power. Thousands of business leaders will flock there alongside Trump and Xi.
It seems unlikely that Huang will be among them.
Huang was forced to step down from his position as chief executive of Nvidia among accusations of several extramarital affairs – AFP/Philip Fong
Scandal now hangs too tightly about him. The 63-year-old Huang initially denied the affair with one of his distributors based in Singapore, before admitting it was true.
His mistress, Alice Huang of Megaspeed, is a Chinese national who had worked for the state broadcaster, and it was reported that the purported Singapore headquarters of Megaspeed (which denies wrongdoing) sits practically empty. She’s barred from entering the city-state and is suspected of being a spy for China.
All this matters because in 2022, the US government restricted exports of Nvidia’s most advanced AI chips to China and Russia, citing the risk they could be put to military use. Jensen Huang had consistently lobbied to sell more of these chips to China, dividing opinion in the Trump administration. So his personal affair has now spun into a geopolitical scandal.
Revelation of the affair led to rueful expressions among chip analysts who had long wondered how China had managed to obtain Nvidia chips for its AI developers when those chips were subject to an embargo. They already had their suspicions.
In May, a tiny three-man Wall Street outsider firm that provides advice for short-sellers, Culper Research, founded by 31-year-old Christian Lamarco, was making its own inquiries.
Nvidia had said that after sanctions were tightened in April 2025, its trade with China fell to zero. Lamarco found that in reality, more than 20 per cent of Nvidia’s annual income “remained driven by China”. It turned out that Jensen Huang’s mistress’s firm was a shell, financed by Chinese tech giant Alibaba, and the chief executive had, at best, been turning a blind eye to what Culper called “illegal GPU diversion and Southeast Asian intermediaries”.
A whiff of conspiracy
For some, all is not as it seems, and at least a whiff of conspiracy lurks behind Huang’s dramatic demise. It’s a theory that gets to the heart of a highly political stand-off over who controls the world’s most important technology.
In the first camp are the “accelerationists”, who want to unleash the full power of AI as soon as it can be realised, believing it will bring vast riches and extraordinary scientific advancement. Pitched against them are the “safetyists”, who are concerned about existential risk and believe tough regulation is needed to curb its development and use.
Trump has become the chief cheerleader of the accelerationists, describing calls to restrict AI development as based on a “sick conspiracy” and the leaders of Anthropic, the safety-based AI lab in the US, as “Left-wing nutjobs”. Meanwhile, Musk, the owner of X and the Grok AI bot, acknowledges that “there’s at least some percentage chance” of AI being an existential threat, but said that he’d concluded: “There’s nothing I can do to stop it. So I basically came to the conclusion I’ve got two choices, which is either be a participant or a spectator.”
President Trump hosted leaders of AI at the White House in September 2025 – Reuters/Jonathan Ernst
So, argue some in the US, it is not only foreign competition that looms over American AI. There are also “doomer traitors who want to slow down American AI,” notes Brian Chau, co-founder of the Alliance for the Future, a pro-AI public affairs campaign.
For Chau, the Jensen Huang scandal is simply another part of a campaign of destabilisation against American AI, whose most powerful elements are based in the US itself, belonging to the powerful effective altruism (EA) movement, which is devoted to preventing existential risks but has been compared to a cult. AI companies including Anthropic and OpenAI, as well as much of today’s AI mythology, emerged from these circles, bringing with them a preoccupation with scenarios of world-ending rogue AI.
They have had a significant impact on the Democratic Party’s approach to AI. Under Joe Biden, “there was an attempt to infiltrate the halls of power with EA thinkers,” Suresh Venkatasubramanian, an AI adviser in the early days of the Biden White House, told The Telegraph earlier this year. “There was a vacuum that they filled. No one thought to ask, ‘OK, what agenda are you bringing?'”
Now, with the Democratic Party controlling Congress, the EA movement is getting a more sympathetic hearing once again. Democrats are planning to use their newly won power to defy Trump with a flood of new laws aimed at controlling AI systems, marking a significant shift from the recent Republican-led era, in which Congress repeatedly refused to pass legislation restricting the technology. Democrats are planning to introduce bills containing measures that would require independent testing of AI systems and set out the circumstances in which AI firms could be held liable for harm caused by their models, Reuters has reported. In the superpower AI race, some worry America’s lead over rival powers will inevitably shrink.
But is it fair to describe today’s technological showdown as a “race” at all? All Nvidia really does is provide a decisive cutting edge for AI work. The edge may not last. Costs elsewhere are coming down, and its 75 per cent gross margins are ever-harder to justify. Google and Microsoft, reluctant to pay the “Nvidia tax”, are developing their own AI chips. In addition, radical new chip designs, such as those from the UK start-up Fractile, are thought to be able to perform key parts of an AI task – the part called “inference” – for a fraction of the energy and cost.
China itself has two aces up its sleeve. Clever engineering and the canny strategic decision to give away AI models for free have thrown America’s orgy of capital expenditure on Nvidia chips and data centres into question.
“When you look at the approach to AI by the superpowers, it’s only the United States that has declared a race,” says former Facebook executive Georg Zoeller, who now advises governments on AI and lectures to company boards in Singapore.
In January 2025, China had its own “Sputnik moment”, when it released an AI model that was engineered to run at a fraction of the cost of the American rivals.
“A mid-sized hedge fund had achieved for just $5m what cost Elon Musk $400m to create,” Zoeller explains – although China’s claims about building DeepSeek on a shoestring budget were strongly disputed by many analysts.
The Open models
Most importantly, China decided to give the model away, ceding control of the technology as soon as it was published. It was a decision partly made from necessity – from not having the latest and greatest Nvidia chips. But it was also strategic. “Open” models mean users can tweak them.
Businesses love them. Not only are they customisable, they are relatively cheap, having not been developed on the back of a vast capital splurge whose costs must be recouped.
China’s economy, based on manufactured goods, benefits from making AI cheap. It can give it away and capture the value in physical manufactured goods. And China has another reason for not wishing to play the game on Washington’s and Silicon Valley’s terms.
Wang Huning, China’s chief political theorist and a strategic adviser to three presidents, believes rampant tech chief executives destabilise society. Unlike Trump, Xi and the CCP Politburo regard tech chief executives not as heroes, but as potential problems. In 2020, Alibaba’s former chief executive Jack Ma disappeared. He later resurfaced, but the crackdown on the moguls has not relented.
One of America’s own moguls, Microsoft chief executive Satya Nadella, has acknowledged the Chinese state may have a point – that reliance on OpenAI’s and Anthropic’s AI may mean US firms lose their autonomy – and perhaps even their profits.
“The last thing any of us want is a world where every company across every sector is ceding value to a few models that eat everything they see,” he wrote. “If all the value is accrued by only a few models, the political economy will simply not tolerate it. There is no societal permission for an AI future that hollows out entire industries.”
He compared it to globalisation. Palantir chief executive and founder Alex Karp has picked up the theme, asking why hand “the means of production” to two companies?
Satya Nadella of Microsoft has commented on the risks if ‘value is accrued by only a few [AI] models’ – AFP via Getty Images)/Jason Redmond
Mathematicians will know how being robbed feels. They were astonished to learn that dozens of OpenAI agents appeared to have used methods they had privately been working on in the chatbot to solve a complex Millennium Prize maths problem. Then one professor alleged that the firm tried to pressure him into working with them on the solution (an OpenAI scientist subsequently insisted that he had only warned the professor of the risk to his career of making “unfounded” claims).
Since the Democrats swept the midterms this month, the tech bros have been discovering how unpopular they really are – and rapacious behaviour doesn’t help.
Anthropic is lobbying to make the open models, including Chinese models, illegal, arguing they are too dangerous. Sceptics call it naked protectionism.
Democrats who favour strong regulation on AI “safety” – an invention of the EA cult – are playing into the duopoly’s hands. “Existential risk is a regulatory capture mechanism,” says Heidy Khlaaf of the AI Now Institute.
Martin Skold, a geopolitical analyst and security expert, thinks the US is suffering from a “digital disease”, where a nation excels in one field, such as oil, and so grows lazy and neglects other wealth-creating sectors. AI is an expensive distraction, he says.
“To beat China, we need chips and aircraft and an industrial base. AI is not a solution to that.”
“China’s economy has worked out how to do cheap and good, and that’s a threat to the financialised American economy,” agrees Zoeller. One with real engineering, not financial engineering at its core.
In the past, technology companies have found their supposedly iron grip on the market dislodged by successors – IBM by Microsoft, Microsoft by Google. But the capacity for self-improvement of AI means it may be the last tech race we ever know.
So as Xi reflects on the touring Silicon Valley tech bros this week, he might recall words attributed to both Sun Tzu, an earlier Chinese strategist, and Napoleon: never interrupt your opponent when he is making a mistake.
