The US economy - China prepares to deliver the sucker punch
Igor Omilaev on Unsplash
Yanis Varoufakis, the charismatic ex-finance minister of Greece and closely watched observer of geopolitics, recently made the case in a podcast titled “Something ugly is about to hit America“ that the US economy is collapsing under the weight of its own unforced errors. No other competing countries need do anything; they can just sit back and watch in glee (or horror, depending on their perspective) as the US economy starts to crack, and then to crumble.
I am going to argue that China is not going to sit back and passively watch the US economy weaken itself – it is going to deliver an additional sucker punch as the US struggles to staunch the bleeding from its own self-inflicted wounds. You know the homily, “don’t kick a man when he’s down”? I suspect that China does not subscribe to that.
First, here is a summary of the Varoufakis case. Yanis Varoufakis is sounding the alarm on what he describes as the “controlled demolition” of the American economy. In his assessment of the US landscape in 2025, he argues that the nation is facing a perfect storm of its own making. It began with April’s aggressive tariffs, which acted less as protectionism and more as a $430 billion tax on domestic businesses, crushing margins and triggering immediate layoffs. This supply chain shock was immediately compounded by mass deportations that drained the labour force, sending agricultural prices skyrocketing simply because there were no hands left to harvest the crops.
It gets worse. Rather than stabilising the ship, the government accelerated the chaos with DOGE-catalysed austerity. Varoufakis argues that the slashing of 300,000 federal jobs under the guise of efficiency removed a critical economic stabiliser just as the private sector began to buckle. Services frayed, such as veterans’ health care access and air traffic control. Meanwhile, a $38 trillion national debt has trapped the Federal Reserve, forcing interest rates to remain punishingly high to service bondholders. This has frozen the housing market and suffocated investment.
Varoufakis concludes that these aren’t isolated problems; they are a “doom loop”. Tariffs and labour shortages drive up inflation, which drives up interest rates, which crush businesses and housing, leading to layoffs. Government cuts then remove the demand that could have saved some of those businesses. It is a vicious cycle that feeds on itself and will force the US into a severe recession or even depression.
OK, perhaps he is right, or maybe partially right. What is not clear is the timing of all of this, because if the US hangs on until elections ’28 or even the midterms ’26, much could change. But the US faces a much larger looming problem, one which seems more immediate and more concerning, and it is China.
Michael Power, a well-known geopolitical commentator, global strategist, and China expert, has written an exhaustive and magisterial paper on what China is up to on the AI front. The paper is titled “No more Moore? So, what then for microchips? And for China?“. This title does not do the paper justice – it is about more than chips – it paints a vivid picture of how carefully China has built an AI ecosystem whose competitive walls now look to be unbreachable.
Power argues that Moore’s Law – the observation that computing power doubles every two years – is basically dying, killed by three converging factors: physics, materials science, and economics. Below 3 nanometres (a measure of the size of a specific transistors and other components on a silicon chip – Nvidia chips live in this rarified space), chips literally fall apart at the atomic level. Electrons tunnel through barriers, copper wires degrade in days, and everything stops working reliably.
Building newer, smaller chips is at a point of diminishing returns – a weak cost/benefit advantage as well as the physics wall. And more importantly, delivering an AI service is bigger than just about chip excellence – it is only one part of a larger ecosystem.
Here is where the US-China divide gets interesting: instead of chasing the impossible nanometre race, China said “nah” and went in a completely different direction. The strategy? “Less Small. More Smart”. Rather than betting everything on one mega-powerful chip (like Nvidia’s Blackwell), China is building what Power calls “Sherman swarms” (referring to the US workhorse mid-level tank of WWII) – moderately powerful chips linked together through an advanced network and clever software. Think “hundreds of decent tanks working together” rather than “one amazing super expensive tank”.
The cost difference is staggering. America’s Project Stargate plans $225 billion (R4 trillion) for computing power. China could achieve the same capability for $861 million (R15 billion) using cheaper components and optimising the entire ecosystem together. More crucially, because they are working as an integrated system, they would deliver roughly 6.7 times more practical computing power than America’s approach. This reflects a fundamental philosophical difference.
Add in China’s massive renewable energy advantage (paying a fraction of US electricity rates) and control over semiconductor supply chains, and the Western tech dominance narrative looks increasingly shaky. China isn’t trying to win a head-to-head chip race – it has relocated the battlefield entirely. (And who can forget the humiliating spectacle of Trump arriving for his summit with Xi armed with his Nvidia negotiating chip in his back pocket, to which Xi basically said - no thanks, and we’re not even going to buy any of your other chips anymore either).
So back to my point. Where exactly is the sucker punch? The US has bet its entire economy on AI. Trillions have been committed by companies and institutions (and the US government) to winning this race. If (or perhaps when) the bubble pops, it will be deafening.
Power’s paper also highlights that American tech companies prioritise quarterly returns – returning cash to shareholders rather than investing long-term. China operates on a national balance sheet with a generational perspective. Beijing’s “Big Fund” can absorb losses for years because it is building strategic autonomy, not quarterly earnings.
Which brings us to this: China’s long view means they can afford to give its AI away essentially free, which it is doing. Its big LLMs are open source. Many US tech companies (especially startups) are already using Qwen from China’s Alibaba rather than any US LLM, because it is free and just as good.
It is China who will pop this bubble by dumping free AI on the US market. OpenAI and the rest will not be able to compete commercially with higher priced offerings of what is essentially the same AI service. They will not live up to their hyperbolic promises. And tariffs can do nothing about that because it doesn’t come through ports; it arrives over the Internet in bits.
Steven Boykey Sidley is a professor of practice at JBS, University of Johannesburg and a partner at Bridge Capital and a columnist-at-large at Daily Maverick. His new book “It’s Mine: How the Crypto Industry is Redefining Ownership” is published by Maverick451 in SA and Legend Times Group in UK/EU, available now.


I fully agree with the points made in this article. Funny I was debating USA v. China with one of my Yankee friends this weekend and him saying that Chinese products and infrastructure are inferior. Look at the high speed rail numbers - China has built close to 50,000 kilometers while the USA built less than 1,000. The USA is a country in decline. The ruling class decided it was better for them to hoard wealth and spend huge sums on a bloated military - end stage capitalism. I am a MMT adherent and don't believe that there is a "debt" problem in the USA. There is a huge inequality problem. A booming stock market makes no difference to the vast majority of Americans. And many economists think the entire economy is built on a flimsy house of cards known as the AI bubble.
Another great article. I love reading your insights. A few minor comments:
Moore's law ran out of steam at least 10 years ago (just compare your current PC's CPU with that of your PC 10 years ago). GPUs were already an architectural circumvention/extension that just happened to work extremely well for AI (but not necessarily other computing tasks, especially those that are hard to parallelize). This obviously *supports* your argument even further.
Nvidia's grip is already loosened - the best LLM to date, Gemini-3 released last week by Google - was trained on non-Nvidia chips (but T4s are not commercially available).
Indeed China is reading to lead the AI race not just because of hardware. Just 3 years ago, virtually all the 'frontier science' AI articles I was reading for my particular interest (hybrid AI architectures) were published in the West (mostly USA with a decent show of Europe). Now, more than 80% of the most novel/innovative papers are entirely from China, and of the remaining 20% I'd say half of the authors have Chinese surnames.
Indeed the giant's feet of clay are crumbling... although one shouldn't under-estimate the resilience and innovation remaining in the US economy. I doubt (or hope not?) that it will be as clear a 'crash' as you make it out to be.
Keep up the good work.