Here’s How the AI Trade Blows Up (Hint: It Has to Do With OpenAI, Anthropic, and China)
Investors who saw the artificial intelligence trade coming 10, five, and three years ago have done extraordinarily well.
The launch of large language models (LLMs) from OpenAI, Anthropic, and others has driven incredible demand for chips, memory, compute, and other parts of the AI supply chain — so much so that investors have been scouring that chain for areas where demand could be constrained.
But as the market has gone up and up, investors have grown wary, whether due to the hundreds of billions of dollars companies are investing in AI infrastructure or regarding valuations that, at times, have seemed ludicrous.
No one knows how this will all end, but here’s the scenario in which the AI trade blows up. It has to do with OpenAI, Anthropic, and China.
Image source: Getty Images.
Over-reliance on OpenAI and Anthropic
As I mentioned, hyperscalers such as Microsoft, Amazon, Alphabet, and Meta Platforms are preparing to spend north of $700 billion on AI-related capital expenditures (capex) for chips, memory, data centers, and other AI infrastructure.
Management teams in their respective second-quarter earnings calls also indicated that capex could reach even higher levels next year. Investors have grown skeptical because this spending has led to significant deterioration in free cash flow, and some of the above hyperscalers are now experiencing negative free cash flow.
Investors also question whether this kind of spend will translate into strong returns on invested capital (ROIC).
The hyperscalers argue that they are not gambling with these investments because they already have customer commitments through 2027 and 2028, giving them a clear line of sight into revenue and strong returns. Who is anyone to argue? After all, the leaders of these companies are incredibly smart and have built some of the strongest companies in history.
However, an increasing number of reports suggest that most of this demand is coming from OpenAI and Anthropic, which need more compute as their models become more complex and more widely used worldwide.
For instance, analysts at Barclays recently issued a research report suggesting that Amazon Web Services (AWS) will generate 73% of its AI-related revenue this year from OpenAI and Anthropic.
Today’s Change
(-2.09%) $-5.82
Current Price
$272.27
Key Data Points
Market Cap
Day’s Range
$271.36 – $278.84
52wk Range
$196.00 – $287.20
Volume
31.9M
Avg Vol
50.3M
Gross Margin
50.77%
Amazon has seen AWS sales increase by about 33% in the first six months of the year, compared to the same period last year.
Meanwhile, UBS projects that 28% of Google Cloud’s AI revenue will come from the two LLM giants this year, and 48% next year. Other companies are in similar positions. For instance, $300 billion of Oracle‘s backlog is reportedly expected to come from OpenAI.
All this demand for data centers has driven demand for chips from Nvidia and for memory to feed the chips’ data from companies like Micron and SK Hynix. It’s a large tangled web.
OpenAI and Anthropic do not appear to be impenetrable
Cracks are starting to emerge in OpenAI and Anthropic. OpenAI lost $38.5 billion in 2025.
Perhaps a bigger issue is that the moats the companies have built may not be impenetrable. OpenAI and Anthropic run closed-loop, proprietary models. While they have reportedly had the best performance thus far, they are also quite expensive for companies to use.
Meanwhile, Chinese companies have begun releasing open-source models, meaning anyone can access the source code and modify them. These models are reportedly 60% to 90% cheaper than those from OpenAI and Anthropic, with performance supposedly catching up, too.
Now, there are concerns about the models coming out of China from a national security perspective, and the belief that they need to be regulated.
Recently, hyperscalers and others in the AI ecosystem, such as Nvidia, issued a joint letter urging lawmakers not to act too hastily in regulating open-source models.
If there is a pricing war and OpenAI and Anthropic lose a significant amount of their market share and pricing power, what happens to all the compute deals they’ve signed? The hyperscalers will have just spent hundreds of billions on AI infrastructure.
If investors see clear evidence that there won’t be strong returns on the invested capital, their stocks are likely to get hit hard.
Are OpenAI and Anthropic too big to fail?
The scenario laid out above is hypothetical, of course, and may never come to fruition. In fact, most market crashes are driven by events that very few see coming. What I’ve laid out above has been discussed widely in the financial media and among many experts and pundits.
We’ve also seen evidence that OpenAI and Anthropic may simply be too big to fail, given how many companies depend on deals they have struck with the two LLM companies.
The Wall Street Journal recently reported that Nvidia is considering guaranteeing $250 billion in debt that OpenAI wants to use to lease a huge AI data center in Pike County, Ohio. The report cited anonymous sources, so it’s not official yet.
There’s also a chance the U.S. government could take a stake in OpenAI. The government might say that it can’t let this company fail due to national security concerns about winning the AI race between the U.S. and China, as well as the economic implications.
So, it’s quite possible that OpenAI and Anthropic will end up being OK. It’s also possible that competitors can fill the void in AI compute if Anthropic and OpenAI suddenly pull back.
I also think most hyperscalers would be able to navigate a potential blowup from one of these LLM giants, though their stocks would still take a big hit.
In the first half of the year, AWS accounted for 21% of Amazon’s total revenue and 61% of operating income, so the loss of AI revenue would hurt the company greatly but likely not kill it. Even before AI, many companies around the world had not transitioned to the cloud.
While a blow-up may never happen, investors should always understand the other side of their trade and the worst-case scenario for stocks they might own or that are lifting the broader market.
