Listen to the audio version of this article (generated by AI).
Key Points
- Chinese AI models continue to take market share from American AI models, even though American models still capture the more profitable portion of the market for now.
- AI users are shifting to Chinese models because they offer “good enough” performance at costs that may be just a few percent of what American models charge.
- Investors need to carefully watch for the signs of a turning point in the AI bubble, since the industry requires significant ongoing investment to remain afloat.
Usage of Chinese artificial intelligence (“AI”) models continues to soar, pushing once-dominant American AI models to the wayside. In the last week of June, Chinese AI models processed 48% of tokens, compared with U.S. models’ 20%, according to AI model brokerage platform OpenRouter. It’s a huge swing from a year ago, when U.S. models held a 74% to 20% lead over China’s.
The key reasons? Chinese AI models offer “good enough” performance at a fraction of the cost of American models, leading many users to switch platforms.
The shift will prove to be problematic for American AI firms such as OpenAI and Anthropic that have taken a “more is more” approach to their models. That is, the American industry is geared to throwing more resources at AI to develop ever more powerful models.
This massive buildout of AI computing resources vastly overextends what’s actually needed to “solve” AI, putting some of America’s largest companies in an increasingly risky position.
That’s led Big Short investor Michael Burry to proclaim “the end is nigh” for AI, and he’s been betting against some of the frothier stocks in the sector, such as Micron Technology (MU).
Burry has criticized the AI industry for focusing on large language models rather than reasoning models. This focus made it appear that AI could be solved if only more resources were put to work, leading to massive investment in the sector. Instead, this approach means that trillions of dollars are being spent on American AI, when Chinese AI is making a product at a much lower cost.
With sky-high valuations on AI-related stocks and a Chinese AI industry that’s rapidly taking market share, investors need to be watching for these three signs that America’s AI bubble is in danger of bursting.
1. Increasing Chinese and Open-Source Share of AI
As recent OpenRouter data shows, Chinese AI has been gaining market share at the expense of American AI modelers, including OpenAI, Anthropic, and Alphabet (GOOGL). While that may not be a problem at the moment, it will become a major problem over time.
Now, AI bulls will argue that Chinese AI is just taking the low-margin business at the bottom, while American AI captures the more profitable segment of the market. That’s true for now. But we’ll see the same thing occur in AI that has occurred in every major industry China has focused on: They’ll capture the “unattractive” low end of the market and work their way up to the top.
We’re already seeing that play out in ways that AI bulls don’t appreciate. OpenAI and Anthropic build highly competitive, even superior, models. But Chinese AI is cheaper and nearly as good. Businesses have discovered that they just don’t need that “extra 10%” performance when the American models regularly cost 10 to 20 times more, and in some cases as much as 150 times more.
This cost is a huge factor. Among startups using open-source software, about 80% are using Chinese models, according to Martin Casado, general partner, Andreessen Horowitz.
With tons of low-cost Chinese models and the potential for no-cost open-source models and on-device AI, consumer-focused AI models are likely to become fundamentally unprofitable. For the moment, that leaves the enterprise space in play. But given the cost of American AI, firms are already reconsidering their AI spending. If cheaper and nearly-as-good solutions exist, many firms will pivot.
While niche applications can continue to exist – focused high-end enterprise models or high-security government models – they may still face substantial pricing pressure.
2. OpenAI Can’t Get Funding or Go Public
The announcement of OpenAI’s initial public offering (“IPO”) and its near-immediate walk-back to 2027 seemed to throw the market for a loop. (We had already analyzed the potential IPO delay before the market decided to freak out.)
The company lost nearly $21 billion in 2025 and an estimated $7 billion in the first quarter of this year. Investors should watch OpenAI closely because it MUST raise money to continue operating. If not, the whole AI ecosystem is at risk of unraveling.
In the wake of the news of the IPO delay, AI-related stocks, including key memory stocks Micron and Sandisk (SNDK), plummeted – an omen of what could happen later.
OpenAI is central to the entire system because of how the AI buildout is financed. OpenAI promises to buy billions in “compute” from AI cloud providers such as Cerebras Systems (CBRS). These providers then borrow against those spending commitments to build the infrastructure needed to deliver this service. Currently, OpenAI has made an estimated $1.4 trillion in spending commitments over the coming years, according to Barron’s.
That’s a massive financial commitment from any company. But OpenAI had just $5.7 billion in revenue in the first quarter of 2026, according to The Information, so it must continue to seek new funding to meet its commitments. If not, the entire AI ecosystem could easily spiral out of control, as the flow of money dries up. OpenAI has even considered a price war to help keep revenue growing, further sacrificing profitability.
But much of the recent money flowing into OpenAI is not venture capital. Instead, it’s money from the companies that benefit from the ongoing AI buildout and need it to continue: Microsoft (MSFT), SoftBank (SFTBY), and Nvidia (NVDA). These three companies provided $110 billion of the $122 billion raised by OpenAI in its most recent funding round in March.
One reason OpenAI has postponed its IPO is that CEO Sam Altman couldn’t get the valuation he wanted. In other words, investors aren’t willing to pay just any price for shares of the firm. The sharply declining stock of SpaceX (SPCX) following its IPO is not helping OpenAI either.
3. Hyperscalers Cut Back AI Spending
OpenAI’s funding is highly dependent on the four major hyperscalers – Microsoft, Amazon (AMZN), Alphabet, and Meta Platforms (META) – to keep pumping money into AI investments. This year, these four players are estimated to spend some $725 billion in capital investments, borrowing significant money or issuing stock (or both) to do so.
But analysts expect even higher levels of spending in 2027, with Goldman Sachs predicting hyperscalers will make capital investments of $1.15 trillion. Even with rapidly rising profits at the hyperscalers – much of it from their AI investments that funnel back as sales to themselves – they’ll need to raise hundreds of billions of dollars in either debt or equity to reach Goldman’s estimates.
But the debt market is already less receptive to Big Tech’s spending spree. Recently, Amazon marketed a $25 billion debt offering and needed to raise the interest rates on its longest-term bonds. The offering attracted weak demand. Orders amounted to just 2.5 times the bonds on offer, compared with a more robust 3.2-times oversubscription in a March debt offering.
Whether hyperscalers cut back may have less to do with their own decision-making and more to do with what investors decide for them. The funding environment is less attractive. To appease the market, Amazon said it wouldn’t issue any more debt in 2026. At the same time, the cost of insuring hyperscalers’ bonds has been rising, suggesting investors are getting more nervous about it.
Signs of the AI Bubble
It’s important to remember that debt is keeping the AI bubble afloat for now, but we’re already seeing plenty of cracks in the market. As the old saying goes, “If something can’t go on forever, it won’t.” But bubbles tend to go on a lot longer than anyone ever thinks possible.
The signs of breakdown are tremendously obvious, a testament to the seemingly unshakeable belief that’s driving the market in the face of basic finance rules. And if this continues for much longer, it won’t end pretty.
Regards,
James Royal, PhD
Editor’s Note: “Frontier AI” is a point of no return when AI surpasses human intelligence and gains free will. Elon Musk warns this moment could hit by the end of 2026. When it happens, 60-year Wall Street legend, Marc Chaikin, predicts Frontier AI could soon become the only thing that determines which stocks make money and which grind to a halt. That’s why he’s giving away a list of stocks to buy and sell absolutely FREE to help you position your money for a virtually unrecognizable world driven by Frontier AI technology. Get Marc’s Frontier AI Hotlist of stocks right here…
