So, renowned investor and Hedge fund manager of the ‘Big Short’ fame, Michael Burry, has made a billion-dollar bet against AI… will he be proven right?
When the $500 billion Project Stargate was announced in January of this year, it elicited largely positive feedback from many tech pundits and industry leaders. After all, ChatGPT went from something no one had heard of to hundreds of millions of users in under three years –a growth trend that only promised to continue for years to come, both in terms of users and usage.
As expected, leaders with a vested interest in the project, such as Microsoft’s Satya Nadella and OpenAI’s Sam Altman, went all-in with their support. Other AI experts, such as Robert Scoble, noted that unless companies decide to channel all their data into a single pipeline, there isn’t enough data in the world to keep the chips busy at this scale. Most conservative geopolitical analysts in the US had mixed feelings about the project. On the one hand, they were elated that the US was unambiguously claiming its position as the global leader in AI; on the other, they worried about who was actually going to foot the bill for all the data centers. Elon Musk was perhaps the only big name who criticized Project Stargate for its scale and opacity—especially at a time when the US government was underfunded. Some experts in India remarked that, regardless of whether the entire sum is spent, the project signaled a major US commitment to making AI the next internet, aimed at rapid adoption; others raised concerns around data sovereignty and autonomy and alluded to various ways data can be weaponized against other countries.
Fast-forward nine months, and the general sentiment appears to be shifting in the opposite direction. What began as little more than internet murmur planting doubts about the overhyped profitability of LLMs has now reached the corridors of high finance, which are abuzz with fears that the AI cartel has overplayed its hand—and may end up dragging the entire market down with it.
Many analysts, investors and intellectuals have raised concerns regarding the current state of affairs in the AI space. The following are some of the major ones.
1. The Core Offerings: The core offering put forth by leading AI companies is a massive productivity boost—think 50× or even 100×. The immediate sales pitch is that any organization can become far more efficient than it ever dreamed possible. From a longer-term perspective, however, the ambitions are far loftier: Sam Altman of OpenAI has repeatedly stated that building AGI is the ultimate goal; Mark Zuckerberg at Meta has set his sights on an even more ambitious target—creating the world’s first superintelligent machine. Other major players have
articulated similar visions, though Alphabet (Google) has chosen a different path, pursuing quantum computing as its route to AGI. Unfortunately, AI companies are now facing serious questions on both fronts.
A new MIT study released in August 2025 rattled many tech-enthusiast investors. It revealed that, despite more than $40 billion invested in enterprise AI applications, over 95% of those projects have yielded zero financial returns—many have actually proved net-negative for the companies involved. In the United States alone, 42% of companies abandoned AI initiatives in 2025, up from just 17% the previous year. Managers who bet heavily on “disruptive AI” now admit they are seeing little to no real disruption. In a widely discussed article titled “AI- Generated ‘Workslop’ Is Destroying Productivity”, researchers from Stanford and Harvard highlight a now well-understood reality: LLM outputs remain superficial and lack the creative spark needed to solve complex problems meaningfully. In a recent interview, JP Morgan Chase CIO Lori Beer shared that the company has restricted employee use of ChatGPT. The near-term efficiency narrative is taking a severe beating. Worse still, corporate AI evangelists continue to insist that the pursuit of AGI or superintelligence is being held back solely by insufficient data- center scale—and that everything will fall into place once larger facilities come online. That claim has become a point of intense debate, especially as every successive leap in model size delivers sharply diminishing returns.
2. A Rocky Road to Profits:
Even if AI offerings eventually live up to the hype, a fundamental question remains: can these companies ever turn a profit large enough to justify their astronomical valuations?
OpenAI alone expects to burn through $115 billion in cash over the next four years while projecting revenues of $200 billion by 2030. Yet investors are deeply worried—ChatGPT remains heavily unprofitable even for users paying $200 a month for the premium plan. In a recent interview, OpenAI CFO Sarah Friar claimed that many customers would gladly pay $2,000 a month for full capabilities. Independent analyses from consultants and academics paint a far less rosy picture, however. For starters, there iswidespread agreement that never before has so much capital been poured into a single technology with suchan unproven track record of generating profits. A full- blown spending frenzy is underway. When asked about the possibility of an AI bubble, Mark Zuckerberg acknowledged the risk but stressed the far greater danger of missing the wave by failing to “invest enough”. Sam Altman has gone further, declaring that he wants to see trillions—yes, trillions— of dollars invested in AI infrastructure over the coming decade. So how exactly will investors get their money back? No clear answer has emerged.
Against a backdrop of underperforming enterprise projects, massive capital outlays, and uncertain revenue models, analysts foresee a foggy future. Bain & Company estimates that the leading AI players—OpenAI, Meta, Google, and NVIDIA—will need combined earnings of roughly
$2 trillion by 2030 simply to fund the computing power required to meet the projected demand. Their current combined annual revenue is just over $650 billion. If McKinsey’s “continued momentum” scenario (published in its August 2025 quarterly) proves accurate and total AI
investment reaches $7.2 trillion by the end of the decade, the industry will be under even greater pressure to deliver returns. All of this ignores intensifying competition from China and
the skyrocketing cost of electricity needed to keep the data centers running.
3.Inflated Profits: Investor Michael Burry—famous for “The Big Short”—has recently accused major AI-related companies of artificially inflating earnings through questionable accounting practices. His primary charge centers on the treatment of depreciation, particularly for semiconductor chips. Burry argues that hyperscale AI players, including Meta, Oracle, and Palantir, are significantly understating depreciation expense by extending the useful life of GPUs and custom accelerators to five or six years, when a more realistic period would be two to three years. This practice materially boosts reported earnings in the short term. To back his thesis, Burry has taken large put positions totaling more than $1.5 billion against these companies (and related stocks). Although markets have taken notice of both the allegations and his bets, there has—so far—been no widespread domino effect or forced restatements. This is largely because accounting standards still grant companies considerable latitude in determining the useful life of depreciable assets.
4.Circular Deal Making: Over the past year, leading AI companies have woven an intricate web of circular deals among themselves. At the center stand two dominant players: NVIDIA—the world’s most valuable company, currently valued at over $4 trillion—and OpenAI. NVIDIA recently agreed to co-invest roughly $100 billion in OpenAI’s massive new data-center build- outs. In return, OpenAI has committed to equipping those facilities almost exclusively with NVIDIA chips—a textbook example of circular deal-making. Similar arrangements are proliferating: OpenAI has signed a comparable contract with AMD under which it will purchase large volumes of AMD chips while AMD becomes one of OpenAI’s largest shareholders. This tightly interconnected network of cross-investments, revenue guarantees, and equity swaps now encompasses virtually every major name in the AI ecosystem, including Oracle, CoreWeave, Microsoft, NScale, Anthropic, Meta, xAI, and others.
Investors have always distrusted circular deals. As the old saying goes, money that merely sloshes around a closed loop eventually comes back to the same pockets—often after creating the illusion of explosive growth. The result is a self-reinforcing ecosystem that looks impressive on paper but raises serious questions about long-term sustainability and genuine economic value.
What worries the market most is the lack of independent, organic demand. This tangled web of cross-investments, revenue commitments, and equity swaps has dramatically increased systemic risk across the entire AI sector: the failure of one major player could quickly cascade into multiple others. In short, many of these companies now appear too interconnected to stand—or fall—alone. As one recent article in The Register pointedly observed, AI giants are effectively using investor capital to fund their own customers, who then turn around and spend that money back on the original suppliers. Unlike traditional vendor-financing arrangements, however, these customers are not yet seeing meaningful financial upside. Perhaps this explains why Sam Altman, during a recent investor call, cautiously floated the idea of a federal backstop for OpenAI’s massive investments—essentially a pre-emptive government lifeline. CFO Sarah Friar later walked back the suggestion, insisting Altman was not asking for a bailout but rather for some form of federal loan guarantees that would “de-risk” the sector as a whole. Whatever the wording, such remarks do little to inspire traditional investor confidence.
Will Michael Burry’s massive AI short positions pay off? No one can predict the future with certainty. What does seem inevitable, however, is that some form of correction—whether orderly or chaotic—is now looming over the AI sector.
The top five U.S. technology companies, all deeply committed to AI hyperscaling, now command a combined market capitalization exceeding $17 trillion—roughly 35–40% of the entire U.S. stock market. The federal government in general, and the second Trump administration in particular, has placed an enormous bet that these AI hyperscalers will deliver transformative economic growth for the country. Yet that wager is being made while the sword of Damocles—an unsustainable national debt—continues to hang precariously overhead. – Prof. S Abhijith