Michael Burry shorts Nvidia as artificial intelligence bubble concerns grow

Michael Burry has opened a short position against Nvidia, focusing market attention on mounting financial and structural pressures within the artificial intelligence sector, according to financial reporting. The investor's latest market move mirrors his historic contrarian bet against the United States mortgage market. Industry observers note that a vast proportion of stock market gains since the year began has relied heavily on artificial intelligence enterprises. Yet confidence in these valuations is beginning to face scrutiny over underlying infrastructure costs and sustainable revenue models.

The wider financial climate has seen volatility following the collapse of hedge funds that bet exclusively on market growth, such as the fund managed by Leopold Aschenbrenner, which dropped significantly. Despite broader market jitters, Nvidia has maintained its standing as one of the world's most valuable firms, valued around $5tn and frequently swapping places with Apple. Driven by strong demand for advanced microprocessors that allow artificial intelligence to perform complex tasks faster, the chipmaker commands premium prices for its hardware.

Media additions

Proponents of the hardware giant often invoke the gold rush idiom regarding picks and shovels, arguing that equipment suppliers profit regardless of which artificial intelligence applications ultimately succeed. However, scepticism surrounds the long-term durability of this model. Startups like Anthropic and OpenAI face difficulties in building a competitive moat, as features are rapidly copied by cheap, open rivals that businesses can run at a much lower cost, causing billions spent on research and development to vanish almost instantly.

Furthermore, Michael Burry has identified a new vulnerability for Nvidia, pointing to a startup chip rival called Etched that he says can produce 10x the performance at lower cost per die — referring to the manufacturing cost to produce a single functional microchip. Venture capital is increasingly flowing into semiconductor companies addressing the inference bottleneck, which involves calculations performed after a user sends a prompt. British companies Fractile and Olix, alongside a startup touted by bullish artificial intelligence pundit Andrew Côté, are among those operating in this space.

Nvidia's primary competitive defence rests on a socket interface or application programming interface called CUDA, which developers have relied on to build tools and models for two decades. Comparisons have been drawn to historical industry strategies, such as IBM designing socket interfaces for mainframe equipment or Microsoft defining application programming interfaces to create impregnable defences through compatibility. However, potential shifts are emerging. China, despite a United States embargo, forbids its developers from using CUDA to avoid dependency, while an artificial intelligence infrastructure agent called Infinity uses teams of agents to optimise models on specific hardware, a process that could potentially render CUDA redundant.

Questions also linger over the immense financial commitments fueling data centre construction. According to The Economist, the data centre splurge requires $2.5tn in annual artificial intelligence revenue just to break even, while former International Monetary Fund and Goldman Sachs analyst Peter Berezin suggests the required figure could reach $10tn. Both analysts agree that current demand cannot match this scale of investment, leaving data centres to depreciating rapidly while experimental agentic artificial intelligence applications remain unreliable. Innovations identified by analysts and market participants threaten to turn these facilities into expensive stranded assets.

Beyond traditional financial metrics, prominent technology leaders maintain distinct ideological frameworks regarding humanity's technological trajectory, as detailed in an analysis published by Time. OpenAI Chief Executive Officer Sam Altman articulated early concerns about artificial intelligence during a startup conference, stating,

"I think that AI will probably, most likely, sort of lead to the end of the world. But in the meantime, there will be great companies created with serious machine learning."

Sam Altman, Chief Executive Officer, OpenAI, via Time Altman's perspective has since shifted toward viewing technological advancement and uncapped demand for human capability as a moral obligation, as outlined in a blog post.

Elon Musk, another prominent industry figure, has repeatedly voiced fears of an artificial intelligence apocalypse, referencing a terminator scenario during a high-profile legal dispute. Musk's worldview incorporates multiplanetary colonisation as a hedge against existential risks, including artificial intelligence. Legal battles involving OpenAI's structure concluded when an Oakland jury unanimously dismissed Musk's case against Altman, OpenAI, and president Greg Brockman after Judge Yvonne Gonzalez Rogers intervened during court proceedings regarding extinction discussions.

Meanwhile, entrepreneur Peter Thiel has argued that caution and regulation represent apocalyptic threats, describing resistance to technological advancement in lectures and criticising international oversight proposals. Market analysts remain divided on the timeline for any potential market correction, weighing whether infrastructure spending has outpaced realistic consumer demand or if computing capacity will continue to command premium pricing as these competing visions and financial pressures collide.