Nvidia, the world's most valuable company by market capitalization, failed to quell Wall Street's concerns about an artificial intelligence bubble despite posting record-breaking revenue. More than 90% of its sales depend on data centers where hyperscalers—big tech companies operating cloud businesses—are investing with borrowed money.
The company's purchase commitments to semiconductor manufacturers surged sixfold over the past year, raising concerns that Nvidia has positioned itself precariously between upstream suppliers and downstream customers.
According to CNBC on February 26 (local time), Michael Burry, CEO of Scion Asset Management and the real-life protagonist of the film "The Big Short," highlighted in a post on newsletter platform Substack that Nvidia's purchase commitments had jumped from $16.2 billion (approximately 23.31 trillion won) to $95.2 billion (approximately 135.99 trillion won)—a sixfold increase in one year.
Purchase commitments represent quantities that Nvidia, a fabless semiconductor design company, has promised to buy from chip manufacturers including TSMC, Samsung Electronics, and SK hynix. Total supply obligations, including purchase commitments and inventory, amount to approximately $117 billion—equivalent to Nvidia's annual operating cash flow.
"Nvidia has committed to purchasing large quantities of products before accurately gauging future demand, tying up more cash for extended periods," Burry wrote.
Burry drew parallels between Nvidia's purchase commitment structure and Cisco during the dot-com bubble era. He recalled how Cisco entered into massive purchase commitments only to write off $2.5 billion worth of inventory in 2001 when demand collapsed.
"What's happening to Nvidia right now is not temporary risk," Burry warned.
Wall Street believes that if the debt-fueled investments of Nvidia's customers—Amazon, Google, Microsoft, Meta, and Oracle—collapse, the purchase commitments Nvidia made to semiconductor manufacturers would become stranded inventory. Hyperscalers are essentially buying Nvidia's graphics processing units (GPUs) with borrowed money rather than profits generated from AI. Should credit problems emerge among hyperscalers, the impact would cascade through Nvidia to Korean semiconductor manufacturers.
Wall Street estimates hyperscalers' actual capital expenditure this year will reach approximately $700 billion (about 1,000 trillion won). This was the first question investment banks raised during Nvidia's earnings call. However, Nvidia CEO Jensen Huang remained optimistic, stating that hyperscalers "currently have a computing infrastructure base of $300 billion to $400 billion, but this is still at a low level."
Nvidia's results also failed to exceed market expectations by a significant margin, dampening Wall Street enthusiasm. Fourth-quarter revenue announced the previous day only slightly exceeded the $66.2 billion forecast compiled by financial information firm LSEG. With Nvidia now recording all-time high revenue for 11 consecutive quarters, Wall Street's sensitivity has diminished. Meanwhile, analysts reacted more sharply to negative developments including surging memory semiconductor prices, difficulties re-entering the Chinese market, and reduced investment in OpenAI.
"Given the company's growth trajectory, investors' standards are currently too high," said Adam Phillips, Chief Investment Officer at EP Wealth Advisors. "It's becoming increasingly difficult to impress Wall Street."
On February 26, the day after its earnings announcement, Nvidia plunged 5.46% in a single session, erasing recent gains. The daily decline was the largest since April 16 last year, when news emerged that the U.S. administration would restrict exports of Nvidia's H20 chips to China—a 10-month high. Nvidia shares have fallen 0.9% year-to-date. According to The Wall Street Journal, Mark Spiegel, who leads New York hedge fund Stanphyl Capital Partners, recently shorted Nvidia stock.
However, substantial counterarguments maintain that Nvidia remains the leading company in AI hardware. According to LSEG, 61 of 66 analysts who recently wrote reports on Nvidia issued "buy" or "strong buy" recommendations.