AI Investment Discussion Intensifies
Discussion regarding the nature of artificial intelligence (AI) investments has intensified, with some industry figures expressing confidence in a sustained growth cycle while others raise concerns about potential overvaluation. Jensen Huang, CEO of Nvidia, stated on a recent earnings call that the company observes a different trend than an "AI bubble."
White House AI advisor and venture capitalist David Sacks described the current period as an "investment super-cycle." Investor Ben Horowitz commented that current demand, supply, and growth multiples do not indicate a bubble. JPMorgan Chase executive Mary Callahan Erdoes characterized the capital flow into AI as representing a "major revolution" in company operations, dismissing the concept of a bubble.
Conversely, Paul Kedrosky, a venture capitalist and research fellow at MIT, expressed doubt regarding the current state of the AI industry. He noted a significant capital inflow into what he termed a "mostly speculative" revolution, arguing that the pace of technological improvement has "ground to a halt."
Capital Inflow and Market Skepticism
The rate of capital investment into AI has drawn attention from financial experts. OpenAI, the developer of ChatGPT, has reported $20 billion in annual revenue and plans to invest $1.4 trillion in data centers over the next eight years. This growth projection assumes increasing sales of its AI services.
However, research suggests that most firms have not observed a direct impact on their financial performance from chatbots, and one analysis indicates that 3% of individuals pay for AI services. Daron Acemoglu, an MIT economist and Nobel laureate, stated that while future AI technologies will add value, much of the current industry rhetoric involves "exaggeration."
Amazon, Google, Meta, and Microsoft are projected to collectively invest approximately $400 billion in AI this year, primarily for data centers. Some of these companies plan to allocate about 50% of their current cash flow to data center construction. To manage cash expenditure, companies like Meta and Oracle are utilizing private equity and debt to finance data center development.
Debt Financing and Special Purpose Vehicles
An assessment by Goldman Sachs analysts indicated that hyperscaler companies have incurred $121 billion in debt over the past year, representing a 300% increase from the sector's typical debt levels.
Analyst Gil Luria of D.A. Davidson investment firm noted that some financial arrangements in Silicon Valley are structured to avoid reporting debt on company balance sheets, often through the use of "special purpose vehicles" (SPVs). In these structures, a tech firm invests in a data center, outside investors provide most of the capital, and the SPV borrows funds to acquire necessary chips. This allows the tech company to benefit from increased computing capacity without adding the debt to its balance sheet.
An example involves an SPV funded by Wall Street firm Blue Owl Capital and Meta for a data center in Louisiana. Blue Owl secured a $27 billion loan for the facility, backed by Meta's lease payments. Meta holds a 20% ownership stake in the entity and accesses all computing power generated by the data center. The $27 billion loan is not reflected on Meta's balance sheet. Should the data center fail, Meta would be obligated to make a multi-billion-dollar payment to Blue Owl for the facility's value.
Luria compared the use of such financial arrangements to those employed by Enron, an energy company that collapsed in 2001, noting that while current practices are transparent, their reliance for future development warrants scrutiny.
Revenue Projections and Market Stability Concerns
Companies are incurring this new debt based on expectations of substantial future AI revenues. Morgan Stanley analysts estimate that major tech companies will spend approximately $3 trillion on AI infrastructure through 2028, with their existing cash flows projected to cover only half of this amount.
Luria warned that if the growth of the AI market stabilizes, there could be an oversupply of capacity, potentially rendering the debt worthless and causing financial institutions to incur losses. He drew a parallel to the dot-com bubble 25 years ago, when debt-financed fiber-optic cables were built for a future demand that had not yet materialized.
Intercompany Transactions and Investment Patterns
The AI investment landscape also includes intercompany transactions that have drawn attention. A $100 billion deal between Nvidia and OpenAI involves Nvidia investing capital into OpenAI to fund data centers, which in turn will be equipped with Nvidia's chips. Some analysts suggest this structure, where Nvidia effectively subsidizes a major customer, may influence perceived demand for AI. Kedrosky noted the scale of such "circular deals" as unusual, comparing it to practices during the dot-com bubble.
Lesser-known companies have also entered this market. CoreWeave, originally a crypto mining startup, has transitioned to building data centers for AI. OpenAI has entered into multi-billion-dollar agreements with CoreWeave, renting chip capacity in exchange for CoreWeave stock. OpenAI could potentially use this stock to cover its rental fees. Nvidia, which holds an ownership stake in CoreWeave, has an agreement to acquire any unused data center capacity through 2032.
Acemoglu commented that these types of deals could eventually reveal an unstable financial foundation.
Investor Reactions and Executive Statements
Some prominent investors have adjusted their positions regarding AI companies. Peter Thiel reportedly sold his entire stake in Nvidia, valued at approximately $100 million, while SoftBank sold a nearly $6 billion stake in the company. Michael Burry, known for his bet against the housing market in 2008, has taken a position against Nvidia, alleging the AI industry employs accounting practices and circular deals. Burry has previously issued market warnings with mixed outcomes. He stated that "True end demand is ridiculously small. Almost all customers are funded by their dealers."
Several tech executives have acknowledged aspects of market exuberance. OpenAI CEO Sam Altman stated in August that investors are likely "overexcited about AI" while also affirming AI's long-term importance. Google CEO Sundar Pichai told the BBC that "elements of irrationality" are present in the current AI market and that "no company is going to be immune" if the market experiences a downturn, including Google.