AI demand continues to grow, but raising the capital to build computing infrastructure is becoming more challenging. Meta lifted its 2026 capital expenditure outlook and said it plans to issue shares to support AI investment, while OpenAI lowered its target valuation for a proposed funding round. Together, the moves have sharpened investor focus on the cost and availability of financing across the AI industry. A rally in US technology stocks failed to hold last week: the Nasdaq 100 fell 0.79% and the S&P 500 slipped 0.47%, while the Dow Jones Industrial Average gained 0.41% as investors shifted toward defensive and value stocks.
Meta’s share sale puts AI spending under scrutiny
Meta reported revenue and profit above expectations, but also disclosed significantly negative free cash flow and raised its 2026 capital expenditure forecast. The company plans to issue shares to fund its AI buildout and is considering creating a new class of shares. It also assessed debt financing, but concluded that borrowing enough to meet its needs could put its credit rating at risk, prompting it to opt for equity instead.
Meta said funding needs at AI labs, SpaceX and major technology companies including Alphabet and Amazon are intensifying competition for capital. The company is raising funds early because it is concerned that other large borrowers could absorb available financing and make conditions tighter. The decision has drawn particular attention because Meta has substantial cash flow from advertising yet is choosing to issue equity in a challenging market, underscoring its concern about future access to capital.
Financing strains spread beyond neocloud companies
Until recently, investor concern centered mainly on the financing capacity of neocloud providers. These companies typically borrow heavily to buy Nvidia GPUs, then rent computing capacity to cloud providers and AI labs. Demand is not the most immediate constraint: Nvidia CEO Jensen Huang and analysts have pointed to strong demand for AI computing, while results from TSMC and ASML indicate that infrastructure construction is still accelerating.
The pressure instead comes from high capital requirements, debt burdens and business models that may prove vulnerable if markets weaken. CoreWeave’s credit spreads have widened to distressed levels, Firmus withdrew its IPO plans, and the proposed listing valuation for UK-based Nscale has faced questions. South Korea’s Elice recorded the worst first-day performance for a large local IPO. Nebius and IREN have each fallen about 30% to 40% over the past month. With interest rates still high and competition for capital increasing, the model of borrowing to buy GPUs and then leasing them out is becoming more expensive to finance.
OpenAI cuts valuation target as capital constrains expansion
OpenAI has lowered its target valuation for a planned funding round of about $30 billion to $1.25 trillion, below its initial proposed pre-money valuation of $1.4 trillion but still above its current valuation of about $852 billion. The company made the adjustment after investors challenged the original pricing. Demand for this round is also said to be weaker than for earlier rounds. Anthropic’s expected IPO next month could offer further insight into how markets value AI model companies’ cash burn, computing needs and path to profitability, although the listing plans remain uncertain.
AI companies can draw on operating cash flow, bond markets, private equity and private credit. Microsoft remains one of the few large cloud providers with positive free cash flow. Meanwhile, Alphabet, Amazon, Meta, Microsoft, SpaceX and Oracle have issued hundreds of billions of dollars in bonds in 2026, raising questions about how much more the investment-grade corporate bond market can absorb. Private investors’ willingness to keep backing AI projects is also uncertain: funds have already faced pressure this year from distressed software loans and lower equity valuations.
Financing structures involving Nvidia and Broadcom include off-balance-sheet vehicles, residual-value guarantees, computing purchase commitments, lease support and convertible notes. They are becoming one way for the industry to bring in additional capital. Executives at Broadcom and Intel have previously warned that the availability and cost of funding could constrain AI development. Huang has described the scarce resource needed to support growth as working capital, not just GPUs or electricity. The market debate now extends beyond access to chips and power to whether these financing structures can supply the capital required for continued expansion.