Citi Sees Big Tech Spending Even More on AI Next Year—Here's How Much

Citigroup analysts have significantly raised their 2025 forecast for hyperscaler AI infrastructure spending to $490 billion, up from $420 billion, citing strong demand and recent cloud deals. This increased buildout is increasingly financed by debt rather than cash flows, exemplified by Oracle's recent $18 billion bond sale, introducing new vulnerabilities such as default and interest rate risk to the AI boom. While some express concerns about an 'AI bubble,' particularly regarding deals like OpenAI's with Nvidia, Citi highlights growing enterprise adoption and clear monetization prospects for AI services as a key differentiator from past speculative periods, benefiting infrastructure providers like Nvidia.

Analysis

Citigroup analysts have materially increased their 2025 AI infrastructure spending forecast for hyperscalers to $490 billion, a significant rise from the previous $420 billion estimate and slightly above Wall Street consensus. This heightened spending projection, driven by urgent enterprise AI adoption and a flurry of recent partnerships, reinforces a bullish outlook for core infrastructure suppliers, prompting Citi to raise its price target on Nvidia (NVDA) to $210. However, a critical development noted is the market's transition from a cash-flow-funded to a debt-funded investment cycle, introducing new vulnerabilities such as default and interest rate risk. This is exemplified by Oracle's (ORCL) recent $18 billion bond sale to expand cloud capacity after two quarters of negative cash flow, with its capital expenditures projected to nearly triple to $58 billion by fiscal 2027. While novel financing structures, like OpenAI's equity-for-systems deal with Nvidia, have sparked concerns about a speculative bubble reminiscent of the dot-com era, Citi analysts highlight a key differentiator: a tangible 'off-ramp' created by growing external demand and clear monetization prospects for AI services from enterprises.

Key Decisions for Investors

• Investors should recognize the sustained bullish thesis for AI infrastructure providers like Nvidia, as evidenced by rising capex forecasts, while remaining aware of the concentrated nature of hyperscaler demand.
• It is prudent to increase scrutiny of the balance sheets and cash flow statements of hyperscalers, particularly Oracle, as the shift to debt-financed expansion elevates their sensitivity to credit market conditions and interest rate changes.
• Monitor the pace of tangible enterprise-level AI adoption and monetization, as this is the key factor differentiating the current investment cycle from a speculative bubble; any slowdown in real-world AI revenue generation could signal a significant shift in market sentiment.