Jefferies: AI Capex to Consume 92% of Hyperscalers' Cash Flow, Powering DRAM Surge

Why It Matters

The projected 92% cash‑flow allocation to AI capex reshapes the financing landscape for cloud giants, potentially increasing reliance on debt markets and altering capital‑structure decisions. For investors, the heightened exposure of DRAM suppliers to AI spending creates a new risk‑return profile: while pricing power may lift margins, a sudden capex slowdown could trigger sharp earnings volatility. Beyond the immediate sector, the trend signals a broader macroeconomic impact. At $800 billion, AI‑related capex will consume a sizable slice of U.S. GDP, influencing fiscal policy discussions and prompting regulators to monitor systemic risk in technology‑focused credit markets. The concentration of DRAM supply also raises supply‑chain resilience concerns, with implications for any industry that depends on high‑performance memory, from automotive to data‑center services.

Key Takeaways

• Jefferies projects hyperscalers will spend 92% of operating cash flow on AI capex in 2026, up from 41% in 2023.
• Total AI‑related capex from major U.S. tech firms is forecast at $700 billion in 2024 and $800 billion in 2025.
• DRAM is expected to capture about 28% of hyperscalers' operating cash flow this year.
• Only three global DRAM suppliers remain, giving them unprecedented pricing leverage.
• Edison Lee warns that rising compute, memory and power costs keep sustainable profitability distant for pure‑model AI firms.

Pulse Analysis

Jefferies' warning reflects a classic case of capital intensity outpacing cash generation, a scenario that historically forces firms to tap external financing or trim growth ambitions. In the 2000s, similar dynamics unfolded in the telecom boom, where operators poured cash into network build‑out only to confront liquidity crunches when subscriber growth stalled. The current AI wave differs in that the underlying technology—high‑performance compute and memory—has fewer substitutes, which could sustain higher margins for DRAM makers even if capex slows.

However, the concentration of DRAM supply introduces a systemic vulnerability. With just three players, any production hiccup—whether from geopolitical tensions, natural disasters, or capacity constraints—could amplify price volatility and affect downstream cloud providers' cost structures. Investors should therefore monitor inventory levels, capacity expansion announcements, and the health of long‑term supply contracts as leading indicators of sector stability.

From a financing perspective, the near‑full‑cash‑flow deployment suggests hyperscalers may increasingly rely on revolving credit facilities or securitized debt to fund ongoing AI projects. This could tighten credit spreads for technology‑sector borrowers and push banks to reassess risk‑weighting for AI‑related loans. Market participants that can offer flexible, long‑duration financing—such as private credit funds—may capture a growing share of the funding pie, reshaping the competitive landscape of tech‑focused capital markets.