The AI Capex Trap: Tech Giants’ Cash Flow Under Strain as DRAM Makers Seize Control

New Delhi | May 2, 2026

The global race for Artificial Intelligence dominance has reached a critical financial juncture. According to a comprehensive report by global brokerage firm Jefferies, the massive infrastructure spending by U.S. hyperscalers is beginning to strain their financial health, even as memory chip manufacturers emerge as the new power brokers of the tech industry.

The 92% Surge: Capital Expenditure Reaches a Breaking Point

The report highlights a dramatic shift in how tech titans—including Microsoft, Alphabet, Amazon, and Meta—are allocating their capital. Jefferies projects that by 2026, these four major hyperscalers will spend a staggering 92% of their operating cash flow on capital expenditure (capex).

• The Rapid Climb: This is a massive jump from just 41% in 2023.
• Economic Scale: Total AI-related capex is expected to hit $700 billion this year and $800 billion in 2025. This investment accounts for nearly 2% of the US GDP and 30% of all US non-financial pre-tax profits.

DRAM Manufacturers: The Beneficiaries of Supply Constraints

While much of the AI hype focuses on software, the real “pricing power” has shifted to Dynamic Random Access Memory (DRAM) suppliers. As Moore’s Law—the principle that chip density doubles periodically—slows down, increasing supply has become fundamentally harder.

Key Structural Changes:

1. Market Consolidation: The industry has shrunk from 12 global suppliers in 2012 to just three dominant players today (Samsung, SK Hynix, and Micron).
2. Long-term Lock-ins: Fearing shortages, giants like Nvidia are now signing 3-5 year supply agreements. This shift makes the memory industry more stable and profitable, mirroring the “foundry model” of TSMC.
3. Revenue Shift: Hyperscalers are expected to allocate roughly 28% of their operating cash flow specifically toward memory chips this year.

The Monetization Dilemma: Is the Profit Sustainable?

Despite the “arms race” to build the biggest AI models, Jefferies raises a red flag regarding profitability. Edison Lee, Jefferies’ Head of China Tech, notes that the skyrocketing costs of compute, memory, and power mean that “sustainable profitability is far away” for companies focusing solely on AI models.

The report warns of a potential “reset” if investors or companies realize they have over-invested in infrastructure without a clear, immediate path to high-margin returns. However, in the long term, the structural demand for high-performance computing remains undisputed.

Disclaimer: This report is based on financial analysis and brokerage data from Jefferies. The information provided is for educational and informational purposes only and does not constitute financial advice. Investors should conduct their own research or consult with a certified financial advisor before making investment decisions.