Is the AI Infrastructure Capex Supercycle Accelerating?
Recent data confirms a massive surge in capital expenditure for AI infrastructure. Microsoft’s Azure revenue hit $100 billion annually in FY 2026, up 33% YoY, driven by AI-driven customer demand. NVIDIA is negotiating a $250 billion credit wrap to backstop OpenAI’s Ohio data center, with potential expansion to $350 billion for silicon. NextEra Energy and Brookfield Corporation partnered for a $100 billion AI data center in Kentucky, combining natural gas power with compute capacity. These moves signal a structural shift, not a cyclical boom. Quote: “AI computing may need 1,000x more energy than currently available” – NVIDIA CEO Jensen Huang.
Why Semiconductor Memory Is in a Supercycle
The demand for high-bandwidth memory (HBM) is exploding. SK hynix’s HBM4 began shipping in Q2 2026, with NVIDIA securing a $500B+ deal for wafer capacity. Seagate’s revenue jumped 50% YoY to $3.6 billion, guided for $4.1 billion in Q1. NAND flash prices more than doubled in H2 2026 as AI data centers outpace supply. Stat: TrendForce projects AI data centers will account for 44% of NAND demand in 2026, rising to 51% by 2027.
Tech Earnings Divergence: Winners and Losers
Not all tech companies are benefiting equally. Meta Platforms reported 24% YoY revenue growth in Q2 2026, while Alphabet’s cloud capex rose to $195-205 billion annually. CoreWeave, despite a $740M net loss, secured $35.2 billion in Meta contracts. The gap highlights divergent strategies: some companies are building AI infrastructure, others are monetizing it. Quote: “Meta’s operating margins compressed from 48% to 41% on GPU costs” – Meta CFO.
Quality Compounders: Who’s Winning Long-Term?
Companies with durable moats and consistent growth stand out. Microsoft’s Azure cloud backlog reached $678 billion, up 84% YoY. Amazon’s $200 billion 2026 capex for data centers positions it for 15%+ annual growth. NVIDIA’s CUDA platform and HBM supply deals create barriers to entry. Stat: NVIDIA’s revenue is 54% concentrated among three hyperscalers, but its ecosystem lock-in ensures sustained demand.
Energy Transition & Nuclear: Powering the AI Future
AI’s energy demands are reshaping energy markets. Cameco’s uranium price surged to nearly $100/lb as nuclear capacity projections doubled to 900 GWe by 2030. NextEra and Brookfield’s $100 billion AI data center in Kentucky uses natural gas turbines. Quote: “AI could consume up to 12% of U.S. electricity by 2030” – Bloom Energy projections.
Macro Crosscurrents: Rates, Oil, and Volatility
The Fed’s hawkish stance and oil price swings are creating volatility. Rates held at 3.5-3.75% despite inflation warnings, while Brent Crude hit $90/barrel after Iran’s missile attack. Energy stocks like Chevron rose 2.28%, but tech faces margin pressure from higher borrowing costs. Stat: Goldman Sachs forecasts U.S. data center power demand will double from 31 GW to 66 GW by 2027.
Investment Takeaways
Focus on companies directly enabling AI infrastructure: NVIDIA, Microsoft, and energy providers like Cameco. Avoid overvalued names with high capex burn without clear pathways to profitability. Quote: “The AI infrastructure buildout is shifting toward a leveraged, vendor-financed credit cycle” – Michael Burry.