The Deep Dive
Big Tech's AI infrastructure investment has reached escape velocity. In 2025 alone, four hyperscalers—Alphabet, Amazon, Meta, and Microsoft—deployed $350 billion in capex, with Amazon alone committing $200 billion for 2026, the majority earmarked for AI infrastructure. The sector's cumulative capex nearly tripled from $162 billion in 2022 to $448 billion in 2025. Yet this unprecedented capital mobilization is colliding with a supply-side reality that no amount of spending can immediately overcome: geopolitical fragmentation of energy and semiconductor supply chains.
The Strait of Hormuz crisis crystallizes the problem. Following the closure on March 4, 2026, Brent Crude surged past $120 per barrel, with QatarEnergy declaring force majeure on all exports. This isn't merely an oil price story—it's an energy cost shock propagating directly into AI infrastructure economics. Data center power consumption markets are already valued at $70.59 billion by 2035 projections, with power distribution units alone commanding 20% market share. A sustained $120+ oil environment raises the marginal cost of backup power generation, cooling systems, and grid stabilization infrastructure that hyperscalers depend on for redundancy.
Simultaneously, semiconductor supply chains are undergoing structural realignment. TSMC's $165 billion US expansion and India's notification of the Dholera chip SEZ signal a deliberate decoupling from concentrated Asian foundry capacity. But this diversification carries a hidden cost: redundant capex, longer qualification timelines, and geographically distributed supply chains that are harder to optimize. Europe's semiconductor vulnerabilities amid Middle East conflict underscore that geography now matters more than efficiency.
The math is brutal. If hyperscalers need $3 trillion through 2028 to support AI infrastructure, and geopolitical fragmentation adds 10-15% to per-unit infrastructure costs through redundancy, supply chain diversification, and energy hedging, the true capex requirement could exceed $3.3 trillion. That's not a rounding error—it's a structural shift in the cost basis of AI compute. The market is pricing in Moore's Law and energy abundance. It's getting geopolitical fragmentation and energy volatility instead.
Signal Watch
Energy demand acceleration outpacing grid capacity: AI-driven data center electricity demand skyrocketed 50% last year, far exceeding the 3% growth in global electricity demand. The IEA projects demand will triple by 2030, but grid infrastructure cannot scale at that velocity—especially when geopolitical disruptions constrain fuel supply and raise backup power costs.
Power management infrastructure becoming a capital-intensive bottleneck: The data center power management market is forecast to reach $45.32 billion by 2035, growing at 7.4% CAGR from a 2025 base of $23.04 billion. This is no longer a commodity input—it's a strategic constraint requiring dedicated capex and vendor lock-in.
Onsite battery storage emerging as grid hedge: The IEA identified onsite battery storage as critical for next-generation AI data centers, positioning them as grid assets with the right incentives. This is a tacit admission that centralized grid reliability cannot be assumed—hyperscalers must now self-insure against supply shocks.
The Bottom Line
Watch for hyperscaler capex guidance revisions in Q2 earnings. If any of the Big Four signal slower infrastructure deployment due to energy cost inflation or supply chain delays, it signals that the $3 trillion thesis is already encountering friction. The real risk isn't that AI buildout stops—it's that it becomes more expensive, more geographically fragmented, and more dependent on state-level energy partnerships (Middle East, Africa, Central Asia) that introduce new political and currency risks.
Bitcoin Macro
Brent Crude above $120 per barrel increases the marginal energy cost for all compute-intensive operations, including Bitcoin mining. Higher oil prices typically correlate with higher electricity costs in regions dependent on thermal generation. Miners operating at the margin face margin compression; those with access to stranded renewables or geopolitical safe-haven jurisdictions gain relative advantage. Watch for hashrate consolidation toward jurisdictions with stable energy policy and away from regions exposed to Middle East supply shocks.
Capital flows toward infrastructure, but geopolitics determines where it lands.