Hyperscaler Capex Just Hit $725 Billion: The Infrastructure Buildout Is Now Bigger Than Most Economies

Hyperscaler AI capex spending reached $725 billion in 2026 with 77% year-over-year growth, signaling that the AI infrastructure cycle has entered a phase where capital deployment, not chip supply, is the binding variable.

The Week in Supply: Nvidia's $81.6B Quarter Reshapes Hyperscaler Capex Math

Record GPU revenue and $725B hyperscaler capex guidance signal sustained AI infrastructure demand, but supply chain strain and packaging bottlenecks are forcing suppliers to accelerate spend.

Nvidia's $81.6B Quarter Exposes the Real Bottleneck: Not Chips, But Supply Chain Velocity

Record GPU revenue masks accelerating strain on packaging, memory, and component suppliers as product cycles compress faster than supply chains can follow.

Iran's Strait Blockade Just Made Chip Shortages a Gulf Energy Crisis

With the Strait of Hormuz closed for 11 weeks, Taiwan and South Korea's fabs are running on borrowed LNG and helium—and the chip shortage is actually an energy shortage wearing a different mask.

The Week in Supply: TSMC's CoWoS Breakthrough Masks a Deeper Capacity Crisis

Advanced packaging yields soar while foundry capacity constraints force hyperscalers to delay AI data center buildouts by 7 GW—a supply chain fragmentation that threatens to reshape the entire semiconductor ecosystem.

AMD's Samsung Bet Signals TSMC's Advanced Node Monopoly Is Cracking

With TSMC's 2nm capacity locked until 2028, AMD's defection to Samsung for Venice and Verano CPUs marks the first major foundry shift in the AI era—and threatens the supply-chain assumption that underpinned the entire hyperscaler capex surge.

The Week in Supply: Foundry Wars Heat Up as Apple Hedges TSMC

Intel and Samsung advance 2nm capacity while Apple explores alternatives, reshaping the semiconductor supply chain hierarchy in 2026.

Apple's TSMC Escape Plan Hits Reality: Why Intel and Samsung Can't Close the Foundry Gap

As Apple explores diversifying away from TSMC, new reports reveal the scale and consistency problem that makes foundry competition nearly impossible—and what that means for the entire AI chip supply chain.

The Week in Supply: $725B in Hyperscaler Commitments, One Foundry Controls the Pace

TSMC dominates advanced node capacity while AMD and Intel fight for scraps; hyperscaler capex surge collides with EUV tool constraints through 2027.

TSMC's A16 Slip Signals the Foundry Roadmap Is Breaking Under AI Demand

TSMC delayed its A16 node from late 2026 to 2027 at its tech symposium, revealing that even the world's most advanced fab cannot keep pace with hyperscaler appetite—and signaling margin pressure ahead.

The Week in Supply: Power Management Chips Become the Forgotten Bottleneck

As AI servers surge, power IC shortages eclipse compute—and metrology tool delays expose fragile fab qualification timelines across Taiwan, Korea, and the U.S.

The $3 Trillion AI Infrastructure Bet Is Running Into Geopolitical Reality

As hyperscalers commit $350B+ annually to AI buildout, Middle East energy disruptions and semiconductor supply chain fragmentation are forcing a reckoning with the true cost of decentralized compute.

The AI Infrastructure Supply Chain: A Field Guide to Who Builds What

A free resource for investors, semiconductor analysts, energy market participants, and data center operators.

The Week in Supply: Energy Becomes the Binding Constraint

Data center power demand is reshaping semiconductor capex, geopolitical oil shocks are spiking miner energy costs, and regulatory resistance is hardening across U.S. states.

Power, Not Chips, Is Now the AI Bottleneck—and It's Reshaping Infrastructure Economics

As data center electricity demand surges 220% by 2030 and the Hormuz blockade pushes oil above $100/barrel, energy constraints—not semiconductor scarcity—are becoming the binding constraint on AI expansion.

The Week in Supply: Hyperscale Bet, TSMC Standard, CoreWeave's $21B Moment

Meta commits $21B to CoreWeave through 2032 as hyperscalers consolidate two-thirds of global data center capacity by 2031.

The 100-Kilowatt Problem: AI's Energy Density Is Breaking the Grid Before It Breaks the Bank