Compute & Market Power
Chips, cloud capacity, data centers, energy, capex, and the supply chains that determine who can build AI.
The AI race ultimately reaches the physical world: accelerators, memory, fabs, networking, data centers, electricity, cooling, and the financing required to assemble them. This topic explains how those constraints shape model availability, cloud competition, national strategy, and the bargaining power of builders.
Coverage follows the stack from semiconductor materials and export controls to hyperscaler capex and grid queues. The goal is to make infrastructure legible as an operating constraint and a source of market power—not to treat every chip launch or data-center announcement as an isolated event.
-
Physical AI Funding Hit $47.4B and Got Top-Heavy
-
ONEOK Turns AI Power Demand Into Gas Capex
-
GPU Kernel Benchmarks Can Be 55.8 Points Too Lenient
-
NVIDIA's Photonic Switch Prices the Network Layer
-
Oracle's AI Buildout Eats Its Payroll
-
SK hynix Breaks Ground on $3.87B HBM Plant
-
Intel's $20B Raise Is a Capacity Test
-
HD Hyundai Power Costs About 8.4% More per MW
-
Amazon’s 7.65 GW Shortcut Carries Permit Risk
-
Firmus Puts $2B of Equity Under a $10B Debt Bet