From the mine to the mind — every layer of the AI economy, live. The audited value chain from energy and raw materials up to the models and apps, priced by the daily worker feed.
Two answers now, and July separated them. PRICING power still sits where it did: ASML's EUV monopoly, TSMC leading-edge plus CoWoS, the Cadence/Synopsys EDA duopoly, the SK Hynix/Micron/Samsung HBM oligopoly, Nvidia's CUDA/NVLink lock-in, and firm interconnected baseload under long-dated PPAs. EQUITY value in July accrued somewhere else — to the hyperscaler-distributors bundling AI into seats customers already pay for (Microsoft +24.6%, Amazon +15.3%), to the accelerator duopoly, and to the de-rated application layer bid up as money rotated out of hardware. The middle of the stack — memory, wafer-fab equipment, interconnect, thermal, merchant GPU rental — was sold as one basket regardless of any individual name's pricing power. Pre-revenue option plays still carry theme exposure and no present pricing power: uranium equities fell ~11% in July while U3O8 spot held ~$86/lb, the cleanest illustration of the distinction.
Firm energized power + multi-year grid interconnection — the stack's true binding constraint; new baseload can't be plugged in faster than queues, NRC licensing, and enrichment allow, so incumbents with interconnected assets price at scarcity.
Nvidia (NVDA) — owns the accelerator layer via the CUDA + rack-scale NVLink moat, taxing every training and inference run regardless of which lab or cloud wins.
11 layers, from energy (L1) up to apps (L11), each sized by the combined market value of the public companies in it. Open a layer or sub-stack for QAI’s take; click any company for its conviction, bull vs bear and value-chain role.
Only US-listed public names are sized; private / foreign names are shown for context but not counted — so a bottleneck layer can read smaller than it really is.
Live prices & fundamentals from the QAI worker feed. Foreign / pre-IPO names shown as context, not investable calls. Not investment advice.