Adaptation Curve

Prediction Ledger

Dated, sourced predictions about the AI supply chain — tracked to resolution. Every call is either confirmed correct, partially right, wrong, or unresolvable, with reasoning. The ledger is the memory.

30 total0 scored30 pending

Pending

If TSMC fully met Jensen's demand, Nvidia could sell ~$2-3T of GPUs in 2026-27 (but would trigger overbuild)

pending
Gavin Baker2026-05-20mid termtarget: 2027-01-01source

Terafab (SpaceX/Tesla + Intel) succeeds in building the world's largest fab in America

pending
Gavin Baker2026-05-20end of_decadesource

OpenAI and Anthropic each exceed ~$200B ARR this year via usage-based pricing + more compute

pending
Gavin Baker2026-05-20nowtarget: 2026-12-31source

If Anthropic were unconstrained on compute it would already be doing ~$100-200B vs ~$50B ARR

pending
Gavin Baker2026-05-20nowsource

Frontier tokens keep capturing the majority of model-layer economic value

pending
Gavin Baker2026-05-20mid termsource

A 'bitter-lesson' violation via ASI self-optimization is the single biggest risk to the AI trade

pending
Gavin Baker2026-05-20mid termsource

Continual learning arrives soon, producing a fast takeoff

pending
Gavin Baker2026-05-20mid termsource

Disaggregation of prefill/decode extends GPU useful lives to ~10-15 yrs and lowers financing toward 5-6%

pending
Gavin Baker2026-05-20mid termsource

Frontier labs eventually stop releasing top models via API (new prisoner's dilemma)

pending
Gavin Baker2026-05-20mid termsource

Nvidia could build a near-frontier in-house model whenever it wants (though likely won't)

pending
Gavin Baker2026-05-20mid termsource

Amazon shows real P&L efficiencies from robotics in retail over the next ~18 months

pending
Gavin Baker2026-05-20mid termtarget: 2027-01-01source

Google stays well-positioned (most compute, YouTube/robotics data, search); Google I/O is a key Pareto-frontier test

pending
Gavin Baker2026-05-20nowtarget: 2026-12-31source

Near-term power/watts shortage begins to ease as new energy sources come online

pending
Gavin Baker2026-05-20mid termtarget: 2028-01-01source

Orbital compute (racks in space linked by lasers) becomes real and solves power long-term, mainly for inference

pending
Gavin Baker2026-05-20end of_decadesource

TSMC's capacity decisions are the single most important variable for whether AI gets a bubble

pending
Gavin Baker2026-05-20nowsource

TSMC stays in a 'Goldilocks' zone — expands enough to block a >30%-share second source, not enough to overbuild

pending
Gavin Baker2026-05-20mid termsource

Intel or Samsung eventually breaks capacity discipline, forcing others to follow

pending
Gavin Baker2026-05-20mid termsource

~half of new power capacity will be behind-the-meter

pending
Dylan Patel2026-03-13end of_decadetarget: 2030-01-01Behind the metersource

Anthropic & OpenAI each reach ~5-6 GW

pending
Dylan Patel2026-03-13nowtarget: 2026-12-31source

Anthropic & OpenAI each reach ~10 GW

pending
Dylan Patel2026-03-13mid termtarget: 2027-12-31source

~700 EUV tools installed -> ~200 GW/yr AI-chip ceiling

pending
Dylan Patel2026-03-13end of_decadetarget: 2030-01-01source

China fully indigenizes DUV; EUV 'working' but not volume

pending
Dylan Patel2026-03-13end of_decadetarget: 2030-01-01source

Smartphone volumes fall to ~800M then 500-600M

pending
Dylan Patel2026-03-13mid termtarget: 2027-12-31source

Memory prices double or triple again

pending
Dylan Patel2026-03-13mid termtarget: 2027-12-31source

Space data centers don't make economic sense this decade

pending
Dylan Patel2026-03-13end of_decadetarget: 2030-01-01source

Sam Altman wants 52 GW/yr (~1 GW/week)

pending
Sam Altman2026-03-13end of_decadetarget: 2030-01-01source

Elon Musk wants 100 GW/yr of compute in space

pending
Elon Musk2026-03-13mid termtarget: 2029-01-01source

Google must double AI serving capacity every 6 months

pending
Google2026-03-13nowsource

Elon Musk: build a 'TeraFab' — 1M wafers/month

pending
Elon Musk2026-03-13end of_decadesource

ASML/EUV becomes the #1 AI-compute scaling constraint

pending
Dylan Patel2026-03-13end of_decadetarget: 2029-01-01source