Capital & Compute

Capital & Compute — A family of maintained AI-economics trackers rather than one dataset: an HBM price tracker carrying contract price per gigabit and derived price per gigabyte by HBM generation with dated revisions and forward forecasts, an adjacent RAM and DRAM price tracker, an AI model tracker with per-model release and pricing records plus a month-by-month release series, an inference-provider comparison covering 35 providers across five categories with representative pricing, and a country-level AI capability tracker.. Small independent publisher with unusually broad structured coverag…

Capital & Compute offers (Alternative, Price, Supply Chain) — 188 published analyses with per-topic counts printed on the page (130 AI costs, 86 models and benchmarks, 80 coding agents, 51 AI markets, 32 local AI and hardware, 25 compute infrastructure), plus tracker tables including 35 inference providers with pricing and a per-model release series — implying tens of thousands of individually dated structured rows across the tracker set..

AI accelerator bill-of-materials and inference cost modelling; memory allocation and wafer-margin analysis for the AI-versus-conventional-DRAM capacity fight; token-cost benchmarking across model and inference-provider options for buyers; a longitudinal record of model release and pricing velocity for market sizing; country-level AI capability index construction; and a dated structured corpus for grounding AI-economics research agents that need current pricing rather than stale training data.

The data is with 2 years of history.

Coverage spans US, APAC, UK; Semiconductors; alternative, price, supply_chain; equities.