Compare growth in AI models, research, repositories and products with semiconductor revenue, cloud GPU pricing and marketplace availability.
Each series is independently normalised to 0–100. Tooltips show original values.
Monthly models, papers, repositories, product launches and cumulative stars across tracked repositories.
Quarterly reported Nvidia and AMD company revenue. Annual totals are excluded to avoid double counting.
Average cloud and marketplace GPU prices plus the number of rentable Vast.ai offers. Higher prices or fewer offers can indicate stronger demand or constrained supply.
Data are fetched automatically using AI agents from trustworthy internet public sources. Models are counted by publication date, papers are summed across tracked AI research categories, repositories are counted across tracked AI topics, and Product Launches counts tracked AI products. Daily observations are summed into calendar months.
Tracked Repo Stars is different from the new-output counts: for each repository and month, the latest available cumulative star snapshot is retained, then those repository totals are added together. Research categories and repository topics can overlap, so these series are directional measures of public activity rather than unique-object totals or measures of quality.
Nvidia and AMD revenue uses quarterly reported company revenue and is displayed in US dollars. Records marked as full-year totals are excluded so the same revenue is not counted once in quarterly figures and again in an annual total. Revenue reflects the whole company and is therefore a broad indicator of accelerator supply activity, not a direct count of GPUs manufactured or shipped.
AWS Spot is the monthly average across tracked GPU instance spot prices. Azure and Lambda are monthly averages across the tracked GPU virtual-machine or instance prices. Vast.ai Price averages the recorded median hourly price across GPU models and daily snapshots.
Vast.ai Offers first sums rentable offers across GPU models for each daily snapshot, then averages those daily marketplace totals across the month. It measures publicly rentable listings, not idle GPUs or utilisation. Prices and offer counts are point-in-time market observations; changes may reflect model mix, provider inventory, pricing decisions, supply constraints or demand.
The comparison and AI Output charts independently transform each series to a 0–100 index using (value − series minimum) ÷ (series maximum − series minimum) × 100. This allows counts, prices, offers and revenue to share one visual scale; it does not make their units equivalent. Tooltips show original values. Every chart, summary card and chart-visibility check excludes the newest month because its observations may still be incomplete.