H100$6.39/hr 1.2% 7d
A100 80GB$2.45/hr 0.5% 7d
H200$10.29/hr 0.8% 7d
L40S$1.28/hr 0.3% 7d
T4$0.24/hr 0.6% 7d
L4$0.45/hr 1.1% 7d
H100$6.39/hr 1.2% 7d
A100 80GB$2.45/hr 0.5% 7d
H200$10.29/hr 0.8% 7d
L40S$1.28/hr 0.3% 7d
T4$0.24/hr 0.6% 7d
L4$0.45/hr 1.1% 7d
Weekly Pulse
Tool

AI Adoption Curve

Where does the AI compute market sit on its S-curve? This tool plots cumulative AI compute spend against four long-run TAM scenarios and shows the corresponding Rogers diffusion stage.

Long-run TAM scenario

Base — sustained growth, AI compute reaches scale of current global cloud.

Time window

Cumulative AI compute spend

$446.6B

% of selected TAM

44.7%

Diffusion stage

Early Majority

Against the $1T TAM scenario, cumulative AI compute spend through 2026-Q2-MTD sits at 44.7% of the long-run ceiling — placing the market in the Early Majority band of the Rogers diffusion curve.

We are here (2026-Q2-MTD)Cumulative spend (actuals)S-curve overlay (logistic, $1T TAM)Signwl data boundary (Jan 2026)Milestone event

Milestones in view

  • 2022-11-30ChatGPTOpenAI launches ChatGPT. Inflection point — Q2-Q3 2023 NVDA data-centre revenue triples.
  • 2023-03-14GPT-4OpenAI releases GPT-4. Frontier model arms race begins; hyperscaler capex commitments accelerate.
  • 2024-03-18BlackwellNVIDIA announces Blackwell (B100/B200/GB200). Forward orders extend NVDA backlog through 2026.
  • 2025-01-20DeepSeek R1DeepSeek R1 launches at low cost on Ascend silicon. First credible non-Nvidia frontier model — inference deflation begins.
  • 2026-04-24DeepSeek V41.6T-parameter MoE at 27% of V3 FLOPs. Per-token inference cost compresses to $1.74/M tokens.
  • 2026-05-17IREN/NVDAIREN signs $3.4bn 5-year Nvidia contract — Childress, Texas, 60 MW Blackwell deployment from early 2027.
Track the curve

Get the Weekly Pulse — Signwl's read on where AI compute sits and what moves the needle.

A weekly synthesis of investable hypotheses and the underlying pricing tape — derived from Signwl's proprietary GPU pricing data. No fluff, no spam, unsubscribe anytime.

Methodology

Spend proxy. AI compute spend is approximated by quarterly NVIDIA Data Center segment revenue (2019 Q1 through 2025 Q4), the cleanest single-source proxy for global AI compute spend given NVIDIA's dominant accelerator share. From Q1 2026 onwards the series transitions to Signwl's proprietary read derived from the live pricing tape (blended hourly rental rate × estimated active capacity × utilisation factor), cross-checked against NVDA's continued growth.

Cumulative series. Plotted on the chart is the running sum of quarterly spend — the cumulative invested AI compute capital. This is the curve typically compared to an S-curve.

S-curve overlay. The dashed purple line is a logistic fit scaled to the selected long-run TAM. It uses standard logistic-diffusion parameters with a midpoint anchored to early 2028. The fit is illustrative — it is meant to show the shape of the curve under each TAM scenario, not to forecast specific values.

Diffusion stage. The stage label (Innovators / Early Adopters / Early Majority / Late Majority / Laggards) is read from the % of TAM penetrated using the Rogers diffusion of innovations thresholds: 2.5%, 16%, 50%, 84%, 100%.

Milestones. Annotated events on the chart are drawn from public announcements (ChatGPT, GPT-4, Blackwell, DeepSeek launches) through 2025; from Q1 2026 onwards, events surface from Signwl's daily investment briefs and weekly pulses (IREN-NVDA contract, etc.).

This tool is provided for informational purposes only and does not constitute investment, financial, or other professional advice. TAM scenarios are illustrative, not forecasts.

Research

Weekly synthesis, daily briefs, monthly reports.

Signwl GPU Pricing Index

Live composite GPU pricing index.

GPU ROI Calculator

Model your GPU investment returns.