Indicators
One object, two addresses. Unpublished rows have a definition and data but do not yet meet the publishing rules (a status, counterevidence, and either two sources or one primary source), so they carry no status and no page.
| Indicator | Bucket | Layer | Status | Conf | Grade | Latest | Obs |
|---|---|---|---|---|---|---|---|
| Rent concentration in semiconductors | — | Compute & physical | concentrating | 85 | grade A | 75.0%as of 2026-07-26(2 obs) | 128 |
| Cloud backlog (remaining performance obligations) | — | Compute & physical | concentrating | 80 | grade A | $638Bas of 2026-05-31 | 73 |
| Frontier-lab revenue run-rates | — | Model | concentrating | 55 | grade C | $40.0Bas of 2026-08-13 | 34 |
| Lab recoupment ratio | — | Model | emerging | 50 | grade C | 0.22×as of 2026-08-13(12 obs) | 27 |
| Consumer surplus from generative AI (willingness to accept) | — | Labour & consumers | dispersing | 65 | grade C | $172Bas of 2026-03-31 | 4 |
| Application layer's share of enterprise generative-AI spend (Menlo) | — | Deployment & application | emerging | 55 | grade B | 51.4%as of 2025-12-31(2 obs) | 4 |
| Semiconductors' share of stack operating income | — | Compute & physical | concentrating | 70 | grade A | 57.3%as of 2026-06-30(8 obs) | 42 |
| Cloud segment operating margins | — | Compute & physical | stable | 75 | grade A | 39.4%as of 2026-06-30(2 obs) | 20 |
| Circular financing scale | — | Compute & physical | concentrating | 75 | grade A | $903Bas of 2026-08-17(14 obs) | 14 |
| Concentration among covered chip filers (HHI) | — | Compute & physical | unclear | 60 | grade A | 0.85 indexas of 2026-06-30(2 obs) | 13 |
| Concentration among covered cloud filers (HHI) | — | Compute & physical | dispersing | 60 | grade A | 0.35 indexas of 2026-03-31(3 obs) | 18 |
| Concentration of business AI spend across labs (Ramp, HHI)unpublished Not yet measurable: the direction rule needs four monthly readings of Ramp's per-lab paid shares and only July 2026 exists; each month arrives as a manual row from Ramp's letter. | — | Model | no status until published | — | grade B | 0.46 indexas of 2026-07-31(3 obs) | 3 |
| Public cloud software valuation multiples (Clouded Judgement) | — | Deployment & application | emerging | 50 | grade B | 4.40×as of 2026-09-04 | 40 |
| Price per unit of capability (cheapest METR-hour on OpenRouter) | — | Model | not yet measurable | 40 | grade A | $0.30as of 2026-09-10(32 obs) | 32 |
| Frontier labs' share of tracked venture dollars | — | Model | stable | 50 | grade A | 94.8%as of 2026-06-30(32 obs) | 102 |
| Vertical versus horizontal application funding | — | Deployment & application | dispersing | 45 | grade A | 2.30×as of 2026-03-31(11 obs) | 19 |
| Frontier lab valuation to run-rate multiple | — | Model | emerging | 45 | grade C | 28.3×as of 2026-06-16(10 obs) | 48 |
| Neolab capital raised per dollar of revenueunpublished Not measurable from free sources: the neolabs' rounds are filed or reported, but none of them has a revenue figure in Epoch or a filing, so the denominator does not exist. | — | Model | no status until published | — | unmeasured | 0 | |
| Lab vertical integration (acquisitions and acquihires, trailing four quarters) | — | Model | concentrating | 55 | grade A | 10as of 2026-09-30(11 obs) | 16 |
| Venture flow to revenue lagunpublished Not yet measurable: the estimate needs at least eight quarters of both venture dollars and revenue for the same sub-layer; the venture series starts in 2023 and revenue exists only for the frontier labs. | — | Model | no status until published | — | unmeasured | 0 | |
| Open-weights gap to the closed frontier (ECI points) | — | Model | dispersing | 55 | grade A | 5.77 indexas of 2026-08-20(3 obs) | 141 |
| Non-Nvidia share of AI compute shipped (H100-equivalents) | — | Compute & physical | emerging | 50 | grade B | 26.0%as of 2026-03-31(3 obs) | 51 |
| Hyperscaler capex to AI revenue (trailing year) | — | Compute & physical | dispersing | 40 | grade A | 2.51×as of 2026-06-30(16 obs) | 48 |
| Hyperscaler server useful lives (10-K estimates) | — | Compute & physical | not yet measurable | 65 | grade A | 5.00 yearsas of 2025-12-31 | 4 |
| Neocloud borrowing cost (CoreWeave facility rates) | — | Compute & physical | not yet measurable | 60 | grade A | 10.5%as of 2026-06-30(4 obs) | 12 |
| Enterprise multi-model usage share (Menlo) | — | Model | not yet measurable | 45 | grade B | 60.0%as of 2023-11-10 | 3 |
| Startups' share of enterprise AI application revenue | — | Deployment & application | emerging | 50 | grade B | 63.0%as of 2025-12-31 | 2 |
| Frontier lab gross marginunpublished Not measurable from free sources: lab gross margins are reported only by The Information and Sacra behind paywalls; the paid stubs exist and are off. | — | Model | no status until published | — | unmeasured | 0 | |
| Compute share of lab spendingunpublished Not measurable as a series: Epoch published a one-off estimate range, not a table; a manual row can carry the point estimate once quoted. | — | Model | no status until published | — | unmeasured | 0 | |
| Application gross margin net of inference (model tax)unpublished Not measurable from free sources: needs SemiAnalysis unit inference costs (paid) joined to reported ARR and Ramp token spend; the paid stub is off. | — | Deployment & application | no status until published | — | unmeasured | 0 | |
| GPU rental price per hourunpublished Not measurable from free sources: the SemiAnalysis index is paywalled; the paid stub is off. | — | Compute & physical | no status until published | — | unmeasured | 0 | |
| Secondary-market marks between roundsunpublished Not measurable from free sources: secondary marks are behind Caplight, Forge and EquityZen paywalls; the paid stub is off. | — | Model | no status until published | — | unmeasured | 0 | |
| Revenue per employee in AI-exposed sectorsunpublished Not measurable yet: no free source pairs sector revenue with employee counts at a usable cadence; BLS industry mappings to exposure are unverified. | — | Adopters | no status until published | — | unmeasured | 0 | |
| Model-layer token concentration on OpenRouter (HHI, top table) | — | Model | not yet measurable | 40 | grade B | 0.11 indexas of 2026-09-09(19 obs) | 19 |
| METR 50% time horizon | Methods | Model | faster than normal | 80 | grade A | 17.4 h(8.5 h–55.1 h)⚑disputedas of 2026-04-07 | 30 |
| METR 80% time horizon | Methods | Model | faster than normal | 70 | grade A | 3.1 h(1.6 h–6.6 h)as of 2026-04-07 | 26 |
| 50%/80% horizon ratio | Methods | Model | consistent with normal | 70 | grade A | 6.34×as of 2026-03-05(2 obs) | 50 |
| US labour productivity, year on year | Adaptation | Labour & consumers | consistent with normal | 80 | grade A | 2.2%as of 2026-06-30 | 76 |
| US total factor productivity, private nonfarm business | Adaptation | Labour & consumers | consistent with normal | 75 | grade A | 0.8%as of 2025-12-31(2 obs) | 9 |
| Developer productivity uplift (METR RCTs) | Products | Deployment & application | consistent with normal | 70 | grade A | -4.0%⚑disputedas of 2026-02-24 | 2 |
| Enterprise pilots with measurable P&L impact | Products | Deployment & application | consistent with normal | 55 | grade C | 5.0%⚑disputedas of 2025-06-30 | 2 |
| Share of work hours assisted by generative AI | Early adoption | Adopters | consistent with normal | 80 | grade B | 6.3%as of 2026-06-30 | 27 |
| US firms using AI (Census BTOS) | Early adoption | Adopters | consistent with normal | 70 | grade A | 22.4%(21.7%–23.1%)as of 2026-08-09 | 21 |
| US businesses paying for AI (Ramp AI Index) | Early adoption | Adopters | emerging | 60 | grade B | 56.1%as of 2026-08-31 | 2 |
| Agent-workdays per human workday in frontier research (self-reported) | Return arrow | Model | emerging | 30 | grade D | 3.10×as of 2026-08-15 | 1 |
| Human interventions on 4–8 hour agent tasks (self-reported) | Return arrow | Model | emerging | 30 | grade D | 50.0%as of 2026-07-31 | 1 |
| Safety brakes inside the labs (events, trailing year) | Adaptation | Model | emerging | 40 | grade D | 3as of 2026-09-01(3 obs) | 3 |
| Frontier training compute growth (doubling time) | Methods | Model | emerging | 60 | grade B | 170 daysas of 2025-07-09(66 obs) | 66 |
| Training power draw growth (doubling time) | Methods | Compute & physical | emerging | 50 | grade B | 506 daysas of 2026-08-07(213 obs) | 213 |
| Hardware price-performance growth (doubling time) | Methods | Compute & physical | consistent with normal | 40 | grade B | 1282 daysas of 2025-11-06(20 obs) | 20 |
| Inference price at fixed capability (halving time) | Methods | Model | emerging | 45 | grade B | 68 daysas of 2024-12-13(6 obs) | 6 |
| Public text data exhaustion year (Epoch projection) | Methods | Training input | emerging | 40 | grade B | 2028 yearas of 2024-06-06 | 1 |
| ARC-AGI-2 best score | Methods | Model | emerging | 55 | grade B | 95.0%as of 2026-09-03(203 obs) | 203 |
| Continual-learning ladder (highest production rung) | Return arrow | Training input | emerging | 45 | grade D | 3.00 rungas of 2026-05-27(2 obs) | 2 |
| FDA-authorised AI-enabled devices, growth | Adaptation | Adopters | consistent with normal | 70 | grade A | 26.7%as of 2026-06-30(2 obs) | 102 |
| US state AI bills introduced (legislative year) | Adaptation | Adopters | emerging | 55 | grade B | 1035as of 2025-12-31 | 2 |
| Active RL-environment vendors (directory count) | Return arrow | Training input | emerging | 35 | grade B | 24as of 2026-07-15 | 1 |
| Ord half-life misfit (observed 50/80 ratio vs constant hazard) | Methods | Model | consistent with normal | 60 | grade A | 2.04×as of 2026-03-05(2 obs) | 50 |
| FDA-authorised devices built on a large language model | Adaptation | Deployment & application | consistent with normal | 60 | grade A | 1as of 2025-12-23(3 obs) | 3 |
| Reliability-science framework adoptionunpublished Not measurable as a number: adoption is a yes/no per system card, recorded as evidence rows rather than a series. | Products | Model | no status until published | — | unmeasured | 0 | |
| Agent calibrationunpublished Not measurable yet: no public leaderboard reports calibration for agents; HAL and tau-bench report success, not confidence. | Products | Model | no status until published | — | unmeasured | 0 | |
| Agent operational safety incidentsunpublished Not measurable yet: incidents are anecdotal; no incident registry with a denominator exists. | Products | Model | no status until published | — | unmeasured | 0 | |
| Deployment data share of trainingunpublished Not measurable: no lab discloses the provenance mix of its training data; recorded only if a system card states it. | Return arrow | Model | no status until published | — | unmeasured | 0 | |
| Employment in AI-exposed occupations, year on year | Adaptation | Labour & consumers | consistent with normal | 80 | grade C | -0.2%as of 2026-06-30 | 3 |
| Entry-level employment shortfall in AI-exposed occupations | Adaptation | Labour & consumers | emerging | 70 | grade B | 19.0%as of 2026-06-30 | 4 |
| Exposed vs unexposed occupation employment gap (Revelio) | Adaptation | Labour & consumers | emerging | 65 | grade B | -6.0%as of 2026-07-31 | 1 |
| California unemployment claims from AI-exposed occupations | Adaptation | Labour & consumers | consistent with normal | 70 | grade A | 1.0%as of 2026-07-31 | 2 |
| Recent-graduate unemployment rate (NY Fed) | Adaptation | Labour & consumers | consistent with normal | 75 | grade A | 5.6%as of 2026-06-30 | 1 |
| Cross-tracker concordance (labour) | Adaptation | Labour & consumers | consistent with normal | 70 | grade B | 0as of 2026-07-31(4 obs) | 4 |
| Augmentation vs automation share (Anthropic Economic Index) | Early adoption | Adopters | consistent with normal | 60 | grade B | 51.4%as of 2026-05-31 | 11 |
| US labour share of income, year on year | Adaptation | Labour & consumers | emerging | 65 | grade A | -3.4%as of 2026-06-30(2 obs) | 38 |
| Customer-support productivity uplift from a generative-AI assistant | Adaptation | Adopters | consistent with normal | 75 | grade B | 14.0%as of 2023-04-30 | 2 |
| Executives reporting no AI impact on their own firm | Adaptation | Adopters | consistent with normal | 70 | grade C | 90.0%as of 2026-02-28 | 2 |
| Frontier models completing expert legal tasks end to end (Harvey LAB) | Products | Deployment & application | emerging | 40 | grade D | 7.1%as of 2026-05-26 | 2 |
| Agent reliability across repeated runs (tau-bench pass^k) | Products | Model | emerging | 55 | grade B | 56.2%as of 2026-05-18 | 4 |
| Progress against the AI 2027 quantitative predictions | Methods | Model | emerging | 50 | grade B | 75.0%as of 2026-07-31 | 3 |
| Work-related share of ChatGPT consumer messages | Early adoption | Adopters | emerging | 65 | grade B | 27.0%as of 2025-06-30 | 2 |
| Usage as a share of theoretical exposure | Early adoption | Adopters | consistent with normal | 60 | grade B | 0.49×as of 2026-06-30(2 obs) | 9 |
| Workers highly exposed to AI with low adaptive capacity | Adaptation | Labour & consumers | emerging | 45 | grade B | 3,300,000 workersas of 2026-01-31 | 1 |
| Occupational-mix dissimilarity since ChatGPT (Yale Budget Lab) | Adaptation | Labour & consumers | consistent with normal | 65 | grade B | 5.26 ppas of 2026-03-01 | 103 |
| Chatbot effect on earnings and hours in exposed occupations (Denmark) | Adaptation | Labour & consumers | consistent with normal | 80 | grade B | 0.0%as of 2024-12-31 | 3 |
| Freelance writing earnings after ChatGPT (Upwork) | Adaptation | Labour & consumers | consistent with normal | 70 | grade B | -5.2%as of 2023-12-01 | 3 |
| Wage premium for AI skills (PwC AI Jobs Barometer) | Adaptation | Labour & consumers | emerging | 50 | grade C | 62.0%as of 2025-12-31 | 2 |
| Expert-data market run-rate (Mercor) | Methods | Training input | faster than normal | 70 | grade C | $2.0Bas of 2026-06-30 | 2 |
| Internal-to-public deployment gap at the frontier | Products | Model | consistent with normal | 45 | grade C | 4.00 monthsas of 2026-06-30 | 2 |
| Waymo paid rides per week | Adaptation | Deployment & application | consistent with normal | 65 | grade C | 500,000 rides per weekas of 2026-03-27 | 2 |