DVC State of AI · Patch notes

What's newAugust 4, 2026

August 3 lens: the model market became a barbell, the funding mix changed, and healthcare gave us the cleanest natural experiment of the year. Anthropic passed OpenAI on both valuation and US enterprise penetration; premium agentic tiers priced up to $10/$50 while capable output fell to $1–$6, with token efficiency emerging as a third economic axis; Chinese open weights took the parameter-count crown; the accelerator monopoly cracked in a single quarter; hyperscaler capex rose to ~$725B at the midpoint and is now 32% debt-funded with Alphabet's free cash flow negative for the first time since 2004; Cursor was acquired for $60B all-stock; and outcome pricing acquired a published price list. The deck expanded from 15 to 18 slides — adding Silicon & Power, Voice AI and Healthcare AI. The Healthcare AI section was substantially updated for the first time since it shipped. Previous entry (May 19) retained below.

Earlier lens (May 19, 2026): agents became infrastructure. Microsoft and AWS shipped the governance and toolkit layers that turn agent platforms into managed enterprise products. NIST/CAISI's first independent eval of DeepSeek V4 Pro shows the open-weight gap is real on held-out benchmarks while cost efficiency holds up — the month-gap figure previously printed is not in the source and has been removed. The Pentagon's IL6/IL7 vendor list and NVIDIA's equity-backed IREN deal complete the picture: frontier AI is now a national-security supply chain and an AI-factory build-out priced in gigawatts. The main report's structure and the new Healthcare AI section are untouched.

Updated August 4, 2026 · primary sources cited per item · older entries below the August 4 and August 3 blocks

August 4, 2026 · fact audit

Accuracy

Full fact audit: 306 material claims checked, outdated legacy cards retired, services matrix rebuilt

A line-by-line audit of every figure in the report and the deck — 306 material claims across index.html, slides.html, slides-data.js, attack-angles.js, healthcare-data.js, app.js and this page, with 91 source URLs fetched and read. Four things changed structurally.

1. Legacy cards were retired rather than layered over. The August blocks had been added above March–May 2026 material that was never removed, so the document contradicted itself: Anthropic appeared at both $965B and $380B, Cursor at both $4B and $2B+ ARR, ElevenLabs at both $600M and $330M, and slide 13 still called Claude Opus 4.7 the flagship while slide 6 priced Claude Fable 5. Each of those is now single-valued. The Ramp enterprise-penetration figure — the most-repeated new claim in the refresh — was three months stale in five files and is now Anthropic 42.4% vs OpenAI 39.5%, cited to Ramp directly instead of through an SEO aggregator. Anthropic’s run-rate is now shown as a time series — $47B in June, up from the official $30B in April — not as a source conflict.

2. The services matrix was rebuilt and single-sourced. It now lives in one data file consumed by both the report and slide 14, so the two can no longer drift. The $1.5T+ mapped headline and all four quadrant dollar totals are gone: the quadrants summed to ~$2.29T against the headline, and Sequoia publishes no quadrant totals. Attribution is narrowed to what Sequoia actually states. Three companies that are no longer independent private challengers (Sana → Workday, Uizard → Miro, Navan → public) were removed and are disclosed separately; duplicate categories and duplicate entities were consolidated; sixteen stale funding figures were refreshed with per-cell dates and source URLs; and every hard-truncated adoption statistic with a severed citation link was removed.

3. Unsupported claims were deleted rather than hedged. Where a second search pass produced no source that states the claim, the claim or its number is gone: the 92%-of-the-Fortune-500 figure, the ~$1.2T Anthropic secondary mark, the $1.3B → $158.7B AV projection, the unattributed layoff-regret quote, the asset-manager concentration statistic, Qwen3.8-Max-Preview in full, DeepSeek V4 active-parameter splits, Grok 4.5 and Gemini 3.6 Flash pricing mechanics, and the exact figures behind several healthcare, export-control and defence-procurement specifics. Groq is no longer described as an acquisition anywhere: it was a $20B non-exclusive technology licence plus asset purchase and acqui-hire, and NVIDIA states it did not acquire Groq.

4. Source and date labels are current everywhere. Capex is reconciled to ~$725B midpoint / up to ~$745B across index, slides, charts, tooltips and the data file, with 2025 at ~$402–410B and $224B labelled as the 2024 figure it is. Waymo is 11 public driverless metros on ~3,871 vehicles with rides flat since March; Tesla’s paid mileage fell ~36% quarter over quarter, and the 22 NHTSA crash reports now sit beside the “zero notable incidents” claim. NOHARM is 100 cases, not 1,100, and the paper’s severe-harm finding is included. The EU AI Act timeline agrees with itself on the two clocks. Cerebras is stated on one basis throughout. Etched is disclosed as a DVC portfolio company, correcting a disclaimer that contradicted DVC’s own public statement. Entity-level facts were promoted into a single keyed data file so this class of contradiction cannot recur.

Updates: every section of the report, all 18 slides, and the underlying data files Jump to report → 306 rows audited, 0 unresolved. Source hierarchy applied throughout: first-party filings and company newsrooms, then regulators, then wires and major business press, then named research firms — with SEO aggregators replaced wherever an alternative existed.

August 3, 2026 · superseded by the August 4 fact audit

Superseded snapshot

The August 3 refresh introduced the new narrative; the August 4 audit replaced its unsupported figures

The August 3 release added the model barbell, funding-mix, Silicon & Power, Voice AI and Healthcare AI narratives. A line-by-line audit the following day found that several precise model, pricing, healthcare, physical-AI and services figures rested on secondary or inaccessible sources. Those figures are not repeated here. The current report, the 18-slide deck and the August 4 entry above are the source of record.

Status: historical narrative preserved; superseded numbers retired Read the August 4 audit entry ↑

May 19, 2026 · previous entry

Report

New AI in Healthcare section now in the report

The report now includes a standalone AI in Healthcare section that treats healthcare as value chains, reimbursement paths, incentives, liability, regulation, and data flows, not just "models for clinicians." The section itself was not edited in this run; mentioned here so readers know where to find it.

Updates: Healthcare AI Jump to section →
Models & Evals

NIST/CAISI eval: DeepSeek V4 Pro measurably behind US frontier on held-out benchmarks

CAISI's independent evaluation used non-public ("uncontaminated") benchmarks across cyber, software engineering, sciences, abstract reasoning, and math. Public-style evals look close to the US frontier; CAISI's held-out evals do not. V4 Pro lands roughly comparable to GPT-5 (released about 8 months earlier), with an IRT Elo near 800 versus GPT-5.5 around 1260. The cost-efficiency story holds: V4 Pro was cheaper than GPT-5.4 mini on 5 of 7 benchmarks. The cleanest framing: public evals look close, held-out evals show a gap, cost efficiency is real.

Updates: Model Wars, Geopolitics Jump to section → Source: NIST/CAISI, May 2026
Agents

Agents became infrastructure: Microsoft Agent 365 GA, AWS Agent Toolkit, Anthropic managed-agent features

Microsoft Agent 365 went GA at $15/user/month as the enterprise governance and control plane for agents (discovery of shadow agents, identity, observability, policy). AWS made the Agent Toolkit and AWS MCP Server generally available, treating agent tooling as a first-class AWS interface alongside the SDKs. Anthropic shipped Dreaming (cross-session memory consolidation in research preview), Outcomes (rubric-graded loops), and Multi-agent Orchestration on its managed-agent platform. A2A protocol crossed 150+ organizations with Google, Microsoft, and AWS integration. Net: there is now a real "agent control plane" layer above models.

Updates: Agent Revolution, Agent Stack Jump to section → Source: Microsoft, May 1 2026 · AWS, May 6 2026 · VentureBeat, May 12 2026 · Linux Foundation, Apr 9 2026
Business Models

Outcome pricing is the direction; hybrid usage pricing is the reality

Coatue's "services as software" framing reads AI as a re-pricing of the software stack from per-seat to per-output, with the cited 3.8% consumption share of traditional SaaS as the gap between narrative and commercial reality. Orb's State of AI Agent Pricing reinforces the same point from a billing-system vantage: most production agent companies are on hybrid pricing today (subscription floor plus usage and outcome components). Outcome pricing is the strategic direction, hybrid/usage pricing is current commercial reality, and subscriptions are increasingly used as a revenue floor inside hybrid models rather than as the value-capture engine.

Updates: Business Models Jump to section → Source: Coatue C:\Takes, May 15 2026 · Orb, 2026
Geopolitics

DoD IL6/IL7 classified AI agreements: eight named vendors, GenAI.mil platform

The Department of War's May 1 release named eight vendors cleared for IL6 (Secret) and IL7 (most sensitive compartmented) deployment on GenAI.mil: OpenAI, Google, NVIDIA, Microsoft, AWS, SpaceX, Oracle, and Reflection AI. The Pentagon framed this as a diversity-of-supply move including open-source weights. Brookings put DoD's potential AI contract value at $90.7B in 2026 and 98.9% of all federal AI spending. The exclusion narrative around Anthropic is covered separately in earlier reporting; we are not re-attributing causality here.

Updates: Geopolitics Jump to section → Source: Department of War, May 1 2026
AI Factories

NVIDIA / IREN: 5 GW DSX pipeline + equity warrant for up to 30M IREN shares

NVIDIA and IREN announced a multi-part deal: a $3.4B five-year managed GPU cloud contract at IREN's Childress, Texas campus; a five-year warrant for NVIDIA to buy up to 30M IREN shares at $70 (about $2.1B notional); and an intent to build up to 5 GW of DSX-aligned AI factory capacity across IREN's pipeline, with the 2 GW Sweetwater campus on the Texas grid as the flagship. The structural read: NVIDIA is now using equity to anchor preferred infrastructure partners, extending DSX as the reference architecture for AI factory construction.

Updates: $700B Sprint Jump to section → Source: NVIDIA, May 7 2026

May 13, 2026 · previous patch notes

Report

AI in Healthcare section: no changes in this run

The Healthcare AI section in the main report was added in a previous refresh and is intentionally untouched this week. Mentioned here only so readers know the section is current as of its original publish and was reviewed, not rewritten, on May 13.

Updates: None — reference only Jump to section →
Coalitions

Anthropic / SpaceX / Cursor: a dependency web, not a vertical empire

Anthropic adds SpaceX compute (more than 300MW and over 220,000 NVIDIA GPUs, per Anthropic's own framing) on top of Amazon for primary training and Bedrock distribution, Google/Broadcom for TPU capacity, and Microsoft/NVIDIA for Azure capacity. SpaceX/xAI is simultaneously circling Cursor, one of the coding distribution points where Anthropic's models are economically powerful (Cursor and GitHub Copilot together represented roughly 30% of an earlier Anthropic revenue milestone, per reporting). The result is a checks-and-balances structure: everyone gets stronger, everyone gets more exposed.

Updates: Incumbents Jump to section →
Cloud

OpenAI / Microsoft exclusivity reset — and OpenAI's multi-cloud stress

Under the Apr 27 partnership reset, Microsoft is no longer OpenAI's exclusive cloud and the IP license through 2032 is non-exclusive; OpenAI continues a (capped) revenue share to Microsoft through 2030 while Microsoft no longer pays revenue share to OpenAI. The strategic cost of cutting the tether is real: OpenAI now has to assemble compute and distribution from multiple patrons (Microsoft, Oracle, AWS, plus its own buildout) without a single dominant protector. The Incumbents section now frames OpenAI/Microsoft/Oracle/AWS as one of three coalitions under stress, alongside the Anthropic/Amazon/Google/SpaceX/Cursor diversified-but-interdependent stack and Meta's separate vertically controlled stack.

Updates: Incumbents, $700B Sprint Jump to section →
Cleanup

Model-version and stale-number reconciliation across report and deck

Aligned model labels and tooltips across index.html, slides.html, slides.js, and slides-data.js so the report and deck no longer contradict each other on the OpenAI / Claude / Gemini lineup (GPT-5.4 + GPT-5.5 default Instant, Claude Opus 4.7, Gemini 3.1 Pro). The inference-cost collapse window is reconciled to 36 months end-to-end (2023–2026). The Manus → Meta acquisition card was softened to a reported deal with no in-repo primary source. The OpenClaw narrative is now an explicit rise-and-fall arc (Jensen-era hype → systematic absorption by Anthropic). CapEx is anchored to ~$695B midpoint / up to $725B top end with per-company numbers consistent between report and deck. Rolling phrases like "this week", "active talks", and "doubled in 3 months" were replaced with dated snapshots. "Microsoft restarted Three Mile Island" softened to the 20-year PPA / targeted-restart framing.

Updates: Model Wars, $700B Sprint, Agent Revolution, Slides Jump to section →

May 6, 2026 · earlier patch notes

Compute

Anthropic rents SpaceX's Colossus 1 capacity in Memphis

Anthropic signed for the full compute capacity of Colossus 1 (300+ MW), with the deal reportedly extending to interest in space-based compute. Rival labs are now willing to rent each other's factories — compute scarcity beats ideology.

Updates: $700B Sprint Jump to section → Source: CNBC, May 6 2026
Deployment

OpenAI and Anthropic stand up PE-backed deployment vehicles

OpenAI raised $4B+ for The Deployment Company at a $10B valuation (TPG, Brookfield, Bain) with access to 2,000+ portfolio clients; Anthropic was reported as forming a parallel adoption entity with Blackstone, Hellman & Friedman, Goldman, Apollo, GIC, and Sequoia. Frontier labs are copying the Palantir FDE model and becoming deployment companies, not just model APIs.

Updates: Agent Revolution, Business Models Jump to section → Historical May 2026 note. The unavailable financing link and its numeric terms were retired from the live report during the Aug 4 fact audit.
Models

GPT-5.5 Instant becomes the default ChatGPT model

Replaces GPT-5.3 Instant: AIME 2025 jumps 65.4 → 81.2, MMMU-Pro 69.2 → 76, with lower hallucination in law, medicine, and finance and preserved low latency. API exposed as chat-latest. The default model is compounding quietly at consumer scale.

Updates: Model Wars Jump to section → Source: TechCrunch, May 5 2026
Cloud

Microsoft / OpenAI partnership goes non-exclusive

Microsoft remains primary cloud partner but OpenAI can now serve products on any cloud; the Microsoft IP license through 2032 is non-exclusive; OpenAI revenue share to Microsoft continues through 2030 (capped) while Microsoft no longer pays revenue share to OpenAI. AI cloud exclusivity has cracked — distribution and compute are becoming multi-cloud.

Updates: $700B Sprint, Incumbents Jump to section → Source: Microsoft / OpenAI, Apr 27 2026
CapEx

Hyperscaler 2026 CapEx resets toward $725B

As reported on May 6, 2026 — superseded by the ~$725B midpoint / up to ~$745B guidance in the August 4, 2026 entry above. Big 4 hyperscalers plus Oracle guided to roughly ~$695B midpoint, up to $725B at the top end — about $100B above prior projections. Per-company anchors used across the report and deck (CapEx explorer in sec-7 and slide 7): Amazon ~$200B, Alphabet ~$180B, Microsoft ~$145B, Meta ~$125B, Oracle ~$50B. The infrastructure chapter got bigger, not smaller; the bottleneck is memory, power, and data-center execution.

Updates: $700B Sprint, Follow the Money Jump to section → Source: Yahoo Finance / Business Insider, Apr 29–30 2026
Defense

Pentagon signs 8 frontier AI vendors for classified networks — Anthropic excluded

DoD signed formal agreements with SpaceX, OpenAI, Google, NVIDIA, Reflection, Microsoft, AWS, and Oracle to deploy frontier AI on classified networks for lawful operational use. Anthropic was excluded amid an ongoing supply-chain-risk dispute. Frontier models are now national-security infrastructure — safety posture has become a go-to-market constraint.

Updates: Geopolitics Jump to section → Source: DefenseScoop, May 1 2026
Agents

Anthropic ships 10 ready-to-run finance agents

Ten templated workflows (pitchbooks, KYC, month-end close, etc.) deployed via Claude Cowork, Claude Code plugins, and Managed Agents — with Excel, PowerPoint, Word, and Outlook add-ins inbound and a Moody's MCP app covering 600M+ public and private companies. Agents are moving from generic chat into packaged vertical workflows.

Updates: Agent Revolution Jump to section → Source: Anthropic, May 5 2026

Recurring patch-notes log. Reported items are sourced from primary releases or first-party reporting; dollar figures and benchmarks reflect what was published in-window and may be revised in the next refresh. Each item is also reflected, in concise form, in the relevant section of the main report — this page is the running changelog with full sources.

← Back to State of AI report ▶ 20-min Slideshow