Record revenue numbers arrived in a month when no two of them shared a basis
Between Sep 24 and Oct 23, 2026, AI companies published more revenue numbers than in any earlier month. Five conclusions explain the month; pick one to see its evidence and the module that proves it.
Five conclusions
The top revenue figures are large, fast-growing and measured differently
Anthropic's run-rate reached $65B at the end of July (reported, counting cloud-partner sales gross), while OpenAI told investors about $50B annualized at the end of September on its own net basis. The $68–70B figure that circulated in late September was an investor gross-up, not an OpenAI number. Anthropic's quarterly revenue went from $4.7B to $11.6B to $16.8B (+147%, then +45%).
Usage is enormous; paying share is small
ChatGPT showed 1.2B weekly users on Sep 29, and OpenAI last gave 50M consumer subscribers, which is 5.6% of the 900M weekly users it reported on Feb 27. Microsoft 365 Copilot passed 30M paid seats against 450M+ commercial seats, at most 6.7% (mixed dates).
Prices moved in opposite directions
The Kingy AI small-model median price fell from $0.53 to $0.26 per million tokens in October, and Anthropic's Haiku 5.5 launched at $0.10 / $0.50 against Haiku 4.5's $1 / $5. At the frontier, Gemini 4 Argon's introductory $2 / $10 doubles to $4 / $20 on Dec 1.
Applications and infrastructure report revenue faster than they can fund growth
Podium announced more than $150M of AI-specific ARR and Chai reported $100M ARR in July (a $120M unaudited run-rate on Oct 1). CoreWeave's Q2 revenue was $2.58B (+112%) with a $104B backlog and about $29.5B of net debt.
Profitability is claimed, contested and set against huge commitments
Anthropic told investors it was profitable on an adjusted basis, without a published definition, while its draft S-1 shows a first-half net loss of $80.7B (about $72B non-cash) and compute commitments of at least $518B. The Oct 8 report on OpenAI's revenue moved suppliers.
Metrics at a glance
| Company | Metric | Value | Basis | Status (Oct 24) |
|---|---|---|---|---|
| Anthropic | Run-rate revenue, end of July | $65B | Annualized; partner sales gross | Reported Bloomberg |
| OpenAI | Annualized revenue, end of September | About $50B | OpenAI's own net method | Reported FT, Bloomberg; company declined to comment |
| Anthropic | Q3 2026 revenue | $16.8B | Recognized quarterly | Reported investor day, Oct 14 |
| Microsoft | AI business run rate | $37B | Company-defined run rate; date not stated | Company figure |
| ChatGPT | Weekly users | 1.2B | Weekly; recap says "collective" | Company-stated Sep 29 |
| Microsoft 365 Copilot | Paid seats | 30M+ | Seats | Company-stated Jun 2026 |
| Gemini 4 Argon | API price per 1M tokens | $2 / $10, then $4 / $20 | Introductory to Nov 30 | In effect Upcoming Dec 1 |
| Podium | AI ARR | More than $150M | Company-defined | Company-announced |
Sources: Bloomberg, Financial Times, CNBC, Alphastreet, Engadget, Microsoft, Google, PR Newswire.
Anthropic's $65B and OpenAI's $50B are the largest figures, and they are not like-for-like
Unless marked otherwise, every figure is a press report of investor materials or a company release, not an audited public-company number. Only annualized figures exist for all four companies, so the ranking orders that family by value.
Three views of the revenue numbers
SpaceX reported Q3 on Oct 20: revenue $14.6B (+161%), of which Super Intelligence was $7.56B, up 830% from $813M a year earlier, about 52% of revenue. It combines cloud services (customers include Anthropic and Google), Cursor, Grok and $538M of X advertising, so it must not be added to Anthropic.
Musk's target of $500B to $1T annual recurring revenue by end-2027 is an aspiration against LSEG analysts' roughly $110B of 2027 total revenue.
Up 123% year on year and about 11% of revenue per Bloomberg research. Its fiscal Q4 2026 call gave only a seat count (over 30 million paid Microsoft 365 Copilot seats), not revenue. Google and Meta gave no AI-specific revenue figure in the window, so neither is ranked.
Recognized revenue was $5.7B in Q1 and $6.7B in Q2 2026; Q2 × 4 = $26.8B, far below the end-September run rate, because a run rate captures the fastest-growing latest period. The same gap showed in 2025: $13.07B booked against a run rate above $20B at year-end.
Ranking table: one family of numbers, four bases
| Rank | Company | Metric and value | Basis | Date | Status |
|---|---|---|---|---|---|
| 1 | Anthropic | Run-rate revenue $65B | Annualized run-rate; gross of cloud-partner sales; investor update | End of July 2026 | Reported Bloomberg |
| 2 | OpenAI | Annualized revenue about $50B | OpenAI's own method, net of partner share; told to investors | End of Sept 2026 | Reported FT, Bloomberg; company declined to comment |
| 3 | Microsoft | AI business run rate $37B, +123% YoY | Company-defined run rate | Fiscal Q3 2026; exact date not stated | Company figure via Alphastreet, Nasdaq |
| 4 | SpaceX | Super Intelligence Q3 × 4 $30.2B (7.56 × 4, our arithmetic) | GAAP segment revenue, not a company run rate | Quarter to Sept 30 | Recognized reported by SpaceX |
The ranking is not like-for-like: the dates differ, and Anthropic counts cloud-partner sales gross while OpenAI counts only its own share.
On Oct 8 the Financial Times reported OpenAI told investors its annualized revenue was "approaching $50 billion". Bloomberg (Oct 9) said roughly $50B at end-September, shared with investors in talks to raise $30B or more at a $1.4T pre-money valuation (reported, unconfirmed). Q3 run-rate growth was 77% overall and 107% in enterprise, per CNBC.
Axios and others reported OpenAI near $70B in late September; Yahoo Finance dates a $68B version to Sept 29. Investors started from an estimated $40B in August and added the 70% growth OpenAI had cited: $40B × 1.7 = $68B (our reading). CNBC confirmed it counted partners' gross revenue. It was never an OpenAI number.
70 − 50 = $20B (68 − 50 = $18B), a ratio of 1.36–1.40×. FutureSearch's roughly 1.4× gross-up is consistent, but it is an aggregator estimate.
Bloomberg reported on Oct 9 that OpenAI expects $70B or more of annualized revenue by end-2026, driven mainly by enterprise (not a result). It continues the $50B series, so we infer net basis. It needs 70 ÷ 50 = +40%, or $20B, in about one quarter. It is not September's $70B gross-up.
The same cloud-partner sale is booked gross by Anthropic and net by OpenAI
Both companies are GAAP-compliant, so neither treatment is an error. The choice turns on who controls the customer relationship and delivers the product, per accounting professor Francine McKenna quoted by Axios.
Books the full $100 as revenue and the provider's cut as an expense.
Books only its own share of certain partner sales as revenue.
Size of the effect
Anthropic books $100 of a cloud-partner sale as revenue and the provider's cut as an expense. Business Insider says it is unclear whether the $16.8B for Q3 is gross or net, and that Anthropic has historically counted gross cloud-partner sales.
OpenAI books only its share of certain partner sales. Anthropic's channel effect (16% of that channel, 7.6% of revenue) is far below OpenAI's 29% gap, so the two adjustments are not interchangeable.
- End of 2025Run-rate about $9B
- Apr 6, 2026Over $30B
- Early May 2026Over $47B
- End of July 2026$65B (reported); Q3 × 4 = $67.2B is only about 3% above it, a quarter average against a point-in-time figure
- End of 2026 · expectations$110B per an attendee (reported); Jefferies: market expected $80B or more now and $105B by year-end (analyst estimate)
Run-rate path: a compilation citing Anthropic and CNBC. Sources: Axios, Business Insider, FutureSearch, Reuters via TradingKey.
Conflicts the page resolves
Q1 is $4.7B at the investor day but $4.8B per the WSJ; Q2 is $11.6B against $11.5B per Bloomberg. The page follows the investor-day figures, which are the latest and sum to the draft S-1 half-year revenue of $16.3B.
Q2 grew 11.6 ÷ 4.7 = +147%; Q3 grew 16.8 ÷ 11.6 = +45%. One article's "about 14%" Q1-to-Q2 growth is a slip; others round Q3 growth to 50%.
Told to investors at a San Francisco investor day on Oct 14, 2026, reported by attendees. Anthropic declined to comment and it is not a filing. Whether it is gross or net is not stated.
Application ARR is arriving at record speed, but every announcement defines ARR differently
One layer below the model vendors sit the applications that buy the models and the infrastructure that sells the compute. Infrastructure companies report recognized revenue and backlog, not ARR.
Two views of the numbers
More than $150M of AI-specific ARR about 2.5 years after expanding into AI agents; its original software business needed 8 years. It reports 20,000+ deployed AI agents and says customers generated more than $13B in revenue with its tools (company claim). No announcement date and no ARR definition beyond "AI-specific".
(120 − 80) ÷ 80 = +50% from Q1 to October, about six months. $100M ÷ about 15 staff ≈ $6.7M of ARR per employee in July. Chai says it is profitable excluding sales and marketing.
Cursor added $608M of recognized revenue from Aug 14 to Sep 30. The $60B all-stock deal was about 60 ÷ 4 = 15× the roughly $4B run-rate of May (secondary sources). Annualizing 47 days gives about $4.7B (our arithmetic, mixing recognized revenue with an estimated run rate).
ARR milestones and what each number measures
| Company | Layer | Figure | Basis, date and status | Label |
|---|---|---|---|---|
| Podium | Application: AI agents for local businesses | More than $150M AI-specific ARR, about 2.5 years after expanding into AI agents (8 years for the original business) | "AI-specific ARR", no further definition; date not stated | Company-announced |
| Chai | Application: consumer AI chat | $120M; $100M in July 2026; $80M at end of Q1 2026 | Unaudited run-rate as of Oct 1, 2026, relayed by Dealroom | Company-stated |
| Cursor (Anysphere) | Application: coding | $608M revenue, Aug 14 to Sep 30 | Recognized revenue in SpaceX's Q3 results; not ARR | Recognized |
| Axya | Application: manufacturing procurement | "Multi-million-dollar" ARR; doubled year on year | CEO claim, no number; Sep 24, 2026 | Claim, no figure |
| CoreWeave | Infrastructure: neocloud | $2.58B Q2 revenue (+112%); $104B backlog | Recognized revenue and backlog, Q2 as of Jun 30 | Reported company releases |
| Nebius | Infrastructure: neocloud | $582M Q2 revenue (+454%) | Recognized revenue, Q2 as of Jun 30 | Reported company releases |
Sources: Podium release, Dealroom, BetaKit, New Market Pitch, Technology.org, I/O Fund.
Infrastructure: demand strong, financing the constraint
CoreWeave net debt of about $29.5B is 29.5 ÷ (2.58 × 4) ≈ 2.9× annualized revenue; Nebius about $0.5B is 0.2×. CoreWeave's backlog is $104B plus more than $25B of new commitments in early Q3. An undated $55.6B figure conflicts, so the page follows $104B.
Nebius raised H100 on-demand prices 17%, from $3.85 to $4.50 per GPU-hour, from Oct 1 (H200 +20%, B200 +19%, B300 +21%). On Oct 22 Jensen Huang told CNBC neocloud demand is "off the charts" but limited working capital holds expansion back (reported; about the industry).
CoreWeave's Q2 net interest expense was $640M against a $626M net loss and $1.51B adjusted EBITDA; its stock fell 9.67% on Oct 22 and 2032 bond yields touched 15% on Oct 18. Nebius funds more through customer prepayments (50% to 60% of capex on its largest contracts).
Context: Nvidia's fiscal Q2 revenue was $96.2B (+106%), Data Center $89B, with a Q3 guide of about $108B ± 2% (about +12% quarter on quarter); dates are not stated.
Usage is enormous, but only 4% to 7% of ChatGPT users and Copilot seats pay
The people and seats behind the revenue lines, and how few of them pay. User counts here sit on four bases that cannot be added: weekly users, monthly users, paid seats and internal employees.
Users, payers and adoption
The 50M is a "more than" count and seven months older than the 1.2B, so 4.2% is a floor if subscribers have not fallen. We follow OpenAI's split as relayed by Whizi: 50M consumer subscribers, with 9M paying business users counted separately (4.5× the 2M of Feb 2025).
A seat is a purchased licence, not a user, and seats rise partly through bundles such as E7. The Register's 3.3% (Copilot Chat payers) has a different denominator and matches 15 ÷ 450 by coincidence. GitHub Copilot paid counts stopped after January.
Google names no paying Gemini subscribers; Anthropic discloses no paying count. Aggregator estimates for US consumers: 4.5% pay for ChatGPT, Gemini or Claude (a16z analysis, Oct 5), and 2.2% of households paid for any AI subscription as of April 2026.
| Product | Value | Basis | Date | Disclosed by |
|---|---|---|---|---|
| ChatGPT | 1.2B | Weekly users | Sep 29, 2026 | Company-stated OpenAI DevDay slide |
| Gemini app | >1B | Monthly active users | Aug 11, 2026 | Company-stated Google |
| Microsoft 365 Copilot | >30M | Paid seats | Apr–Jun quarter, said Jul 29, 2026 | Company-stated Microsoft |
| GitHub Copilot | 50M | Users | Same call | Company-stated Microsoft |
| Claude | 245M | Monthly users, estimate | May 2026 | Estimate Sensor Tower panel, aggregator-only |
| Meta Muse Code | 6,000+ | Internal employees | Reported Oct 5, 2026 | Reported The Information, not Meta |
Gemini's 1B covers the app only; AI Overviews (2.5B) and AI Mode (1B+) are Search features counted separately. Meta's figures are employees, not market adoption.
Prices moved in opposite directions: small models halved, the frontier is set to double
Per-token prices fell at the small-model end and were set to rise at the frontier, while list prices of existing models barely moved. USD per 1M tokens; "in effect" means the date is on or before Oct 24, 2026.
Prices and meters
| Model | Input | Output | Date | Status | Source |
|---|---|---|---|---|---|
| Gemini 4 Argon, introductory | $2 | $10 | GA Oct 13; to Nov 30 | In effect time-limited | |
| Gemini 4 Argon, standard | $4 | $20 | From Dec 1 | Upcoming | |
| Gemini 3.8 Flash, introductory | $0.75 | $3.75 | To end of 2026 | In effect standard price not stated | |
| Claude Haiku 5.5prompts up to / over 100k | $0.10 / $0.50 | $0.50 / $2.50 | Oct 7 | In effect | Anthropic |
| Claude Haiku 4.5 | $1.00 | $5.00 | Not stated | Superseded | Anthropic |
| Claude Sonnet 5.5 | $2 | $10 | Sep 28 | In effect | Anthropic |
| Claude Opus 5.5 | $4 | $20 | Sep 22 | Reported Anthropic page not in material | Substack |
| GPT-6.1 Sol | $2 | $10 | Sep 29 | Reported OpenAI page not in material | Improvado |
| GPT-6 Astra | $10 | $50 | Not stated | Reported OpenAI page not in material | Substack |
| Meta Muse Wonder | $4 | $20 | Oct 12 | Reported Meta page not in material | Shattered.io |
Google calls Argon's $2 / $10 an "introductory price", with cached input 95% cheaper. It was announced on Sep 30 to trusted cyber defenders only and opened publicly on Oct 13.
The median dropped from $0.53 to $0.26 (down 50.0%). Kingy says none of the 17 models and plans present in both editions changed list price; every move came from a lab swapping in a new model. Haiku 4.5 at $2.00 blended gave way to Haiku 5.5 at $0.20: (3 × $0.10 + $0.50) ÷ 4.
Kingy's frontier median fell from $4.50 to $4.00 because Argon at its introductory rate replaced Gemini 3.1 Pro Preview. At the standard rate Argon blends to $8.00; if the other six entries stay put, the median would be $6.00 (our estimate). GPT-6 Astra at $10 / $50 is reported at 2.5 times GPT-5.6 Sol.
Sonnet 5.5 kept $2 / $10 and halved cache reads from $0.20 to $0.10, with up to 30% lower cost per task claimed by Anthropic. Artificial Analysis (via a Substack review) records about 193,000 output tokens per task at max effort, so its cost per task exceeds Opus 5.5's.
Free users are limited to Gemini 3.5 Flash-Lite, down from three models; Plus subscribers lost Pro, and Deep Think moved down to AI Pro. No Google post explains it. Ultra is $99.99 per Tech Insider but $199.99 per Kingy; the material does not settle it.
A new Pro 500 tier, and Pro 200 reopened at the same $200 with a lower allowance; holders active Sep 22–29 keep the old allowance until Oct 29 (reported, upcoming). Go at $8 "may include ads". Fast mode draws the allowance at 2.5× and Ultrafast at 8×. A US student offer of four free Plus months runs through Oct 31.
Since Oct 7, Max and Team plans include monthly API credits; the amount is not stated. Kingy's subscription medians were unchanged at $20 (standard) and $200 (top): the changes were in access and allowances, not sticker prices.
| Product | Unit | List price |
|---|---|---|
| HubSpot Customer Agent | Resolved conversation | $0.50 (cut from $1.00, Apr 14, 2026) |
| Fin | Outcome | $0.99 |
| Zendesk AI agents | Verified resolution | $1.50 committed, $2.00 pay-as-you-go |
| Salesforce Agentforce Help Agent | Resolution | $2 |
| Salesforce Agentforce Flex Credits | Action | About $0.10 |
| Salesforce Agentforce add-on | User per month | $125 |
At three actions per conversation, Flex Credits cost 3 × $0.10 = $0.30, or $300 per 1,000 conversations (Macha, vendor blog); the $2 conversation meter is cheaper only above 20 actions. List prices were read on Oct 18, 2026 by QueryNow, a vendor-run comparison, so treat them as secondary.
Profit is claimed but contested, and commitments dwarf revenue
What margin is left after serving costs and compute bills, who carries them, and how the market reacted on Oct 8. The only company statement here is Anthropic's June 1 announcement of a confidential draft Form S-1; the rest is press-reported or an analyst estimate.
Five lenses on profit and cost
At an Oct 14 investor day Anthropic reportedly said Q3 revenue was $16.8B (reported by attendees; no filing; gross or net not disclosed) against $11.6B in Q2, and that it was profitable "on an adjusted basis". No profit amount or methodology was disclosed; the FT says the adjustment excludes stock-based compensation. Outside figures disagree (an 11% adjusted operating margin from New Market Pitch, 5.4% on $17.3B from First Page Sage), and neither is treated as fact.
| Item | Period | Value | Source |
|---|---|---|---|
| Revenue | FY2025 | About $4.6B | Reuters via TradingKey |
| Operating expenses / compute and infrastructure | FY2025 | $12.65B / $7.33B | Reuters via TradingKey |
| Operating loss | FY2025 | Over $8B, excluding financing write-downs | Reuters via InvestmentNews |
| Net loss | FY2025 | About $42B, of which about $34B non-cash | Reuters via TradingKey |
| Revenue / compute and infrastructure | H1 2026 | $16.3B / nearly $9B | WSJ via Dealroom |
| Net loss | H1 2026 | $80.7B, of which about $72B non-cash | WSJ via Dealroom |
The draft S-1 is unpublished; figures come from Reuters (Sept 29) and the WSJ. "Adjusted" remains undefined.
At least $518B over ten years to six partners, about 80% non-cancellable. Named items: Google $111B, AWS $110B, Microsoft $31B and $161.2B of Broadcom-related leases, which sum to 413.2, or 79.8% of 518. xAI (up to $84.5B, mostly cancellable on 90 days' notice) and AMD (over $20B) are the flexible items. The WSJ review gives "up to $750B"; the page uses the lower, firmer $518B.
| Ratio | Arithmetic | Result |
|---|---|---|
| $518B ÷ FY2025 revenue | 518 ÷ 4.6 | 113× |
| $518B ÷ annualized Q3 revenue | 518 ÷ (16.8 × 4 = 67.2) | 7.7× |
| Even annual spend ÷ annualized Q3 | (518 ÷ 10 = 51.8) ÷ 67.2 | 77% |
| Non-cancellable only | (413.2 ÷ 10 = 41.3) ÷ 67.2 | 61% |
The even ten-year spread is our assumption; no payment schedule is reported. Revenue is booked gross, so net ratios would be higher.
A July presentation reported by the FT on Sept 18 projects $856B of compute spend through 2030, $840B of cumulative revenue (856 ÷ 840 = 1.02×) and $278B of negative free cash flow for 2026–2030 (internal projections). Against about $50B annualized, 856 ÷ 50 = 17×.
2026 guidance: Alphabet $195–205B, Meta $130–145B, Microsoft about $175B, Amazon $200–220B, Oracle $50–70B. The sum of ranges is $750–830B against about $413B in 2025, nearly double (1.9×). Backlogs: Microsoft $678B, Oracle $664B.
If spend falls short, Anthropic "must pay Google the difference". Two customers supplied about 12% each (24%) of Anthropic's 2025 revenue. Amazon, Google and Microsoft are each its investor, customer, supplier and competitor.
80% software and 76% total gross margin, 62% for usage-only pricing (FY2025). Aleph finds no median compression, with software margin at 79–81% over four years. CloudZero's illustration: AI cost at 8% of revenue turns an 80% margin into 72%.
OpenAI's gross margin was about 39% in Q1 2026 against about 33% a year earlier (reports). Operating loss was $9.3B on $5.7B in Q1 and $12.3B on $6.7B in Q2, which is 163% and 184% of revenue. Anthropic: 80% before partner revenue share (FT) against a modelled 52% for Q2 with partner shares as costs (First Page Sage, estimate); not comparable.
The FT reported OpenAI told investors annualized revenue was roughly $50B; CNBC confirmed it. $68–70B had circulated in late September, and the $18–20B gap was attributed mainly to accounting, not lost sales. The stated cause of the selloff was this measurement difference, which made investors ask whether a major AI customer can pay for the compute it has ordered.
CoreWeave: $104B backlog, largely debt-funded. Oracle: $664B remaining performance obligations. Broadcom: first custom OpenAI accelerator. AMD: also Meta and Anthropic deals. Nvidia: about $169B of value lost, about 12 GW of OpenAI commitments. CoreWeave's loan clause on a "Cash Trap Event" is a risk, not a breach that has occurred.
| Step (Bain, Sep 29; every value an estimate) | Value | Arithmetic |
|---|---|---|
| Hyperscaler capex, 2026 | Up to $780B | About 5× three years earlier |
| AI infrastructure spend, 2031 | $1.5T a year | Bain scenario |
| Revenue needed | About $6T a year | $1.5T ÷ 25% capex share |
| Consumer plus enterprise AI | $1.2T–$1.8T | $200B–$400B + $1.0T–$1.4T |
| Gap to fund | About $4.2T | $6.0T − $1.8T (at the low end, $4.8T) |
Search and ads ($100B–$200B), autonomous systems ($400B) and physical AI ($900B) total up to $1.5T, leaving $2.7T ($4.2T − $1.5T). A year earlier Bain put the need at $2T by 2030, with an $800B shortfall, so the requirement has tripled.
Commentary notes undated estimates: Goldman's roughly $300B break-even AI revenue for five hyperscalers against about $70B of AI cloud revenue above trend, a gap of about $230B a year, and Sequoia's David Cahn at about $600B. Semafor (Sep 15) put four hyperscalers' spending about $66B above operating cash flow over six quarters.
The State of AI Report (Oct 8) says OpenAI and Anthropic's combined reported run rates reached about $105B by late summer, from about $30B at the start of the year, roughly 3.5×. On Ramp's sample, 1% of customers drove about 80% of observed OpenAI and Anthropic spend.
Compare AI revenue only on one basis, and read every profit claim with its definition
This module reads the earlier modules together and adds a dated watch-list. It uses only facts shown above.
Anthropic $65B (end of July, gross) against OpenAI about $50B (end of September, net). The gap that moved markets on Oct 8 was 70 − 50 = $20B, 29% of the gross-up figure. Anthropic's own gross-versus-net effect looks smaller: 16% of its cloud channel, 7.6% of $4.6B revenue.
Anthropic's quarterly growth fell from +147% to +45%, and Q3 × 4 = $67.2B is only about 3% above the July run-rate. OpenAI needs about +40% in one quarter to reach the reported $70B-plus year-end expectation (70 ÷ 50).
5.6% of ChatGPT weekly users (50 ÷ 900, from Feb) and at most 6.7% of Copilot's commercial seats. Ramp's index shows 55.9% of its US business customers paid for AI in September, against 19.8% of US businesses in the Census survey, so the sample matters.
Small models got cheaper at once (median $0.53 to $0.26), while frontier medians moved less ($4.50 to $4.00) and Argon's price doubles on Dec 1.
Compute of $7.33B on revenue of $4.6B is 159% of revenue in FY2025. Commitments of at least $518B are 7.7 times annualized Q3 revenue. AI product gross margins are estimated at 45% in 2025 against 80% for software.
Lower per-token prices can still raise spend if usage grows faster; the material does not show that directly. Enterprise growth is shifting from adding payers to deepening use, which raises the value of usage-linked metrics such as tokens and agent use over seat counts.
Layer snapshot: the same month on three layers
| Layer | Company | Metric | Value | Basis |
|---|---|---|---|---|
| Model vendor | Anthropic | Run-rate, end of July | $65B | Gross of partner sales |
| Model vendor | OpenAI | Annualized, end of September | About $50B | Net |
| Application | Podium | AI ARR | More than $150M | Company-defined |
| Application | Chai | ARR (July) | $100M | Company-stated |
| Infrastructure | CoreWeave | Q2 revenue | $2.58B (+112%) | Reported |
| Infrastructure | Nebius | Q2 revenue | $582M (+454%) | Reported |
What it means by reader
Small-model prices halved; Argon's price doubles on Dec 1.
Route simple work to small models and budget for the Dec 1 step.
Small-model median $0.53 to $0.26 (Kingy); Haiku 5.5 at $0.10 / $0.50 (Anthropic); Argon $4 / $20 from Dec 1 (Google).
Seat and usage prices differ widely; paying share is low.
Ask for the basis of any quoted revenue or ARR, and pay for measured use.
Copilot seats at most 6.7% of commercial seats (Microsoft); meters from about $0.10 per action to $125 per user per month (list prices, vendor-run comparison).
Bases differ between labs; suppliers fell on one report.
Compare revenue on one basis before comparing multiples.
Anthropic $65B gross against OpenAI about $50B net; Oct 8 closes: CoreWeave −7.77%, Oracle −5.48%, Nvidia −2.94%.
Key dates
| Date | Event | State on Oct 24, 2026 | Source |
|---|---|---|---|
| Oct 7, 2026 | Claude Haiku 5.5 launches at $0.10 / $0.50 | In effect | Anthropic |
| Oct 13, 2026 | Gemini 4 Argon GA at $2 / $10 | In effect | |
| Oct 14, 2026 | Anthropic investor day: Q3 revenue $16.8B | In effect | CNBC via Jawlah |
| Oct 29, 2026 | ChatGPT Pro 200 old allowance ends (reported) | Upcoming reported | Skillmammoth |
| Nov 30, 2026 | Argon introductory API price ends | Upcoming | |
| Dec 1, 2026 | Argon moves to $4 / $20 | Upcoming | |
| End of 2026 | OpenAI expected at $70B-plus annualized; Anthropic expects a $110B run rate (both reported expectations) | Upcoming reported | Bloomberg; Mint |
| End of 2026 | Gemini 3.8 Flash introductory $0.75 / $3.75 ends | Upcoming |