Industry

Vol. 1·Wednesday, November 18, 2026

Anthropic vs. OpenAI Enterprise Market Share: What the S-1 Settles, and What It Doesn't

The filing shows how Anthropic counts a dollar. The newest outside readings put it ahead of OpenAI by 3 to 24 points, depending on the panel.


26 min read

Tips, corrections, or questions? support@omniscient.media

Anthropic vs. OpenAI Enterprise Market Share: What the S-1 Settles, and What It Doesn't

[Editor's Note: This piece was researched, drafted and checked with AI models made by Anthropic, one of its two subjects. Omniscient Media is also a paying customer of Anthropic's API. One model that reviewed the draft was made by OpenAI, the other subject. Neither company saw or approved the piece before publication.]

The numbers used to rank Anthropic against OpenAI each count something different, and Anthropic's S-1 pins down only one lab's yardstick. About 44% of Anthropic's revenue in the first nine months of 2026 came through cloud providers' marketplaces, down from about 48% a year earlier, and the company books every one of those sales at full value, counting the platforms' cut as an expense.[1] OpenAI "recognizes only its share," per Bloomberg; Anthropic is described as counting the gross amount of cloud-partner sales.[2] Netting the channel fees out would trim Anthropic's nine-month revenue by roughly 14% to 17%, by our arithmetic from the filing's own fee figures.[1] That one accounting choice sits under most revenue comparisons between the two labs, and Anthropic's S-1 supplies the fee figures that let a reader size it.

The filing and data: Anthropic's amended S-1 (Amendment No. 2) on EDGAR; the original S-1 (filed October 26th); Ramp AI Index; OpenRouter rankings; Menlo Ventures, 2026: The State of Enterprise AI. Every filing figure below comes from a Fidelity-hosted copy of the amended prospectus, the version we could read in full, unless we say otherwise.

The S-1 doesn't say who leads, though. Anthropic filed it publicly on October 26th,[3] amended it on November 2nd to add third-quarter results,[4] and launched its roadshow on November 10th for 825 million shares at an expected $70 to $78 each; as of that release, the registration statement hadn't yet become effective.[5] Meanwhile, the outside datasets split in September. In the week of September 7th, OpenAI's models took "just over 50% of combined spend between the two labs" on OpenRouter, a service that routes developers' requests to whichever model they pick, and OpenRouter called it "the first week at or above the 50% line since the week of Feb 24, 2024."[6] In the same month, on a panel of companies that track their AI token bills through Ramp, the corporate-card company, Anthropic held about 56% of the two labs' spending and dipped no lower than 50.4%, on September 21st, in the series as we retrieved it on October 12th.[7]

The newer readings agree on the leader and differ on the margin. Ramp's latest weekly panel, for the week ending November 1st, gives Anthropic 49.3% of all token spending and OpenAI 46.1%, which works out to about 51.7% against 48.3% between the two by our arithmetic; four weeks earlier the split was 53.2% to 42.3%.[8] Menlo Ventures' survey, published November 3rd, puts Anthropic at 55% of enterprise LLM API spending and OpenAI at 31%.[9] Ramp's gap narrowed over those four weeks, while Menlo's report has Anthropic's lead widening from a year earlier.[9]

The OpenRouter number is the one that traveled. At the start of the year Anthropic had held about three quarters of the two labs' spending there, and the reversal on that measure became a Wall Street Journal story about OpenAI gaining ground on Anthropic in a price war. One account from Trending Topics noted that "OpenRouter's customers are mostly A.I. start-ups and tech companies."[10][11] Other counts on the same service look different again. Measured in tokens rather than dollars, OpenAI took 78% of the two labs' combined usage in the week of September 14th.[6] Counted by requests in the week beginning October 5th, OpenAI had 18.6% and Anthropic 3.9%, with DeepSeek leading at 24.5%.[12] Menlo, whose own report leans on OpenRouter's traffic data, says the service's users "skew more toward individual developers and startups than the enterprise buyers in Menlo's survey."[9]

Ramp's numbers don't all agree with each other, either. According to a Korean outlet's account of the price-war coverage, a Ramp economist said an analysis of about 70,000 Ramp customers showed spending on OpenAI models passing Anthropic's in mid-September for the first time since the previous December, before Anthropic regained a narrow lead.[13] That broader dataset isn't the token panel Ramp embeds on its site, and differences in which customers and which spending each includes could explain why they diverge.

"Enterprise AI" here means business buyers of AI models and products: API spending, business seats and business subscriptions. None of the measures we examined covers that market directly, and they often produce different rankings, depending on whose buyers they sample and what they count. Anthropic's prospectus shows how soft the category is even from the inside: the company says it doesn't separately define or internally track a category of "enterprise" customers, and instead uses customers with $100,000 or more of trailing-twelve-month revenue as its measure of enterprise adoption.[1] OpenAI hasn't published its financials, and Anthropic's nine-month 2026 figures come from unaudited interim statements.[1]

The scoreboards

Each business dataset in circulation measures a different slice of buyers, and several come from companies with a stake in one lab or in AI spending generally.

Dataset

Who it measures

What it counts

Stake

Latest reading

OpenRouter (2026)

Developers and start-ups routing API calls

Weekly spend, tokens and requests, two-lab or all-model

Listed by Menlo Ventures as a portfolio company

OpenAI just over 50% of two-lab spend, week of Sep 7

Ramp token panel

Ramp customers using its AI spend management product

Provider-billed token spend and volume, including cloud-platform and router API spend; subject to revision

Sells AI cost tooling

Anthropic 49.3%, OpenAI 46.1% of all token spend, week ending Nov 1

Ramp AI Index (monthly)

US businesses on Ramp's card and bill-pay platform

Share of sampled firms spending on a vendor; firms can pay both

Same

Anthropic 44.8%, OpenAI 40.6% of firms (October data)

Menlo Ventures (survey, Sep 2026)

Nearly 400 US enterprise AI decision-makers, fielded by Morning Consult

Estimated API spend, derived from survey responses

Investor in Anthropic and OpenRouter

Anthropic 55%, OpenAI 31%, Google 9% (published Nov 3)

Andreessen Horowitz (survey, early 2026)

Enterprise CIOs

Share reporting production use; firms can use both

OpenAI investor

78% use OpenAI in production, 44% Anthropic (January 2026)

Anthropic S-1 (filed Oct 26; amended Nov 2 and Nov 10)

Anthropic's own customers

Revenue and customer counts by trailing-twelve-month revenue

The issuer; no comparable filing from OpenAI

About 44% of nine-month 2026 revenue via cloud marketplaces (about 48% a year earlier); more than 11,900 customers at $100,000 or more, more than 1,900 at $1 million or more (Sep 30)

OpenRouter's own rankings page warns that its public charts rank models by tokens or requests, and "None of them measures users or spend."[12] The spending series reached the press through posts and a newsletter, whose share charts are "computed within the segment named in each chart's footer," which for the headline figure means the two labs alone.[6] On that measure OpenAI's share of tokens ran well ahead of its share of dollars, which fits a mix of cheaper OpenAI models carrying more of the volume. Another newsletter noted that nobody had separated how much of OpenAI's jump was new usage and how much was "routing changes toward the newer model."[14]

Ramp publishes two different measures. Its headline index defines AI adoption as "the share of businesses with spending on AI models, subscriptions, tokens, and APIs," so a firm with one $20 seat counts the same as one with an eight-figure contract.[8] By that measure 44.8% of US businesses on Ramp paid Anthropic in October and 40.6% paid OpenAI; OpenAI's share was unchanged from September, while xAI rose 2.2 points to 6.3%.[8] Among firms paying for AI, the top 1% spent a median of $10,199.20 per employee per month, against $16.32 for the median firm, which is the gap a share-of-firms count can't see.[8]

The token panel is narrower, and Ramp is explicit about it. The product imports usage and billing data from connected providers, and Ramp counts "API spend paid directly to model providers, in addition to API spend through GCP, AWS, Azure, and several popular routers and inference products." It adds that the panel "is not representative of Ramp's total AI spend volume, though results are directionally similar," and that "historical estimates are subject to revision" whenever a business connects a new provider.[8] Over a longer span the panel tells a different story from this year's. In the series as we retrieved it on October 12th, OpenAI held about two-thirds of two-lab spending in January 2025, Anthropic's monthly share first passed half in October 2025, and it averaged about 68% in March 2026 and about 56% in September, by our count.[7] The page now shows only recent weeks, so we couldn't re-check those monthly figures. In the latest weeks, total token spending rose 10.5% and volume 29.0% against the same week a month earlier, while the blended price per million tokens across the two labs fell 14.8% to $0.53. Anthropic led on volume (55.1% against OpenAI's 41.4%) even as its spend share slipped, which fits the cheaper-tokens pattern seen on OpenRouter.[8]

Ramp also revises. In May a Ramp economist announced Anthropic's first lead in business adoption, at 34.4% to 32.3%,[15] and those numbers no longer appear in the series Ramp published by October, where OpenAI was still narrowly ahead in April.[7]

Two surveys point in different directions, use metrics that can't be compared, and were each published by an investor in one of the two labs. Menlo's 2026 report says Anthropic's share of enterprise LLM API spending rose from 40% to 55% and OpenAI's from 27% to 31%, with Google falling from 21% to 9%; together the two labs hold 86%.[9] Menlo's own description of the method is that "LLM market shares represent estimated dollars spent based on proportion of production API usage," built from what respondents reported about their workloads, so each share is an estimate modeled from survey answers.[9] The report marks both Anthropic and OpenRouter as Menlo portfolio companies, which means two of the scoreboards above lean on the same investor's work. Andreessen Horowitz opened its January survey with "we are investors in OpenAI," then reported that "78% of surveyed enterprise CIOs are using OpenAI models in production," against 44% using Anthropic.[16] A count of who uses a lab in production says little about how much they spend.

Consumer measures sit outside that definition of enterprise, but they're the numbers most readers have seen: ChatGPT had 56.4% of generative-AI web traffic in September on Similarweb's count, against 9.6% for Claude.[17] Ramp says, by its own measure, that the share of firms switching AI model providers "reached a record 8% in September."[18]

Two ways to count a dollar

Anthropic's filing confirms its side of Bloomberg Law's description.[2] The company's reason for gross booking is that "the Company is the principal and recognizes revenue on a gross basis because the Company has a direct contractual relationship with customers, is primarily responsible for providing the intelligence service to customers, and establishes pricing," with the cloud platforms acting as collection agents.[1] The prospectus summary describes the structure in plain terms: Anthropic establishes "direct contractual relationships with our customers while paying a fee to each CSP."[1]

The S-1 is the most detailed disclosure of large-customer counts and channel mix that either lab has made, as far as we found. Cloud marketplaces carried about 11% of Anthropic's revenue in 2023, about 32% in 2024 and about 47% in 2025; for the nine months ended September 30th, the share was about 48% in 2025 and about 44% in 2026.[1] Reliance on the channel has eased slightly in revenue terms even as the dollars grew; marketplaces still accounted for 52% of net accounts receivable on September 30th, down from 60% at the end of 2025, because the platforms bill end customers, collect, and remit to Anthropic net of their fees.[1]

The cost of those channels is visible in the expenses. Sales, marketing and partnership costs rose to $5.53 billion for the nine months of 2026 from $322 million a year earlier, and the filing attributes $4.8 billion of the $5.2 billion increase to fees paid to channel partners.[1] If that $4.8 billion were netted against revenue, nine-month revenue of $33.24 billion would come out roughly 14% lower by our arithmetic, and netting the whole $5.53 billion line would cut it about 17%. Against marketplace revenue of about $14.6 billion (44% of the total), $4.8 billion is about a third, a rate that looks high for a marketplace fee. So the line probably holds payments to other channel partners too, and we found no passage that says how it splits.

OpenAI "told investors that it hit roughly $50 billion in annualized revenue at the end of September," CNBC confirmed, after a figure of $68 billion had circulated that a person familiar with the matter said "included gross revenue from OpenAI's partners, which helps investors make a more direct comparison with Anthropic."[19] The Financial Times, as quoted by TechCrunch, said the higher number came from "attempts by OpenAI's own investors to produce a direct comparison with Anthropic's annualised revenues."[20] We covered the gap between OpenAI's projections and its actuals in our piece on OpenAI's revenue miss.

Both labs have described fast growth in their own terms. Anthropic said in May that its "run-rate revenue crossed $47 billion," and CNBC reported that it claimed a run rate of $65 billion at the end of July.[21][22] The amended prospectus puts a booked number beside that: $17.0 billion of revenue in the third quarter and $33.24 billion for the nine months, against $2.5 billion in the same months of 2025.[1] CNBC reported that quarter-on-quarter growth slowed to about 47% from 144% in the second quarter.[4] Four times $17 billion is roughly $68 billion, by our arithmetic, a gross-basis quarter that shouldn't be set beside OpenAI's roughly $50 billion net-basis run rate without adjusting one of them. It does equal the $68 billion figure that circulated for OpenAI on a gross basis, but that match proves little: one number was assembled by OpenAI's own investors, per the Financial Times,[20] and the other is a single quarter multiplied by four. The filing also defines Anthropic's own run-rate metric: the trailing 28 days of consumption fees multiplied by 13, plus monthly subscription fees multiplied by 12.[1] We found no published equivalent for OpenAI's annualized figure, so only one side of that comparison has a visible method. We covered the lead's origins in our piece on Anthropic passing OpenAI on revenue. OpenAI, for its part, told investors of "107% run rate growth for its enterprise business" in the third quarter.[19] Those are snapshots from different months on different accounting bases, and converting either lab's number to the other's method would take assumptions that can't be checked from public documents, from fee rates to the monthly mix by channel, so a like-for-like ranking isn't possible yet.

Margins are reported even more loosely. The Information reported in January that Anthropic expected a 2025 gross margin of about 40%, against about 46% for OpenAI; with partner revenue counted differently, even those may not be comparable.[23] The filing adds a wrinkle. Its summary income statement has no cost-of-revenue line at all, and channel-partner fees sit inside sales, marketing and partnerships, so a gross margin built from the filing would leave the platforms' fees out.[1] What it does show is a compute and infrastructure line of $18.8 billion against $33.24 billion of revenue for the nine months, about 57% by our arithmetic, which includes $6.5 billion of compute bought on short-term commitments that the company says costs significantly more than compute under its multi-year deals.[1] SemiAnalysis, which judged Anthropic's subscriptions "an overwhelmingly better deal" for users, found that they "can take up over 40% of inference compute" while being about 10% of Anthropic's revenue; its public text gave no comparable figure for OpenAI, though it said subscriptions make up a larger share of OpenAI's revenue.[24]

The revenue itself is mostly metered. Consumption-based revenue was $26.3 billion, or 79% of nine-month revenue, against $6.9 billion from subscriptions, and Anthropic says a growing number of enterprise customers sign non-cancelable annual or multi-year contracts that mainly produce consumption revenue.[1] On large accounts, the prospectus counts more than 11,900 customers with $100,000 or more of trailing-twelve-month revenue as of September 30th (up from more than 950 a year earlier), more than 1,900 with $1 million or more, and roughly 250 with $10 million or more; the $100,000-plus group produced $24.5 billion, about 74% of nine-month revenue.[1] Concentration is moderate by the filing's account: no customer exceeded 10% of revenue in the nine months, and the top five were about 25%.[1] It also says over 50% of the Fortune 500 are $100,000-plus customers, yet spending averages under 0.1% of those companies' total expenses, which the company reads as runway.[1]

Those are the only large-customer counts either lab has put on paper that we found, and they carry limits: they cover one lab, each "customer" is an organization Anthropic assembles by grouping accounts from its own records, and the 2026 financials behind them are unaudited. Anthropic says it holds the direct contract even when a customer buys through a cloud, and that many customers use both channels, which suggests marketplace buyers are counted, though we found no passage that says outright how.[1] The filing also reports net revenue retention of approximately 900% as of September 30th, a company-defined metric that compares the latest quarter's revenue from customers with any revenue a year earlier against what they paid then, and excludes Pro, Max and Ultra subscribers.[1]

Partner channels, including the federal one

Both labs sell through partners that also compete with them, on different terms. OpenAI's April agreement with Microsoft governs how its models reach Azure customers: CNBC reported that "OpenAI products will ship first on Azure unless Microsoft decides otherwise," that OpenAI keeps paying Microsoft a share of revenue, and that "Microsoft will no longer pay a revenue share to OpenAI" on Azure sales of its models.[25]

Microsoft shows how mixed a single partner can be. In Microsoft 365 Copilot, "Microsoft enables Claude models on by default for most customers in commercial cloud (excluding EU/EFTA and UK)."[26] In GitHub Copilot, which Microsoft also owns, "GitHub designated GPT-5.3-Codex as the base model" for business and enterprise customers in March.[27] GitHub's policy of switching new models on by default excludes "models not covered by GitHub's data retention agreement (e.g., Fable 5)."[28] The Information reported that Microsoft is on track to pay Anthropic more than $2 billion a year.[29]

The prospectus also ties the platforms to Anthropic's cap table. A concurrent private placement of non-voting Class N stock, contingent on the offering and scheduled to close right after it, has Google buying about $10.0 billion and Microsoft about $1.0 billion at an assumed $74 midpoint price, alongside Nvidia ($10.0 billion), AMD, Broadcom, Micron and TSMC, for $26.0 billion in all; the filing says Google's purchase funds part of an investment it had already committed to.[1] Convertible notes held by Amazon Web Services also convert into Class N shares at the offering, up to an ownership cap.[1] Anthropic's own risk factors name the exposure: a substantial portion of its revenue and receivables is "driven by a limited number of customers and distribution channels," and its reliance on a few cloud providers and suppliers, "some of which have strategic and investment relationships with us and are our customers," creates "complex dynamics that could give rise to conflicts of interest."[1] So the channels carrying more than two-fifths of Anthropic's revenue are also investors in it, and Microsoft, which turns Claude on by default in Copilot, is among the buyers of the new stock.

In federal work, recent Pentagon access and procurement terms have favored OpenAI, though none of these facts measures government-wide use or spending: a defense official said on October 5th that the Pentagon "has ceased the use of Anthropic products," and a federal appeals court in Washington upheld one of the two designations behind that decision on September 25th.[30][31] OpenAI's ChatGPT Mil is "accredited for CUI at Impact Level 5" on the Pentagon's GenAI.mil platform,[32] and in civilian procurement Anthropic's "$1 per user deal will now be available until Oct. 31," a deadline that has since passed; on October 27th GSA announced an expanded agreement with a 40% discount on token-based usage across the Claude models available to the government and no minimum orders,[33] while OpenAI's government-wide deal offers "discounted, consumption-based pricing until Dec. 31, 2028."[33]

Matched list prices

The gaps between token shares and dollar shares in the scoreboards above fit a mix of cheaper models carrying more of the volume, which makes list prices worth a look. On their published price lists, two pairs of models now carry identical standard rates. OpenAI's GPT-6.1 Sol and Anthropic's Claude Sonnet 5.5 both list at $2 per million input tokens and $10 per million output tokens; GPT-6 Luna and Claude Haiku 5.5 at $0.10 and $0.50.[34][35] The third match that existed in October is gone. Anthropic launched Fable 5.2 on November 9th at $8 and $40, which CNBC reported as 20% below the 5.1 generation, and that now undercuts GPT-6 Astra at $10 and $50.[36][34] The middle tiers don't line up at all: Anthropic's Opus 5.5 lists at $4 and $20, and OpenAI has added GPT-6.1 Terra at $1 and $5.[35][34] When OpenAI launched GPT-6 Sol in September, Quartz reported, it priced it at "$2 per million input tokens and $10 per million output tokens, down from $4 and $20 respectively for GPT-5.6 Sol."[37] OpenAI's pricing page now says "GPT-5.6 Sol's promotional pricing is available until November 30, 2026."[34] Matching list rates don't make the models equivalent, and caching, batch pricing, negotiated discounts and partner markups all change what a buyer actually pays.

The match breaks on long prompts, in both directions. Claude Haiku 5.5 is priced by prompt length, so prompts over 100,000 tokens cost $0.50 and $2.50 per million.[35] When we retrieved OpenAI's list on October 12th it showed a long-context rate for Luna of $0.20 and $0.75, and for Sol of $4 and $15, against Anthropic's statement that its newer models other than Haiku 5.5 "include the full 1M token context window at standard pricing."[38] We couldn't read OpenAI's long-context columns on the current page, so those two OpenAI figures carry the October date.

Tokens aren't identical units either. Anthropic says the tokenizer in Claude 4.7 and later models "produces approximately 30% more tokens for the same text" than the previous one, so the same text can cost about 30% more on a newer Claude model at an unchanged per-token rate (Anthropic says the exact increase depends on the content), though it says nothing directly about OpenAI's tokenizer.[35] Cost per task is a third measure: the prospectus says Opus 5.5 cut customers' costs per task by approximately 40% against Opus 5, a company claim we couldn't test.[1]

What the filing leaves open

The filing can't put OpenAI's numbers on the same basis. OpenAI filed confidentially in June but has signaled a 2027 debut, according to CNBC.[4] OpenAI president Greg Brockman said on November 4th at a Wall Street Journal conference that the company could be IPO-ready in 2027, that enterprise is now the majority of its revenue (it was around 30% before), and that annualized revenue grew 70% in the third quarter, which he attributed to the enterprise side of the business doubling.[39] He also said OpenAI "wasn't even in the conversation" when it came to agentic coding, and described the first two quarters as "pretty rough."[39] Those are management statements, unaudited and without a dollar figure attached. The governance side of both listings is the subject of our piece on the mission and the market.

Anthropic priced its offering at $83 a share on November 18th, above the $70 to $78 range,[5] according to reports from CNBC, Reuters and the Wall Street Journal citing sources familiar with the matter;[40] the stock is due to start trading on Nasdaq on November 19th.[41] A price above the range suggests investors wanted the shares, but it doesn't say how they treated a quarter booked gross. The question of who sells more to enterprises waits for OpenAI's public S-1, which would come ahead of the 2027 debut the company has signaled. Until OpenAI files publicly, the outside panels are the only view of both labs at once, and the newest of them put Anthropic ahead by margins ranging from about 3 points on Ramp's token panel to 24 in Menlo's survey.


Sources

  1. Anthropic, PBC, preliminary prospectus (Form S-1/A): prospectus summary, risk factors, MD&A and revenue note (copy hosted by Fidelity Institutional; the amendment is also on EDGAR at sec.gov/Archives/edgar/data/2130357/000162828026072484/anthropicpbc-sx1ax2.htm, which we couldn't open) Inline ↗

    4 passages checked · read October 12, 2026
    …In the week of Sep 7, 2026, OpenAI models took just over 50% of combined spend between the two labs.…
    …That is the first week at or above the 50% line since the week of Feb 24, 2024.…
    …Measured in tokens, OpenAI models took 78% of combined OpenAI and Anthropic usage in the week of Sep 14.…
    …Share charts are computed within the segment named in each chart's footer.…
  2. Bloomberg Law, article teaser (October 2026) Inline ↗

  3. CNBC, live coverage of Anthropic's S-1 (October 26, 2026) Inline ↗

    1 passage checked · read October 12, 2026
    …OpenRouter’s customers are mostly A.I. start-ups and tech companies, so the data does not give a complete picture of the enterprise market.…
  4. CNBC, "Anthropic reports $17 billion in Q3 revenue, purchase commitments of $895.4 billion in revised S-1" (November 2, 2026) Inline ↗

    1 passage checked · read October 12, 2026
    …start of the year, Anthropic, with its Claude Code and Cowork lead, had about three quarters of the business spending on the two labs tracked by OpenRouter, across some 120,000 companies.…
  5. Anthropic, "Anthropic launches roadshow for proposed initial public offering" (November 10, 2026) Inline ↗

    4 passages checked · read October 12, 2026
    …DeepSeek leads at 24.5% of the week's requests.…
    …in the most recent complete week Rank Author Share of requests Change in requests 1 DeepSeek 24.5% +14% 2 OpenAI 18.6% +3% 3 Google 17.8% -16% 4 Z.ai 13.4% +107% 5 Qwen 5.4% -12% 6 Anthropic 3.9% +59% 7 Xiaomi 2.9% +21%…
    …requests 1 DeepSeek 24.5% +14% 2 OpenAI 18.6% +3% 3 Google 17.8% -16% 4 Z.ai 13.4% +107% 5 Qwen 5.4% -12% 6 Anthropic 3.9% +59% 7 Xiaomi 2.9% +21% 8 Mistral AI 2.0% +1% 9 Tencent 1.7% All other authors 9.7% +19% Tool Calls…
    …None of them measures users or spend.…
  6. OpenRouter, "The Token State of Play" (September 22, 2026) Inline ↗

    1 passage checked · read October 12, 2026
    …Ara Karajian (아라 카라지안), an economist at Ramp, said an analysis of data from about 70,000 of its customers in mid-September showed spending on OpenAI models surpassed Anthropic for the first time.…
  7. Ramp AI Index, page data and methodology (retrieved October 12, 2026); Omniscient Media analysis of its embedded series Inline ↗

    1 passage checked · read October 12, 2026
    …OpenRouter nor the outlets citing its data break out how much of Astra's 19% came from net-new usage versus routing changes toward the newer model.…
  8. Ramp AI Index, October 2026 data and week ending November 1, 2026 (retrieved November 16, 2026) Inline ↗

    2 passages checked · read October 12, 2026
    …43.8% of U.S. businesses paid for subscriptions or tokens from Anthropic, up 0.34 percentage points month over month.…
    …OpenAI underperformed overall AI adoption, rising only 0.09 percentage points to 39.8% of businesses.…
  9. Menlo Ventures, "2026: The State of Enterprise AI" (November 3, 2026) Inline ↗

    1 passage checked · read October 12, 2026
    …per @tryramp data Today's update of Ramp AI Index shows 34.4% of businesses using Anthropic versus 32.3% using OpenAI.…
  10. Trending Topics, on The Wall Street Journal's report (October 8, 2026) Inline ↗

    2 passages checked · read October 12, 2026
    …is generative AI’s first “killer use case,” with $4B in enterprise spend Anthropic commands 40% enterprise LLM API market share, compared to 27% for OpenAI AI Boom vs. Bubble For all the fears of over-investment, AI is spreading…
    …report Data Sources and Methodology Survey Approach This report synthesizes findings from a survey of 495 U.S. enterprise AI decision-makers, conducted in partnership with an independent research firm, November 7–25, 2025.…
  11. Michael Parekh, on The Wall Street Journal's report (October 2026) Inline ↗

    2 passages checked · read October 12, 2026
    …16, 2026 Shawn Carolan , Amy Wu Martin and Sam Borja Report Perspective Portfolio 2025: The State of Generative AI in the Enterprise For all the fears of over-investment, AI is spreading across enterprises at a pace with no…
    …December 09, 2025 Tim Tully , Joff Redfern , Deedy Das and Derek Xiao Report Perspective Portfolio 2025: The State of AI in…
  12. OpenRouter, Rankings (week beginning October 5, 2026; retrieved October 12, 2026) Inline ↗

    3 passages checked · read October 12, 2026
    …First, let’s get one thing out of the way: we are investors in OpenAI.…
    …78% of surveyed enterprise CIOs are using OpenAI models in production, either hosted directly or via cloud service providers (CSPs).…
    …44% of enterprises are now using Anthropic in production, rising to over 63% including testing.…
  13. Digital Today, on The Wall Street Journal's report (October 11, 2026) Inline ↗

    2 passages checked · read October 12, 2026
    …over the same stretch, has moved steadily the other way — climbing from 54.3% in July to 55.6% in August and now 56.4% in September.…
    …Anthropic’s chatbot has held roughly steady in the 9-10% range for three months now — 9.6% in July, 9.3% in August, and 9.6% in September — after a sharp run-up earlier in the year that took it from 8.0% in April to its current…
  14. AI Weekly (October 8, 2026) Inline ↗

    1 passage checked · read October 12, 2026
    The share of firms switching AI model providers reached a record 8% in September as Anthropic and OpenAI competed with new models and price cuts.
  15. Ara Kharazian, Ramp, on X (May 13, 2026) Inline ↗

    8 passages checked · read October 12, 2026
    …Sales through the two companies rose from 11% of revenue in 2023 to 32% in 2024 and nearly half in 2025.…
    …The company paid roughly $351 million to the platforms in distribution fees, according to a Reuters analysis, suggesting that the cloud providers collected…
    …Anthropic records the full value of marketplace contracts, under which customers buy access to Claude through a cloud provider's marketplace, as…
    …About $3.8 billion of Anthropic's revenue came from customers paying based on their use of the company's Claude AI system, while subscription revenue totaled…
    …Two unnamed customers each accounted for 12% of revenue last year.…
    …Anthropic warned that many of its largest customers are not bound by long-term contracts and could reduce or halt their spending.…
    …At the end of 2025, Anthropic had $54.6 billion in non-cancellable hosting and computing commitments.…
    …By early 2026, its total long-term commitments had exceeded $417 billion, covering 3.5 gigawatts of dedicated computing capacity.…
  16. Andreessen Horowitz (January 30, 2026) Inline ↗

    2 passages checked · read October 12, 2026
    …companies said that Microsoft remains OpenAI's primary cloud provider, and that OpenAI products will ship first on Azure unless Microsoft decides otherwise.…
    …However, Microsoft will no longer pay a revenue share to OpenAI, according to Monday's blog post .…
  17. OfficeChai, on Similarweb data (October 2026) Inline ↗

    2 passages checked · read October 12, 2026
    …Microsoft enables Anthropic models on by default for most customers in commercial cloud (excluding EU/EFTA and UK).…
    …Anthropic models aren't available for federal customers in GCC or for any customers in GCC High and Department of Defense (DoD) environments.…
  18. Ramp (October 6, 2026) Inline ↗

    1 passage checked · read October 12, 2026
    …About long-term support (LTS) models Further reading About base models Important On March 18, 2026, GitHub designated GPT-5.3-Codex as the base model.…
  19. CNBC (October 8, 2026) Inline ↗

    1 passage checked · read October 12, 2026
    …Open-weight models (e.g., DeepSeek, Kimi K2.7) and models not covered by GitHub’s data retention agreement (e.g., Fable 5) are excluded from default enablement regardless of your policy.…
  20. TechCrunch, citing the Financial Times (October 8, 2026) Inline ↗

    1 passage checked · read October 12, 2026
    …EVP of Copilot Charles Lamanna, a potential Satya Nadella successor; a source says Microsoft is on track to pay Anthropic $2B+ per year (Aaron Holmes/The Information) Open Links In New Tab Mobile Archives Advertise Site News…
  21. Anthropic, Series H announcement (May 28, 2026) Inline ↗

    2 passages checked · read October 12, 2026
    …However, that estimate has since been slashed by more than a third.…
    …Usage of Anthropic's coding assistant, Claude Code, recently dropped to roughly 30,000 employees—down from approximately 60,000 earlier this year.…
  22. CNBC (October 7, 2026) Inline ↗

    2 passages checked · read October 12, 2026
    …The Pentagon "has ceased the use of Anthropic products", the department official said in a statement on Monday.…
    …OpenAI tools have been more widely adopted in recent months in some military departments, according to two people familiar with AI use in the defence department The Pentagon's…
  23. The Information, "Anthropic Lowers Gross Margin Projection as Revenue Skyrockets" (paywalled; we couldn't open it, and the figures are as retold by Tiger Brokers); Tiger Brokers, syndicating The Information (the copy we read) Inline ↗

    2 passages checked · read October 12, 2026
    …A San Francisco federal judge ruled last month that one designation was illegal.…
    …The designation prevents the U.S. military from using Anthropic's models and blocks defense contractors from using them in their work with the agency.…
  24. SemiAnalysis (October 2026) Inline ↗

    1 passage checked · read October 12, 2026
    …The department said ChatGPT Mil is accredited for CUI at Impact Level 5, a DoD cloud-security standard for certain sensitive unclassified information.…
  25. CNBC (April 27, 2026) Inline ↗

    2 passages checked · read October 12, 2026
    …The agreement now offers discounted, consumption-based pricing until Dec. 31, 2028 instead of the previous $1 per agency deal.…
    …Subscribe Close AI Anthropic extends its GSA OneGov deal for Claude by one month The $1 per user deal will now be available until Oct. 31, according to a website update.…
  26. Microsoft Learn, Anthropic as a subprocessor for Microsoft Copilot offerings (retrieved November 16, 2026) Inline ↗

    4 passages checked · read October 12, 2026
    …GPT-5.6 Sol’s promotional pricing is available at least through November 21, 2026.…
    …Cache writes Output gpt-6-astra $10.00 $1.00 $12.50 $50.00 $20.00 $2.00 $25.00 $75.00 gpt-6.1-sol $2.00 $0.10 $2.50 $10.00 $4.00 $0.20 $5.00 $15.00 gpt-6-luna $0.10 $0.01 $0.125 $0.50 $0.20 $0.02 $0.25 $0.75 All models Batch Short…
    …$20.00 $2.00 $25.00 $75.00 gpt-6.1-sol $2.00 $0.10 $2.50 $10.00 $4.00 $0.20 $5.00 $15.00 gpt-6-luna $0.10 $0.01 $0.125 $0.50 $0.20 $0.02 $0.25 $0.75 All models Batch Short context Long context Model Input Cached input Cache writes Output…
    …Model Input Cached input Cache writes Output Input Cached input Cache writes Output gpt-6-astra $10.00 $1.00 $12.50 $50.00 $20.00 $2.00 $25.00 $75.00 gpt-6.1-sol $2.00 $0.10 $2.50 $10.00 $4.00 $0.20 $5.00 $15.00 gpt-6-luna…
  27. GitHub Docs, "Fallback and LTS models" (retrieved October 12, 2026) Inline ↗

    4 passages checked · read October 12, 2026
    …Claude Haiku 5.5 is priced by prompt length: a request whose prompt is over 100,000 tokens pays higher prices.…
    …pricing  Claude 4.6 and later models (except Claude Haiku 5.5) and Claude Mythos Preview include the full 1M token context window at standard pricing.…
    …This tokenizer produces approximately 30% more tokens for the same text.…
    …Claude Sonnet 4.6 and earlier models use the previous tokenizer.…
  28. GitHub Changelog (July 29, 2026) Inline ↗

    1 passage checked · read October 12, 2026
    …GPT-6 Sol is priced at $2 per million input tokens and $10 per million output tokens, down from $4 and $20 respectively for GPT-5.6 Sol.…
  29. Techmeme, summarizing The Information (October 12, 2026) Inline ↗

    3 passages checked · read October 12, 2026
    …OpenAI told investors that it hit roughly $50 billion in annualized revenue at the end of September, CNBC confirmed, lower than the the $68 billion figure widely reported…
    …A person familiar with the matter said the $68 billion figure included gross revenue from OpenAI's partners, which helps investors make a more direct comparison with Anthropic.…
    …revenue, OpenAI touted 77% total run rate growth during its third quarter, as well as 107% run rate growth for its enterprise business during the same period, the person said.…
  30. BBC News (October 5, 2026) Inline ↗

    1 passage checked · read October 12, 2026
    …That figure was devised via “attempts by OpenAI’s own investors to produce a direct comparison with Anthropic’s annualised revenues,” per the FT.…
  31. CNBC (September 25, 2026) Inline ↗

    2 passages checked · read October 12, 2026
    …Anthropic counts the gross amount of sales through cloud partners like Amazon.com Inc. as revenue.…
    …OpenAI recognizes only its share of ...…
  32. Military Times (August 31, 2026) Inline ↗

    1 passage checked · read October 12, 2026
    …G in February, adoption has continued to grow across global enterprise customers, and our run-rate revenue crossed $47 billion earlier this month.…
  33. FedScoop (September 17, 2026); GSA news release (October 27, 2026) Inline ↗

    2 passages checked · read October 12, 2026
    …Anthropic claimed at the end of July that its annualized revenue run rate was up sevenfold year-over-year to $65 billion.…
    …2026 6:29 PM EDT Isabel O'Brien @in/xisabelobrienx @xisabelobrienx WATCH LIVE Key Points Research firm New Constructs wrote in a note on Tuesday that Anthropic's IPO would be "the most ridiculous IPO of 2026," posing a historic…
  34. OpenAI API pricing (retrieved November 16, 2026) Inline ↗

    2 passages checked · read October 12, 2026
    …projected last month that the gross margin for its artificial intelligence products sold to enterprises and application developers would reach 40% in 2025.…
    …In contrast, OpenAI expects a 2025 gross margin of about 46%, a figure that already encompasses the inference costs for both paying and free ChatGPT users.…
  35. Anthropic, Claude API pricing (retrieved November 16, 2026) Inline ↗

    2 passages checked · read October 12, 2026
    …of models—which both companies market as the intended daily driver for most users—Anthropic is an overwhelmingly better deal, offering ~5x the API-equivalent value across the board.…
    …Source: SemiAnalysis Tokenomics Model Despite being just 10% of overall revenue, subscriptions can take up over 40% of inference compute and lower blended revenue per MW by ~$36M.…
  36. CNBC, "Anthropic releases first frontier models since Amodei's call for slowdown as IPO buzz builds" (November 9, 2026) Inline ↗

  37. Quartz (September 23, 2026) Inline ↗

  38. Anthropic, Claude API pricing, "Long context pricing" (retrieved November 16, 2026); OpenAI's long-context rates are from its API pricing page as retrieved October 12, 2026 Inline ↗

  39. TechCrunch, "OpenAI president Greg Brockman says company could be IPO ready in 2027" (November 5, 2026) Inline ↗

  40. CNBC on X, "Anthropic prices IPO at $83 per share, above expected range" (November 18, 2026; source-reported; the full CNBC article was not available when we checked); Crunchbase News, "Anthropic Prices IPO Shares At $83, Above Range" (November 18, 2026; the account we read, which attributes the price to CNBC, Reuters and the Wall Street Journal citing sources familiar with the matter; we did not open the Reuters or Journal reports) Inline ↗

  41. CNBC, "Anthropic IPO guide: The price range, date and more" (November 16, 2026) Inline ↗

X LinkedIn Email

Get this every weekday.

The Omniscient Bulletin: consequential AI, explained and evaluated. 5 to 7 items a day with the take, not the recap.