~$1.8B of the $64.4B IPO goes to selling stockholders (23.6M shares at $78). [7] Financial statements run to 30 September 2026; pricing is expected the week of 16 November.
Valuation came in well below earlier expectations. At $70-$78 on 825M shares, Reuters puts the top of the range at up to $1.02T, well short of the amounts discussed earlier. [10] CNBC puts it at up to about $1.405T fully diluted. [2] Investors had discussed $1.5T in October and as much as $2T earlier in the year. [11]
Seven partners anchor the book: AMD, Broadcom, Google, Micron, Microsoft, NVIDIA and TSMC are buying $26.0B of Class N at the IPO price. [12] Net proceeds to Anthropic are about $84.6B at the midpoint. [13]
The operating picture got stronger. Run-rate revenue reached about $81B at 31 October, up $10B from 30 September. [4] That compares with Q3 revenue of $16,956M, about $67.8B annualized. [9]
Anthropic is a Delaware public benefit corporation whose stated purpose is the responsible development and maintenance of advanced AI for humanity's long-term benefit. [14] Nine-month 2026 revenue was $33,242M: $26,306M consumption-based and $6,936M subscription-based. [15] It has applied to list Class A shares on the Nasdaq Global Select Market under ANT. [16]
The investor question is whether revenue growing 46.8% QoQ can outrun a commitment stack that now totals about $900.9B. [17][6] Q3 shows the cost of falling behind on capacity: short-term compute pushed the quarter into an operating loss. [5] Against that, the company holds $131.8B of cash and short-term investments. [18]
The falsifiable near-term KPI is the Q4 exit rate. Investors expect $110B of annualized revenue by year-end, per The Wall Street Journal. [19] Run-rate revenue was about $81B at 31 October, so roughly $29B more is needed in two months. [4]
- Adjusted operating income was $4,081M for the nine months, a 12% margin, with adjusted net income of $5,591M. [20][21]
- Nine-month operating cash flow was $6,569M. [22]
- Net revenue retention was about 900% at 30 September. [23]
- More than 1,900 customers spend at least $1M a year, and about 250 spend at least $10M. [24][25]
- SpaceXAI capacity was expected to be fully delivered in December 2026, which should reduce reliance on costlier short-term compute. [26]
- Q3 swung to a $567M operating loss as compute reached 58% of revenue. [5][27]
- Convertible notes were carried at $130.5B at 30 September, up from $108.6B in June. [28][29]
- OpenAI's 20-40% discounts lifted its share of Rippling's coding work from 16% to 52%. [30]
- Amazon rollover notes covering 1,564M shares convert quarterly after the IPO at the lowest of their fixed price, the IPO price and the 10-day average. [31][32]
- The cornerstone placement is not subject to underwriter lock-ups. [33]
- The Founder LLC holds 50.1% of votes, and Anthropic will be a controlled company. [34][35]
The liquidity path is a live IPO. The roadshow launched on 10 November with pricing expected the following week. [36] The range is $70-$78 for 825M Class A shares, 801.4M of them primary. [37][38] A version sent to banks on 9 November, showing about 867M shares at $70-80 and a $69.3B raise, was withdrawn as inaccurate. [39] The final terms cut the raise by about 7%. [40] Anthropic is proceeding despite a sharp sell-off in AI-related stocks last week. [41]
- Where it prices in the range. $70 implies about $916B on outstanding shares, below the $965B Series H post-money. [42][43]
- Underwriters' option: up to 123.75M extra shares. [44]
- Q4 compute mix: whether delivered SpaceXAI capacity replaces short-term compute. [26]
- Lock-up schedule: non-founder insiders can sell 10% from day one, with staged releases of 6% on days 27 and 48 and 7% on days 62 and 76. [45][46]
- Regulatory: the FTC issued a civil investigative demand in October, and the Department of War designation litigation continues. [47][48]
Anthropic monetizes model capability in three ways: metered tokens through its API and cloud marketplaces, seat-plus-usage enterprise subscriptions, and agentic products led by Claude Code. The former head of competitive intelligence at OpenAI describes Anthropic Enterprise as seat fees combined with metered usage. [49] Claude is available on the Claude Platform, Google Cloud's Vertex AI, Microsoft Foundry and Amazon Bedrock. [50]
The company was founded by siblings Dario and Daniela Amodei, and raised a $124M Series A in May 2021 led by Skype co-founder Jaan Tallinn. [51] It is headquartered in San Francisco. [52] Per the S-1, a single Class F share gives the Founder LLC 50.1% of total voting power, recalculated to hold that percentage as other voting shares change. [53] The company remains a Delaware public benefit corporation. [54]
Management's track record is two-sided. Execution on revenue is visible in the S-1: Q2 2026 revenue of $11,555M compares with $787M a year earlier. [8] Headcount rose about 170% in the year to September while TTM revenue rose about 1,200%. [55] The governance record is less settled. The S-1 says the Department of War designation followed Anthropic's public opposition to use of its technology for mass domestic surveillance or fully autonomous weapons. [56] Retention of builders is a live question after two core Claude Code engineers left on the same day in late October. [57] OpenAI CEO Sam Altman called those departures "not a good sign." [58]
- Frontier models
- Claude Opus 5.5, Sonnet 5.5 and Mythos 5.1 sit in the current lineup offered through the cyber access tiers. [59] Opus 5.5 is the first model in the 5.5 family and targets long-running coding and knowledge work. [60]
- Claude Code
- Runs on developer machines, reads and modifies source code, and calls tools and MCP servers under a developer's credentials. [61]
- Claude Enterprise and Team
- Compliance and Analytics APIs expose session transcripts, tool calls and per-user usage to governance teams. [62]
- Claude Cowork
- A health system cut administrative workload by 20% using Cowork and Claude Code, per partner Intuitive.ai. [63]
- Security programs
- Project Glasswing partners found at least 129,000 verified software vulnerabilities between April and July 2026 using Mythos models. [64]
The S-1 splits revenue only into consumption-based ($13,064M in H1 2026) and subscription-based ($3,222M); Claude Code and other products are not broken out. [65][66][67]
Expert-call rows reflect individual practitioners' accounts, not company disclosures.
Concentration is moderate and falling. For the nine months to 30 September 2026, no single customer exceeded 10% of revenue and the top five made up about 25%. [78] In H1 2026 the largest disclosed customer, Customer C, was 11%. [79] In Q3, about 79% of revenue came from customers outside the top five. [80] The filing does not name the top five or say whether they are cloud providers. [81]
Expansion inside accounts is the strongest signal. The S-1 reports more than 11,900 customers with at least $100,000 of TTM revenue at 30 September 2026, up from more than 950 a year earlier. [82] More than 1,900 customers exceed $1M and about 250 exceed $10M. [83] Net revenue retention was about 900%. [84] JPMorgan's executive director reports overall AI usage growing more than 5% month over month, with Claude Code rollout still limited while controls are built. [85]
The ledger also shows where control of the workflow is slipping to intermediaries. At Microsoft, the move to Copilot CLI routes Claude underneath rather than abandoning it. [86] Adobe's FinOps manager says engineers switch between Anthropic and OpenAI on tooling, negotiated discounts and observed value. [75] Both accounts point to Claude winning the hardest workloads while losing direct ownership of the developer seat.
Q3 values marked derived are 9M 2026 less H1 2026. Gross margin and free cash flow are calculated from reported lines; free cash flow excludes $1,955M of nine-month prepayments for leased assets and equipment. [112]
Quarterly revenue per the filings was $4,731M in Q1 2026, $11,555M in Q2 and $16,956M in Q3. [8][9] Q3 grew about 46.8% QoQ and 1,127% YoY. [17] Sequential growth slowed to 47% from 144% in Q2. [113]
Profit moved the other way. Q2 was the first profitable quarter, with $974M of operating income, but Q3 returned to a $567M operating loss. [5] The adjusted operating margin fell to 15% in Q3 from 19% in Q2. [103][114]
The October investor-day figure of $16.8B understated audited Q3 revenue by about 1%. [115]
Management attributes the Q3 operating loss primarily to $5.0B of compute under short-term commitments, which cost significantly more than comparable multi-year compute. [5] Compute and infrastructure expense was $9,863M, 58% of revenue, up from $5,558M and 48% in Q2. [27]
The second driver is $2.0B of non-cash donation-matching charges. [5] The quarterly presentation shows $1,960M of non-cash charitable contributions. [116] Donation-matching expense was $2.1B for the nine months, about 6% of operating expenses. [117]
Strip both and the core business still generated operating leverage. Adjusted operating income rose to $2,463M in Q3 from $2.2B in Q2, but margin compressed because revenue grew faster than adjusted profit. [103][114]
Cash generation held up. Nine-month operating cash flow was $6,569M against $925M of capex and $1,955M of prepayments for leased assets and equipment. [22][112] Even after prepayments, nine-month free cash flow is roughly $3.7B.
Cash and short-term investments were $131.8B at 30 September, with the mix shifting to $36.3B of short-term investments. [18][118] Debt is still $6.2B of Amazon notes principal. [119] Convertible notes rose to $130.5B on the balance sheet. [28] That accounts for the $94.3B nine-month fair-value charge. [106]
Marketplace revenue remains booked gross, as principal. [120] Those sales were about 48% of H1 revenue. [121] The fees CSPs deduct are not quantified, so a net-revenue figure is unavailable.
- Google: at least ~$121.3B from April 2026 to July 2033 regardless of usage, up from ~$114.5B in the first filing. [122][123] Additional TPU capacity for 2028-2031 is estimated at $125-150B and sits outside the obligations table. [124]
- AWS: at least $110.0B from May 2026 to April 2036, regardless of usage. [125]
- Microsoft: at least $44.9B from November 2026 to June 2033. [126]
- Broadcom: about $161.2B of equipment-lease rent excluding interest, of which ~$125.2B relates to the 2026 lease arrangement. [127]
- SpaceXAI: up to $120.5B into 2029, reaching ~$3.1B a month once all capacity is delivered, expected in December 2026. [26]
- Nscale and Lambda: up to ~$44.6B and ~$41.3B, paid as capacity is delivered and accepted. [128][129]
- AMD: above $20.0B if the full capacity is procured; AMD will also buy up to $5.0B of Class N stock against delivery milestones. [130]
The amended filing replaces the earlier partial aggregates with one total: about $895.4B of committed spend as of and after 30 September 2026. [131] The 10 November amendment lifts that figure to about $900.9B. [6] About $447.5B is expected before 31 December 2030, $69.8B falls due within 12 months, and substantially all of the remainder lands in 2031-2045. [131] About $213.8B of the total is subject to termination rights or conditions before commencement. [132]
More sits outside that table. The filing also lists about $451.2B of estimated obligations for leases not yet commenced and about $100.2B of later commitments. [133] For comparison, fixed or minimum commitments were about $265.6B at 30 June in the first filing. [134]
Against revenue, the $69.8B due in the next 12 months is about 1.0x the Q3 annualized revenue of $67.8B (16,956 x 4). [131][9] The $447.5B due before 2030 averages about $105B a year over 4.25 years, roughly 1.5x the current run-rate. Q3 shows the other side of the trade-off: when contracted capacity lags demand, short-term compute costs more and pushed the quarter into an operating loss. [5]
The filing frames compute as the binding constraint. Anthropic says it has operated in a compute-constrained environment, relies on a limited number of suppliers, and may need more expensive short-term compute during shortages. [135] That risk materialized in Q3, when $5.0B of short-term compute drove the operating loss. [5] It names a conflict at the core of the supply chain: Amazon, Google and Microsoft supply substantial compute while also offering competing models. [136]
Owned capacity is arriving through partners. Theseus Infrastructure, formed with Macquarie Asset Management and GIC, will develop sites with Anthropic as anchor tenant. [137] Fluidstack provides data-center development and operations under a colocation agreement, with no dollar figure disclosed. [138] AMD's first gigawatt of MI450-series GPUs starts deploying in H1 2027. [76]
Productivity inside the company is the offset to compute intensity. Anthropic had about 5,300 employees at 30 September 2026. [139] TTM revenue per employee rose from $1.4M to over $6.6M. [55] In September 2026 Claude wrote over 90% of contributions to Anthropic's own codebase. [140]
Distribution is split almost evenly between direct and cloud marketplaces. CSP marketplace sales rose from about 11% of revenue in 2023 to 47% in 2025 and 48% in H1 2026. [81] GitLab's partnership manager says usage across the first-party API, Vertex and Bedrock can count toward both Anthropic commitments and cloud burndowns, which has repeatedly helped Anthropic win business. [141] Sales and marketing spend rose to $3,808M in H1 2026 from $239M a year earlier. [95]
The services layer is being built to move pilots into production. Anthropic and Hg plan a European consultancy launching by year-end, following a similar $1.5B US initiative. [142] Claude Frontier Academy is a $100M programme to train 10,000 engineers. [50] Block's commercial executive still describes Anthropic as less mature than OpenAI or Google in standardized packaging and procurement. [143]
Each cell is one practitioner's account from a dated expert call; spend shares are proxies, not disclosures.
Anthropic holds the high-capability end of enterprise coding, and the price competition is aimed at that end. The former head of competitive intelligence at OpenAI cites enterprise spending data showing Anthropic overtaking OpenAI among paying businesses, followed by signs of OpenAI recovering ground. [157]
OpenAI is buying share back with price. It is offering 20-40% off list for enterprise models including Codex, and Replit and DataCamp received 30-40% offers and switched primary models. [30]
Anthropic has cut prices too. A review of June-October pricing describes aggressive cuts by both labs, driven by low switching costs and open-weight competition. [158]
No defensible TAM figure appears in the S-1 coverage or company releases reviewed, so a top-down market size is unavailable. The bottom-up evidence comes from spend shares. In Block's commercial executive's peer proxy, OpenAI holds over 50% of model spend, Anthropic is close behind, and Anthropic may capture the largest share of marginal new spend. [153] Anthropic's audited Q3 revenue of $16,956M, about $67.8B annualized, is the observable floor for its own serviceable demand. [9]
Pricing power is eroding at the list level and holding at the workload level. The managed-services director at a North American advisory firm estimates negotiated discounts are down about 10% when customers can commit volume or compare platforms. [159] The same expert estimates comparable completed workflows cost about 15% less through optimization rather than vendor cuts. [159]
Model efficiency is the offset. In Kiro's internal benchmarking, Opus 5.5 completes tasks with about 40% fewer tool calls and half the tokens of Opus 5, and its credit multiplier fell to 2.0x from 2.2x. [160] The former OpenAI competitive-intelligence head argues a more token-intensive Claude task can still be cheaper if it finishes correctly with less human intervention. [49]
The orchestration layer is moving to customers and intermediaries. A Microsoft principal engineer says the shift to enterprise wrappers is chiefly economic: routing, context management, token accounting and budget controls let simple requests go to smaller models while complex work stays on Claude or GPT-class models. [161] Block's executive notes sophisticated buyers increasingly use routing and abstraction layers. [162]
Open-weight models are closing the gap from below. A Cohere technical-staff member estimates the best open-weight models are roughly one generation behind leading closed models. [163] Moonshot AI's Kimi K3.1 Pro rivals leading US frontier models on some benchmarks. [164] Adobe is testing Qwen for basic coding tasks and agents because of its low cost. [146]
The counterweight is that on-prem models are neither cheap nor easy to maintain; they require skilled engineers and continuing infrastructure support. [165] The near-term exposure is therefore concentrated in high-volume, lower-complexity tokens rather than in the frontier coding work that carries Anthropic's premium.
Public shareholders will buy economics with limited control. A single Class F share gives the Founder LLC 50.1% of total voting power, recalculated to preserve that share as other voting stock changes. [53] The Class F share sunsets 450 days after the Founder LLC has had fewer than three qualifying members for 90 consecutive days. [166] Anthropic will be a controlled company under Nasdaq rules. [167]
A second layer sits with non-shareholders. The Long-Term Benefit Trust holds the Class T share and can generally elect and remove a bare majority of the board. [168] It is governed by five voting trustees. [169] Harvard astrophysicist Christopher Stubbs was named to the trust on 21 October. [170] Employees other than the co-founders hold Class E shares, which carry limited voting rights and no economic rights. [171] Class N shares are non-voting and convert to Class A on a qualifying transfer. [172]
The third layer is the Anthropic Public Benefit Fund. Anthropic has already contributed 846,840,960 Class N shares to it, equal to 5% of fully diluted shares at 30 June assuming full conversion of the Amazon notes. [173] After the IPO the board may issue up to a further 2% of outstanding shares each time the share price rises 50% above the prior milestone, starting from the IPO price. [174] A milestone requires closes above the threshold on 90 of the trailing 120 trading days, and the filing states no separate stockholder approval for those issuances. [175] For a public buyer, this escalator is dilution that kicks in at +50%, +125% and so on from the offer price.
The risk factors centre on four exposures. Anthropic says it has operated in a compute-constrained environment and may need more expensive short-term compute. [135] It warns that competition could pressure prices and that agentic workloads make profitable pricing harder. [176] It flags that Amazon, Google and Microsoft supply its compute while offering competing models. [136] It also warns that models could pose catastrophic or existential risks if safeguards prove insufficient. [177]
On litigation, Anthropic entered a $1.5B settlement in the Bartz copyright case in 2025 and warns adverse fair-use rulings could force injunctions or retraining. [178] On the Department of War designation, a district court ruled in Anthropic's favour in August 2026, but a divided D.C. Circuit denied its petition on 25 September. [179]
Use of proceeds is general corporate purposes, including about $3.8B for RSU tax withholding, [180] as well as including compute and infrastructure capex, hiring, possible acquisitions and RSU tax withholding. [181] Anthropic receives nothing from shares sold by selling stockholders. [182] The amended filing states there are no agreements for material acquisitions. [183] Co-founders have pledged 80% of their personal equity to philanthropy, though the pledges are voluntary and revocable. [184] Twenty-two underwriters are named, led by Goldman Sachs and Morgan Stanley. [185] New investors face dilution of $65.79 per share at the $74 midpoint. [186]
Severity is an editorial ranking of exposure size and proximity, not a probability.
Consensus belief. The market had priced Anthropic for a $100B+ year-end run-rate, with some investors projecting $120B. [200] What the evidence shows. Nine-month gross margin, derived from the amended S-1, is about 85.5%, and operating cash flow was $6,569M. [91][22] Q3 nonetheless swung to a $567M operating loss as compute reached 58% of revenue. [5][27] The contested point. Whether Q3's short-term compute squeeze is a one-quarter timing gap or the first sign that a commitment stack now at about $900.9B runs ahead of revenue. [6] The signal that changes the read. Q4 compute as a share of revenue once SpaceXAI capacity is fully delivered in December, and the Q4 exit rate against the $110B investors expect. [26][19]
Anthropic amended S-1 (2 Nov 2026). Nine-month operating cash flow was $6,569M and adjusted operating income $4,081M, a 12% margin. [22][20] Management ties the Q3 operating loss to $5.0B of short-term compute and $2.0B of non-cash donation charges. [5] On that reading, both are timing and policy items rather than structural cost. Q3 adjusted operating income still rose to $2,463M. [103]
Anthropic S-1 (26 Oct 2026). H1 2026 operating cash flow was $3,929M against capex of $482M. [110][111] Q2 produced $1.0B of operating income at an 8% margin. [114] That is the quarter that shows the cost structure when capacity is not being bought on the spot market.
Memeburn (22 Oct 2026). Its ratios show compute spend falling from about $2 per $1 of revenue to about 55 cents. [201] Jim Cramer, CNBC (1 Nov 2026). "I think that the goal of the bears out there is to derail the Anthropic IPO." [202] On the bull reading, the $131.8B cash pile funds the $69.8B due over the next 12 months nearly twice over. [18][131]
Aswath Damodaran, NYU Stern (Oct-Nov 2026). He called the adjusted Q3 profit "often a fictional number" and argued $1.5T hinges on 30% margins. [203][204] He has since estimated Anthropic's equity value at about $1.48T, warning that excessive pricing and restrictive voting rights could discourage investors. [205] The amended filing supports his margin point: the nine-month adjusted operating margin is 12%. [20]
Anthropic amended S-1 (2 Nov 2026). Compute and infrastructure rose to $9,863M in Q3, 58% of revenue, from 48% in Q2. [27] Aggregate committed spend is about $895.4B, with $447.5B expected before the end of 2030. [131] The company's own risk factors warn that agentic workloads are more compute-intensive, which could make profitable pricing difficult. [176]
Reuters (via 2 Nov reporting). Some investors lowered valuation expectations as sequential growth slowed and AI-infrastructure financing drew scrutiny. [206] Jason Lemkin, 20VC (22 Oct 2026). "Nobody 10xes forever." [207] On the bear reading, growth slowing to 47% QoQ from 144% while compute intensity rises 10 points is how a mismatch begins. [113][27]
What it is. The question is whether contracted compute is a capacity asset or a fixed liability. The amended S-1 puts aggregate committed spend at about $895.4B, running to 2045. [131] About $213.8B is subject to termination rights or conditions before commencement. [132]
Role. Compute is both the product input and the largest cost. Compute and infrastructure was $9,863M in Q3, 58% of revenue. [27] Q3 showed that under-contracting is also costly: $5.0B of short-term compute drove the operating loss. [5]
Exposure. About $69.8B falls due within 12 months and $447.5B before the end of 2030. [131] Google TPU capacity for 2028-2031 adds an estimated $125-150B outside the obligations table. [124]
Strength and risk. The strength is flexibility in part of the stack: Nscale and Lambda bill only as capacity is delivered and accepted. [128][129] The risk is the take-or-pay core: Google, AWS and Microsoft are owed regardless of usage. [122][125][126]
The numbers. For compute to stay at Q3's 58% of revenue, the $447.5B due before 2030 needs about $772B of cumulative revenue. [27][131] Over 4.25 years that is about $182B a year, about 2.7x the Q3 annualized revenue of $67.8B. [9] Getting back to Q2's 48% ratio would need about $219B a year.
The contested question. Whether the compute ratio reverts toward Q2's 48% once contracted capacity replaces short-term purchases. SpaceXAI capacity was expected to be fully delivered in December 2026, so Q4 is the first test. [26]
Both sides of this debate are drawn from enterprise buyers on dated expert calls. Each voice is one practitioner, so the panel shows direction rather than market share.
AI partnership manager, GitLab (29 Jul 2026). "The flexibility that we've had with Anthropic, I would say, is probably the one reason why they have continuously won our business." [208] Anthropic is GitLab's largest AI vendor by spend, ahead of OpenAI and Google, within an eight-figure budget. [68] Cheaper, more capable models led GitLab to use more, not spend less. [147]
AI transformation executive, Hanger (25 Oct 2026). "Swapping would in fact be a project on its own." [209] Claude replaced ChatGPT Enterprise and delivered a ~40% cut in approval-cycle time and 50%+ more cases closed per day by senior clinicians. [72]
Former head of competitive intelligence, OpenAI (14 Oct 2026). "Workflow-level switching costs are much higher." [210] Replacing an agent wired into repositories, testing and deployment is far harder than swapping models. [154] Director of AI product security, HP (30 Jul 2026). "Claude is very hyperspecific and really well-loved by our engineering teams." [211]
OpenAI pricing campaign, per IT-trade reporting (26 Oct 2026). OpenAI is offering 20-40% off list for enterprise models including Codex. [30] At Rippling, OpenAI's share of coding work rose from 16% in May to 52%, and Replit and DataCamp switched primary models. [30]
Commercial executive, Block (29 Jul 2026). Claude Code uses about four times as many tokens per task as Codex on the same work, making real-world cost high and hard to predict. [145] Internal estimates rose three to five times when usage or plans changed. [212]
Principal engineer, Microsoft (8 Sep 2026). "The biggest issue with the customer that they are complaining is the token economics." [213] Microsoft cancelled most internal Claude Code licences for Copilot CLI. [74] Executive director, JPMorgan Chase (8 Oct 2026). "For 90% of the tasks, you really don't need Opus." [214] FinOps manager, Adobe (2 Sep 2026). Adobe is moving to model-agnostic tools partly because Claude Code is less flexible for other models. [75]
Every voice on both sides agrees Claude leads on the hardest coding work. Even Adobe's FinOps manager says "the Anthropic models still hold a little bit of an edge" for coding. [215] The split is over who captures the value. GitLab and Hanger describe seat-level lock-in through workflow integration. [141][216] Microsoft, JPMorgan and Adobe describe the routing layer moving in-house, which keeps Claude for complex tasks while pushing the 90% of simpler tokens to cheaper models. [161][217][218] If that pattern spreads, revenue per enterprise account grows more slowly than usage.
- How much of the $9,863M Q3 compute bill was short-term spot capacity that falls away once SpaceXAI is fully delivered? [27][26]
- Which of the $213.8B of conditional commitments is likely to be cancelled or deferred? [132]
- Who are the top five customers behind about 25% of revenue, and are any cloud providers? The filing does not say. [81]
- How large are the marketplace fees netted from the 48% of revenue booked gross through CSPs? They are not quantified. [120][121]
- How much of the ~$81B October run-rate is consumption that could fade if price cuts spread? Run-rate annualizes consumption over a trailing 28 days. [4][221]
- Why does the fully diluted value differ between the filing's share components (~17.5B shares) and the ~$1.405T reported? [2]
The range resets the valuation conversation lower. Reuters puts the top of the range at up to $1.02T, "well short of the previously discussed amount," in the latest sign of cooling appetite for AI. [10] On a fully diluted basis the top of the range is worth up to about $1.405T, per CNBC. [2] Both are below the $1.5T investors discussed in October and the $2T talked about earlier in the year. [11]
Which basis is like-for-like with the $965B Series H is the key question. On the 13.09B shares outstanding after the deal, the range spans about $916B to $1,021B, so the low end is below the May post-money and the midpoint ($968B) is roughly flat with it. [42][43] Adding the options, RSUs and Amazon rollover-note shares excluded from that count gives about 17.5B shares, or roughly $1.36T at $78. [227][228][31] The remaining gap to CNBC's ~$1.405T is not reconcilable from the filing alone.
The deal was marketed through a stumble. A version sent to banks on 9 November showed about 867M shares at $70-80 to raise $69.3B before Anthropic withdrew it as inaccurate. [39] The final terms cut the raise by about 7%, or $4.9B. [40] Anthropic is pressing ahead despite a sharp sell-off in AI-related stocks last week. [41] Other AI infrastructure listings have been delayed or withdrawn as higher rates hit shares of operators such as CoreWeave. [36]
Funding-round record, $169.4B total funding across 20 recorded transactions. Per-round lead investors are not separated in the record. [230]
The private-mark ladder steps up without a down round: $61.5B in March 2025, $183B in September 2025, $380B in February 2026 and $965B in May 2026. [230] The two April 2026 tranches at $350-355B sit below the February Series G mark and read as strategic tranches priced off earlier terms, not as a down round.
Press accounts describe the May round as $65B at a $965B valuation. [43] The funding record splits it into a $50B close plus $15B of April tranches. [230]
Secondary pricing ran well ahead of where the IPO range now sits. In August, Rainmaker Securities CEO Glen Anderson said "the few sellers on our books are around $1.5 trillion." [190] Eric Schmidt holds about 1% from a 2021 Series A investment, per Bloomberg. [231]
LTM EV/Sales on the provider basis, after last week's AI sell-off. None is a pure-play frontier-model developer; no listed pure-play exists.
At the top of the range on outstanding shares ($1.02T), Anthropic would trade at about 12.6x the $81B October run-rate. [10][4] On CNBC's fully diluted ~$1.405T it is about 17.3x. [2] That brackets NVIDIA at 14.0x [234] and Microsoft at 11.7x [236], and sits far below Palantir at 54.5x [232].
Two adjustments cut the other way. The denominator is gross: Anthropic books cloud-marketplace sales as principal, and those were about 48% of H1 revenue. [120][121] The numerator overstates enterprise value: pro forma cash is roughly $216B after the ~$84.6B of net proceeds is added to $131.8B at 30 September, which lowers EV at the top of the range to roughly $800B on outstanding shares. [13][18]
Multiples are derived from the anchors, run-rate revenue of ~$81B at 31 Oct [4], Q3 revenue of $16,956M annualized [9] and NVIDIA's 14.0x [234]. Equity values on outstanding shares use the 13,086,672,593 shares after the offering and placement; fully diluted values include options, RSUs and Amazon rollover-note shares. [42]
The anchors are now the range itself: $916B at $70, $968B at $74 and $1,021B at $78 on outstanding shares, ordered bear to bull. Each sets a revenue requirement, not a fair value. At NVIDIA's 14.0x [234], even the top of the range needs only about $73B of run-rate revenue, below the ~$81B Anthropic reported at 31 October. [4] On CNBC's fully diluted ~$1.405T the requirement rises to about $100B, which is still below the $110B investors expected by year-end. [2][19]
The margin test is the harder one. Damodaran argued $1.5T hinges on 30% margins and has since estimated equity value at about $1.48T. [204][205] The range sits below both of those marks. The latest quarter's adjusted operating margin was 15% and the nine-month margin 12%. [103][20] The range therefore prices Anthropic closer to its current revenue than to the margin expansion his framework required at $1.5T.
Prediction markets had priced richer outcomes. In late October Polymarket put the chance of a first-day close above $2T at about 42%. [244] A first-day close above the ~$1.405T fully diluted top of the range now requires a pop of roughly 42% from $78 to reach $2T on the same basis.
The path is a dated IPO, priced into a weaker tape. Signal, implication and source:
- IPO track (live). The roadshow launched on 10 November and pricing is expected the following week. [36] The Nasdaq Global Select listing is under ANT. [245]
- Strategic anchoring. Seven suppliers and partners are taking $26.0B at the IPO price, with Google and NVIDIA at ~$10.0B each. [12][222] That book is not subject to underwriter lock-ups. [33]
- Down-round risk on the basic count. Pricing at $70 implies ~$916B on outstanding shares, below the $965B Series H. [42][43]
- Post-listing supply. Amazon's rollover notes convert quarterly at the lowest of their fixed price, the IPO price and the 10-day average. [32] Broadcom's notes can convert at 90% of the 30-day VWAP. [198]
- Distress (no signal). The funding record shows no primary down round before the IPO and a $42B October debt financing. [230] Pro forma cash is roughly $216B. [13][18]
The gate that would flip the read is pricing below $74, or upsizing above $78, next week. OpenAI has ruled out its own 2026 debut, so Anthropic is the AI-lab test for public demand this year. [246]
- FY2026E revenue
- $57.7B
- +1,159% YoY
- Q4 2026E $24.5B
- FY2027E revenue
- $257.6B
- +346% YoY
- Up from $242.5B in the first model
- FY2027E adj. EBIT margin
- 29.5%
- Peak year of the forecast
- FY2028E adj. EBIT margin
- 18.9%
- -10.6pp YoY
- Compute spend doubles to $171.7B
- FY2030E revenue
- $850.2B
- Adj. EBIT margin 21.9%
- FY2030E cash
- $445.1B
- From $95.6B at 30 Sep 2026
All figures on this tab are Canalyst model output unless a citation is shown. The model was updated on 2 Nov 2026 to Q3 2026 actuals from the amended S-1; no consensus is loaded yet.
Canalyst model. FY2024-FY2025 are actuals; FY2026 onward are forecasts.
The model forecasts compute spend as a fixed ratio of revenue. Actual Q3 2026 came in at 1.72x, down from 2.08x in Q2, as short-term compute raised costs. The model assumes a return to 2.08x in Q4 2026, then 3.0x through 2027 and 2.2x for FY2028-FY2030.
The 2028 step-down is the entire FY2028 margin dip. Compute moves from 33% of revenue in FY2027 to 45% in FY2028. Adjusted EBIT margin falls from 29.5% to 18.9%, and net income falls 29% from $53.9B to $38.1B while revenue still grows 47%.
Q3 is a warning on the near-term assumption. The amended S-1 shows compute at 58% of revenue in Q3, against the model's 48% implied for Q4. [27] The 2028 timing lines up with the filing's capacity calendar: Google's additional TPU capacity is expected between 2028 and 2031. [124]
Canalyst model output. Costs are grouped as compute and infrastructure, S&M and G&A (personnel included), which differs from the S-1's cost of revenue / R&D / S&M / G&A presentation. Quarterly figures use the model's in-period revenue line. In this update, Q3 2027E revenue ($88,767M) sits slightly above Q4 2027E ($88,350M), which looks like a forecast-input inconsistency rather than an intended decline.
The updated model ties to the amended filing. Q3 2026 revenue of $16,956M and the $567M operating loss match the S-1/A. [9][100] The model's Q3 adjusted EBIT margin of 14.5% matches the 15% filed. [103]
The first-look model had forecast Q3 revenue of $17,031M and EBIT of +$478M. Revenue landed 0.4% below that, and EBIT missed by about $1.0B, because $5.0B of short-term compute and $2.0B of donation charges were not in the forecast. [5] Despite the Q3 miss, the update raised FY2027E revenue by 6% to $257.6B.
The model's Q4 2026E revenue of $24,502M annualizes to about $98B. Anthropic's own run-rate revenue was already about $81B at 31 October, up $10B in a month. [4] Another ~$17B of run-rate growth over November and December would get to the model's Q4 exit level. That is below the $110B investors expected by year-end, per The Wall Street Journal. [19]
The model needs Q4 to grow about 44.5% QoQ, close to Q3's 46.8%. [17] Compute must also fall back to 48% of revenue from 58% in Q3. [27] Those two assumptions return Q4 to an $824M EBIT profit.
October's run-rate jump supports the revenue half of that assumption. The margin half is untested until Q4 shows whether delivered SpaceXAI capacity replaced short-term compute. [26]
Cumulative modelled compute spend over FY2026E-FY2030E is about $932B. That is roughly double the $447.5B of committed spend the amended S-1 expects before the end of 2030. [131] It is also above the full $895.4B aggregate, most of which runs to 2045. [131]
On the model's path, contracted capacity falls short of forecast usage, so more capacity has to be signed. Q3 shows what happens when that capacity is bought short-term instead. [5] If revenue undershoots the model, the risk inverts: the fixed Google, AWS and Microsoft minimums become the binding cost. [122][125][126]
On the model's numbers the range is inexpensive on forward revenue. The top of the range on outstanding shares is 4.0x FY2027E revenue, and 5.5x fully diluted. NVIDIA trades at 14.0x [234] LTM. Run-rate revenue of ~$81B at 31 October already exceeds the ~$73B the top of the range needs at NVIDIA's multiple. [4]
The earnings multiples show why the market is not paying for the model. At $78 fully diluted the stock would be 26.0x FY2027E net income but 36.8x FY2028E, as modelled compute intensity rises and earnings fall 29%. A buyer at the top of the range is underwriting three things: revenue rising about 4.5x in FY2027, compute efficiency improving to 3.0x from Q3's 1.72x, and the FY2028 efficiency loss being temporary.
Damodaran's 30% margin test is met only around the model's FY2027 peak: 29.5% adjusted for the year and 33.0% in Q3 2027E, before stock-based compensation. [204] On reported EBIT, the model never exceeds 23.0% in any period shown. The gap is about $25.8B of modelled stock-based compensation in FY2027.
- Growth is audited through Q3. Q3 2026 revenue was $16,956M, up 46.8% QoQ and 1,127% YoY. [17] Nine-month revenue was $33,242M. [15]
- Profitability went backwards in Q3. A $567M operating loss followed Q2's $974M profit, driven by $5.0B of short-term compute and $2.0B of donation charges. [5]
- Cash generation held. Nine-month operating cash flow was $6,569M, and cash and short-term investments were $131.8B at 30 September. [22][18]
- The commitment stack is about $900.9B. [6] At 30 September, $447.5B is expected before 2030 and $69.8B within 12 months; $213.8B is conditional. [131][132]
- The headline loss is still the cap table. $94.3B of the $104.0B nine-month net loss is a fair-value charge. [106][109]
- Concentration is moderate. The top five customers were about 25% of nine-month revenue. [219]
- Control stays inside. The Founder LLC holds 50.1% of votes and the Long-Term Benefit Trust elects a board majority. [34][247]
- The range reset the valuation. $70-$78 implies up to $1.02T on outstanding shares and ~$1.405T fully diluted, against earlier talk of $1.5-2T. [10][2][11] The low end on outstanding shares is below the $965B Series H. [43]
- Demand is anchored by suppliers. Seven partners, led by Google and NVIDIA at ~$10.0B each, take $26.0B at the IPO price. [222][12]
- Growth re-accelerated into Q4. Run-rate revenue was ~$81B at 31 October, up $10B in a month. [4]