The following is FutureSearch's comprehensive forecast for Anthropic's financial and technological developments through May 2027. My central case has ARR reaching $93B by May 2027 (p10 $58B, p90 $145B), based on a Claude Code revenue of $20B, and the IPO pricing at a median of December 20, 2026 with about 88% probability of clearing the May 21, 2027 deadline.
The financial forecast distributions anchor most heavily to information surrounding the Series H close at $965B post-money (Crunchbase), the Opus 4.8 release (TechCrunch) and Anthropic's plan for Mythos, and the June 1 confidential S-1 filing (Anthropic) that starts the procedural clock toward a public listing, a clock OpenAI joined one week later with its own June 8 filing. For the head-to-head IPO race, see our updated Anthropic and OpenAI IPO dates and valuations.
In a sentence, at $965B Series H pricing, Anthropic is trading at 21x current ARR and 10x my median 2027 ARR. I predict the public market will price Anthropic at or above the Series H mark. My median for the 90-day-post-IPO market cap is $1.05 trillion.
Anthropic disclosed $47 billion in run-rate revenue as of mid-May 2026, per the May 28 Series H announcement (Anthropic). That is the anchor for the trajectory shown above. CFO Krishna Rao's number extends Dario Amodei's earlier disclosure walk ($9B Dec 2025, $14B Feb, $19B March, $30B April, $44B early May) and lines up with both SemiAnalysis reporting and the Q2 2026 GAAP projection of $10.9B in quarterly revenue Anthropic shared with Series H investors (CNBC). The company is adding about $96M in annualized revenue per day.
The p50 of $93B by May 2027 assumes that pace decelerates in absolute terms but keeps compounding. Three structural drivers support it: enterprise concentration (1,000+ customers spending $1M+ annually, up from 500 in February per Sacra), the coding moat (54% of coding-specific LLM spend per Menlo), and cross-cloud distribution that no competitor has matched.
Claude Code is the live ammunition. It went from $500M ARR in September 2025 to $8B in May 2026, and the p50 forecast of $20B by May 2027 assumes the doubling-every-six-weeks cadence breaks down to roughly 2.5x annual. The case for the right tail at $45B is that agentic workloads consume 10 to 100 times more compute per developer-day than chat, and Anthropic owns the coding category. Opus 4.8's launch theme reinforces that case. It centers on Dynamic Workflows that orchestrate hundreds of parallel subagents and codebase-scale autonomous migrations described as running "from kickoff to merge" across hundreds of thousands of lines.
The case for the left tail at $7.5B is that Cursor launched its own model in November 2025 and hit $2B ARR with 60% enterprise mix (MLQ). I think the Cursor disintermediation story is overpriced. Custom coding models lag Claude by six months on real workloads, and enterprise procurement loops back to the best model on the next renewal cycle. Cursor will keep buying Claude API at scale even as it markets its own model.
The single biggest tail risk on the revenue line is the gross-revenue accounting question. Anthropic books cloud-reseller end-customer spend on a gross basis (Sacra), which means full customer spend hits revenue and partner payouts hit expenses. A forced restatement to net during IPO due diligence would cut headline ARR by 20 to 40% in one print.
I think this restatement risk is real but overstated by skeptics. Stripe books gross. Shopify books gross. The cloud-reseller relationship is the genuine customer-of-record arrangement, not a flow-through. The auditor risk is non-zero but I'd put it at single digits, not the implicit 20-30% the public-market $400-500B targets imply.
The chart above tracks the Claude Code Max 20x weekly Sonnet hour cap, the most-watched consumer proxy for whether Anthropic has compute headroom. The line stepped from 240 hours in 2024 to 360 hours through most of 2025, then jumped three times in five weeks in early May (Medium), all following the SpaceX/xAI deal that lifted Opus API limits 1500%+ (Anthropic). That looked like the constraint breaking.
The Series H announcement complicates the picture. Anthropic "has struggled to meet demand in recent months, forcing it to institute usage limits during peak hours and incentivize off-peak use" (Reuters). The rationing is not resolved. It is being managed.
Two compute-proxy forecasts tell a coherent story. Flagship output pricing for the next major Anthropic release centers at $25 per million tokens, identical to four consecutive Opus generations (4.5, 4.6, 4.7, and today's 4.8). The p10 is $15 and the p90 is $40. Anthropic's revealed preference is to hold pricing flat even as GPT-5.5 moved to $30 and Gemini 3.1 Pro moved to $12. That is a company protecting share without resorting to price rationing.
The asymmetric right tail at $40 captures the Mythos pricing case. The system card (Anthropic) reports Mythos itself at $125 per million for Project Glasswing partners, which gives Anthropic an anchor far above any current GA price. With Mythos GA committed "in the coming weeks" per the Opus 4.8 message (Tom's Guide), that anchor pricing is about to face market discovery.
The consumer rate-limit forecast sits at a p50 of 500 hours. The p10 is the current 360 hours; the p90 is 2000. The 500-hour median reflects that part of the May spike was off-peak modulation rather than durable expansion, and Reuters's peak-hour rationing language confirms it.
Anthropic announced fresh compute commitments alongside the Series H: AWS expansion of up to 5GW of additional capacity, Google/Broadcom 5GW of TPUs, and SpaceX access to Colossus 1 and 2. Strategic memory and storage suppliers (Micron, Samsung, SK hynix) participated in the equity round, which is unusual and signals long-term supply preference. Capacity is arriving. It is being absorbed by demand faster than the May 2026 evidence alone suggested.
The chart above plots SWE-bench Verified scores for each Claude release, with the restricted Mythos Preview marked as a separate diamond at 93.9%. The capability roadmap split on May 28. Alongside the Opus 4.8 release, Anthropic stated that it expects "to be able to bring Mythos-class models to all our customers in the coming weeks" once cyber safeguards are complete (Axios). That is a categorical change from the April system card framing, which described Mythos's cyber capabilities as emergent from general training and not separable.
Anthropic now positions Mythos-class as a tier above the Opus 4.x line and not as "Claude 5." That bifurcates the capability roadmap into incremental Opus 4.x releases and a Mythos-class premium tier that becomes the genuine frontier product.
Mythos-class GA before May 21, 2027 sits at 75-80% probability. The "coming weeks" commitment is hard to walk back without a fresh safety incident or a regulator action.
The mathematical tension between the Mythos GA probability and the SWE-bench Verified distribution largely dissolves under the new tier structure. The SWE-bench distribution does not need to constrain Mythos GA timing because Mythos is not being released as the headline Claude 5. It is a separate premium tier. The forecast for the next GA Opus flagship's SWE-bench Verified score moves to a 93% median (p10 87.6%, p90 96.5%) on the assumption that Mythos-class scoring gets reported within the window.
The Opus 4.8 release reinforces a less-discussed point about Anthropic's release cadence. The 41-day turnaround from Opus 4.7 is the fastest flagship gap in Anthropic's history. TechCrunch attributes part of that to a chilly reception for Opus 4.7 and major releases from OpenAI Codex and Google Gemini 3.5 Flash.
The Sonnet and Haiku tiers are 3 and 7 months stale, which suggests Anthropic is concentrating resources on the Opus flagship cadence and the Mythos track while letting mid- and low-tier tiers ride. The "Claude 5" question becomes less meaningful in the process. The next generational capability jump may ship as a Mythos-class GA in Q3 2026 rather than as a model branded "Claude 5."
The ASL-4-equivalent forecast moves up to about 40%. The "coming weeks" GA of a model whose system card described emergent zero-day discovery, sandbox escapes, and "less confidence than for any prior model" on the R&D-4 threshold means Anthropic is about to deploy its most capability-flagged model widely. Broad deployment forces a fresh public risk determination, and that is exactly the surface where a formal threshold crossing would land.
Anthropic filed a confidential draft S-1 with the SEC on June 1, 2026 (Anthropic). That is the single most concrete IPO signal short of a public roadshow, and it resolves the open question from the Series H announcement (whether the round was a substitute for an IPO or a bridge to one) in favor of "bridge." The chart above shows the resulting probability distribution. The procedural clock now runs from a known date.
The Series H close at $965B post-money is the highest private valuation any AI company has carried and $113B above OpenAI's most recent $852B mark. Multiple outlets and investors familiar with the company framed it as the pre-IPO bridge round.
From Reuters: "Anthropic's pursuit of private funding coincides with preparations for a public listing, according to investors and bankers familiar with the company. Both Anthropic and OpenAI are planning to tap the public market, possibly as quickly as this year." (TechCrunch) describes the round as "what could be the AI startup's final private fundraise before debuting on the public markets". (Fortune) calls it the "race to own the technology of the century".
The investor list is the strongest piece of evidence. Capital Group co-led the round, with Fidelity and T. Rowe Price participating: mutual funds that take pre-IPO allocations to lock in IPO entry. Mutual funds do not pay $965B for long-term private positions. They pay to be first in line on the public listing. That makes the Series H the pre-IPO round, not a substitute for it.
The procedural calendar math pulls the distribution earlier and tightens it. Confidential SEC review for a company of Anthropic's scale runs 3-6 months, then a public S-1 flip, a 15-day cooling period, a one- to two-week roadshow, and pricing. SpaceX moved from confidential filing on April 1, 2026 to a public S-1 on May 20 to a target IPO around June 12, about eleven weeks total. Airbnb ran four months end to end. Klaviyo and ARM both ran three to four months. Reddit ran nine months on multiple comment rounds.
Anthropic carries specific friction factors that argue against the SpaceX pace. Gross-vs-net accounting (Anthropic counts cloud-reseller passthrough as revenue) is the question SEC reviewers will spend the most time on. The Department of War litigation requires material risk-factor disclosure (CNN). AI capability and safety risk factors are novel disclosure territory with no settled template. Customer concentration (historical Cursor and GitHub Copilot dependence) requires careful treatment. Together these point to four to six months of review rather than eleven weeks.
The percentiles update accordingly. The p10 IPO date is September 28, 2026 (the SpaceX-clone fast track). The p25 is November 10, 2026. The p50 is December 20, 2026, well inside the forecast window. The p75 is February 28, 2027 (the realistic eight-month review timeline given Anthropic's complexity factors). The p90 holds at May 21, 2027. The probability of an IPO before May 21, 2027 rises to about 88%, up from 60-65% post-Series H and from about 30% before the round closed.
The right tail (about 12% no-IPO mass) covers four scenarios: SEC issues on accounting that force restatement or material delay; equity market dislocation closing the window; a material adverse change from the DoW litigation; or Anthropic withdrawing if it judges the public market pricing unattractive against the $965B private mark. None of these are likely individually. Together they retain meaningful probability.
The more important point is on valuation. The public-market $400-500B target, the consensus signal in March-April from investors who skipped the round, assumed the IPO would be deferred. With the IPO now in the window and the Series H closing at $965B with mutual-fund participation, the public-market mark is more likely to clear the Series H mark than to fall below it.
The Fidelity, T. Rowe Price, and Capital Group participation is a price-anchor commitment for the IPO. They are not paying $965B in May to short the public listing in March. They are paying $965B for early access to a listing they expect to print above that mark. The forecast median for the 90-day-post-IPO market cap is $1.05 trillion, with the p10 at $750B and the p90 at $1.6 trillion. The Series H buyers are pricing the median outcome, not chasing the bull case.
The Department of War (renamed from Defense per Executive Order 14347 on September 5, 2025) designated Anthropic a supply chain risk under both 10 U.S.C. § 3252 and 41 U.S.C. § 4713 (FASCSA) on March 3-4, 2026. The chart above tracks the two parallel legal tracks Anthropic filed in response. This is the cleanest example of a story that matters for the brand but stays immaterial to the revenue thesis.
The litigation status: Judge Rita Lin (NDCA) granted Anthropic a preliminary injunction on March 26 (CNN), the D.C. Circuit denied an emergency stay on April 8, and oral arguments happened May 19 (Axios). Judge Henderson called the designation "spectacular overreach." Judge Katsas pushed back on Anthropic's evolving usage policies. The ruling is under advisement.
The forecast puts the probability of a $100M+ DoD or IC prime contract by May 2027 at 3%, and the probability of supply chain designations being formally vacated at 9%. Both numbers are calendar bets, not capability bets. Even a panel ruling for Anthropic faces en banc and SCOTUS review.
The DoW awarded $600M in classified AI contracts to eight competitors in May 2026, excluding Anthropic (CNBC). Alternatives are entrenched. The federal track stays frozen.
This does not move the revenue thesis because Anthropic is on pace to do $93B ARR by May 2027 without federal procurement. A $100M contract would round to noise. The relevant federal scenario is tacit non-enforcement, where the White House lets agencies buy Claude through subcontractors and resellers while leaving the formal designation in place. The probability of that path is higher, around 35-40%.
Anthropic's parallel hedge of UK government courtship with London HQ expansion and dual stock listing offers (FT) is the right strategic response. None of this changes the revenue base by more than a few percent in either direction. One second-order point worth flagging: Mythos-class GA with cyber capabilities going to "all customers" sharpens the dual-use cyber narrative the Pentagon designation leans on, which could complicate the political track even as the legal track moves toward resolution.
The three scenarios collapse to decoupling at 48%, convergence at 42%, containment at 10%.
Decoupling is the base case where Anthropic's commercial trajectory runs independently of the federal standoff. ARR lands $85-105B by May 2027. The Opus 4.x cadence continues with Mythos-class shipping as a separate premium tier in Q3 2026. The IPO prices in Q1-Q2 2027 at $1T to $1.2T. DoW designations remain on the books while enforcement softens.
Convergence (42%) is the capability and capacity bull case. Mythos GA validates the moat, ARR overshoots to $110-140B, the IPO prices above $1.4T with the S-1 calendar holding the SpaceX-pace timeline, and the federal situation resolves through executive deal or D.C. Circuit ruling for Anthropic.
Containment (10%) is the bear case. Mythos GA runs into safety issues delaying broader deployment, compute re-tightening evidence accumulates beyond the Reuters peak-hour reference, gross-revenue accounting forces a restatement during S-1 review (or Anthropic withdraws the filing), and the D.C. Circuit affirms § 4713.
Six predictions I'd take the over on, given those scenarios.
Anthropic prices its IPO at above $1T between November 2026 and February 2027. The confidential S-1 went in on June 1, the Series H mutual-fund anchor is the price floor, and OpenAI's parallel S-1 filing one week later creates competitive pressure. Probability: 65%.
The next Opus release prices at $25 output, holding the four-release anchor. Mythos-class GA prices at $40 to $60 as a premium tier, well above the current $25 anchor but well below the $125 partner figure. Probability: 60% for the Opus tier holding $25.
Mythos-class GA ships before December 2026. The "coming weeks" timeline plus the public Anthropic commitment make this hard to walk back. The only thing that delays it is a fresh safety incident or a regulator action. Probability: 70%.
ARR ends 2026 between $75B and $90B, and the May 2027 print lands at or above $100B. The May 2026 starting point at $47B plus the Opus 4.8 agentic expansion plus Mythos GA stacking into the second half of 2026 supports the high end. Probability: 50%.
The IPO happens before May 21, 2027 at a first-day market cap above $1T. The S-1 procedural calendar, the Series H pricing floor, and the OpenAI competitive pressure all align. Probability: 80%.
Cursor's in-house coding model loses share back to Claude by Q1 2027. The six-month lag between frontier capability and downstream coding models is the structural constraint Cursor cannot beat. By Q1 2027, Mythos-class is GA and the Opus line is on 4.9 or later; Cursor's model is a major version behind. Cursor's Claude API spend rises in absolute terms even as its in-house model marketing continues. Probability: 60%.
Pulling all of this together, at 965B Series H pricing Anthropic is trading at 21x current ARR and 10x my median 2027 ARR. The consensus public-market read of \400-500B is anchored to four priors: a stale view of compute as a binding constraint, an assumption that the IPO would discipline a high valuation downward, an implicit assumption that Mythos stays restricted, and a treatment of gross-revenue accounting as likely to be restated.
The first three priors are wrong as of June 2026. The Series H close at $965B, the Opus 4.8 release, and the June 1 confidential S-1 filing compound the case. The S-1 filing turns the IPO from a calendar bet into a procedural one with a known start date and a median resolution in late December 2026. The Series H buyers, the mutual funds whose participation is the cleanest pre-IPO signal available, are pricing the public listing at or above $965B. So am I. I'd buy Anthropic at $965B if I could.
Run this forecast yourself by connecting FutureSearch to Claude. Then ask Claude to refresh the numbers any time the news cycle moves.