2026 ANTHROPIC
IPO_$65B_
SERIES_H_
$965B_VAL.
2026 — inflection point для Anthropic: бывшие core-исследователи OpenAI, миссия «responsible AI development». За три недели — крупнейший VC round в истории ($65 млрд Series H), confidential S-1 в SEC и три Wall Street bookrunner'а. Hardcore tech-разбор: timeline, Series H cap table, confidential S-1 wire protocol, bookrunners, ARR $10→$47 млрд, enterprise share > OpenAI, IPO race, company profile, risk matrix, FAQ, 5-step playbook. Data cutoff 25.06.2026 — не investment advice. Для инженеров: IPO funding → compute capex → API pricing — transmission lag 2–4 квартала; hedge через Metal/MLX обязателен.
1. Pain points: почему Anthropic IPO — системный signal
1) Valuation narrative: $965B private — намеренно под trillion threshold; analysts IPO range $1.1–1.25T. 2) Enterprise flip: Ramp AI Index (июнь 2026) — Anthropic впервые #1 по US enterprise adoption (41 %) и API spend (40 %). 3) Capital crowding: Anthropic + OpenAI + SpaceX ≈ $5T potential mcap — liquidity drain для других tech IPO 2026. 4) Dev impact: Claude Code = 4 % public GitHub commits; 80 % internal production code at Anthropic written by Claude — post-IPO compute capex перепишет API economics и token throughput pricing.
2. Timeline: core events
| Дата | Событие |
|---|---|
| 12.02.2026 | Series G closed: $30 млрд, valuation $380 млрд |
| Апрель 2026 | Amazon $5 млрд strategic commit; ARR > $30 млрд |
| 28.05.2026 | Series H: $65 млрд, post-money $965 млрд |
| 01.06.2026 | Confidential S-1 filed with SEC |
| 03.06.2026 | Morgan Stanley, Goldman Sachs, JPMorgan — lead bookrunners |
| Октябрь 2026 (expected) | Earliest listing window (Nasdaq/NYSE) |
S-1 через 4 дня после Series H close — orchestrated pipeline, не coincidence.
3. Series H: record round breakdown
Size: $65 млрд — largest single VC round ever.
3.1 Lead investors
Altimeter Capital · Dragoneer Investment Group · Greenoaks Capital · Sequoia Capital
3.2 Co-leads
Capital Group · Coatue Management · D1 Capital Partners · GIC · ICONIQ Growth · XN
3.3 Notable follow-ons
Blackstone · Baillie Gifford · Brookfield · D.E. Shaw Ventures · DST Global · Fidelity · General Catalyst · Jane Street · Temasek · T. Rowe Price · Insight Partners · Lightspeed
3.4 Strategic / industrial
Amazon: $5 млрд (prior commit, in round) · Micron · Samsung · SK Hynix — top-3 memory vendors lock supply chain for HBM/DDR5 bandwidth pipeline.
3.5 Use of proceeds
Official: ① AI safety + interpretability R&D; ② compute infra scale (Amazon 5 GW, Google+Broadcom 5 GW TPU, SpaceX Colossus 1/2 GPU capacity); ③ Claude enterprise products + partner ecosystem.
4. IPO process: technical decode
4.1 Confidential S-1 (JOBS Act)
Eligible EGCs file S-1 draft confidentially — iterate with SEC before public disclosure. Formal prospectus ≥ 15 days before roadshow.
Anthropic (01.06.2026): «Today, Anthropic, PBC confidentially submitted a draft registration statement on Form S-1 to the U.S. Securities and Exchange Commission for a proposed initial public offering of our common stock. This gives us the option to go public after the SEC completes its review.»
Key: confidential filing ≠ mandatory IPO; timing, price, size TBD; listing = market + regulatory function.
4.2 Bookrunners
| Institution | Role |
|---|---|
| Morgan Stanley | Lead Left |
| Goldman Sachs | Lead Bookrunner |
| JPMorgan Chase | Lead Bookrunner |
| Wilson Sonsini | IPO counsel (Google 2004) |
4.3 Listing window & target valuation
Earliest: October 2026; conservative Q4 2026. SpaceX ref (Apr 1 confidential → May 20 public → June listing): Anthropic public S-1 expected Jul–Aug. Private val $965B; analyst IPO range $1.1–1.25T. At $47B ARR ≈ 20× P/S.
5. Financials: ARR curve
| Timestamp | ARR |
|---|---|
| Early 2025 | ~$10B |
| End 2025 | ~$90B |
| Feb 2026 (Series G) | ~$140B |
| April 2026 | ~$300B |
| May 2026 (Series H) | ~$470B |
16 months: $10B → $470B — 47×. Salesforce ~decade to $10B revenue. Feb–May 2026: ~$8B ARR/month net add. Driver: Claude Code — 4 % global GitHub commits (doubling monthly), 80 % internal prod code.
Profitability: operational profit expected Q2 2026 — vs OpenAI «high revenue + high burn» narrative.
6. Market share: enterprise #1 over OpenAI
| Metric (Jun 2026) | Anthropic | OpenAI | |
|---|---|---|---|
| US enterprise AI adoption | 41 % | 32.3 % | — |
| Enterprise API spend share | 40 % | 27 % | 21 % |
| Claude Code / GitHub commits | 4 % (global) | — | — |
Source: Ramp AI Index (June 2026). OpenAI wins consumer brand; enterprise trust + code gen = Anthropic moat.
7. OpenAI competition & AI IPO race
| Dimension | Anthropic | OpenAI |
|---|---|---|
| Private valuation | $965B | $852B |
| Latest round | $65B (Series H, May 2026) | $122B (Mar 2026) |
| ARR | ~$470B | ~$360B (est.) |
| IPO status | Confidential S-1 (Jun 2026) | Planned Sep 2026 |
| Enterprise | API spend #1 (40 %) | #2 (27 %) |
| Company | IPO status | Valuation | ARR |
|---|---|---|---|
| Anthropic | S-1 filed (Jun 1) | $965B | ~$470B |
| OpenAI | Preparing | $852B | ~$360B |
| SpaceX | Roadshow (Jun 2026) | $1.75T | — |
Sam Altman (CNBC): «OpenAI will go public when we think it's right. We're not focused on deciding the exact timing now.»
8. Company profile
Founded 2021 · HQ San Francisco · CEO Dario Amodei (ex OpenAI VP Research) · President Daniela Amodei · Entity PBC · Products Claude 3.5/4/Opus 4.8, Claude Code · Customers global enterprise (finance, healthcare, cybersecurity).
9. 5-step playbook: devs & investors
Step 1: map Claude API / Claude Code dependency by token volume + compliance tier.
Step 2: multi-vendor fallback (OpenAI / Gemini / local MLX) — avoid single-vendor price shock post-IPO.
Step 3: parse public S-1 (Jul–Aug) for net revenue, compute capex, export control risks (Fable 5).
Step 4: evaluate pre-IPO indirect exposure (DXYZ) — liquidity + barrier, not listed equity.
Step 5: benchmark Claude Code workloads on MLX 70B Q4 via Metal — establish tokens/sec baseline before API repricing.
10. Risk matrix
1) Market: IPO slip to 2027. 2) Regulation: US Fable 5/Mythos 5 suspension — S-1 risk disclosure. 3) Price war: OpenAI cuts → ARR growth compression. 4) Valuation: 20× P/S fragile if growth decelerates. 5) Compute capex: $65B may not cover long-term GPU/TPU roadmap. 6) Competition: Gemini, Meta AI, xAI. 7) Throughput risk: API rate limits + latency under load — local Metal hedge mandatory.
11. Case study: Claude Code burst & Mac compute routing
15-engineer team: Claude API Jan 2025 ~$8K/mo; post Claude Code rollout May 2026 ~$42K/mo — correlated with Anthropic enterprise ARR curve. Post-IPO compute capex → API pricing in 2–4 quarters. Routing: Cursor + Claude Code cloud flagship; batch refactor + unit test gen → MLX 70B Q4 on M4 Max 128GB unified memory (~35 tok/s sustained on Metal); 7×24 CI agent → remote Mac node (thermal headroom, no local throttling). $470B ARR proves code agent = fastest-growing enterprise software category — don't wait for post-IPO price cuts; establish local inference baseline now.
12. Investor FAQ
Q: Can I buy Anthropic stock now?
A: Not listed. Pre-IPO via Forge Global, Hiive, EquityZen or indirect DXYZ (Destiny Tech100).
Q: When IPO?
A: Earliest Oct 2026; SEC review 3–4 months; Q4 2026 or early 2027 possible.
Q: Nasdaq or NYSE?
A: Unconfirmed; tech IPOs typically Nasdaq.
Q: Profitable?
A: Operational profit expected Q2 2026; prior losses on compute/R&D.
Q: ARR vs net revenue?
A: ARR = monthly ×12; net lower after discounts, refunds, cloud cost allocation — exact in public S-1.
Q: Anthropic vs OpenAI valuation?
A: May 2026 Series H: Anthropic $965B > OpenAI $852B.
Q: Origin & founders?
A: US company, 2021, Dario & Daniela Amodei et al. (ex OpenAI researchers).
13. Closing: pre-IPO, local Metal throughput = best Claude workflow hedge
Anthropic trillion sprint = expanding cloud Claude capacity — but IPO proceeds, compute buildout and developer API bills lag by quarters. Pure cloud Windows/Linux can hit Claude API, but Cursor/Xcode toolchain parallelism, Claude Code-class local agents, launchd 7×24 daemons run smoother on Apple Silicon with Metal unified memory. If IPO narrative and API cost volatility affect your stack, MACGPU remote Mac nodes: 70B quant on unified memory, Cursor/LiteLLM dual-stack — controlled compute before Nasdaq bell.