Anthropic Is in Talks to Raise $30B–$50B at a $950B Valuation — and That Number Edges Past OpenAI

Anthropic Is One Term Sheet Away From Being Bigger Than OpenAI

Here's the deal: on May 12, Bloomberg and The New York Times simultaneously reported that Anthropic is in early talks for a $30–50B funding round at a valuation as high as $950B. Term sheet is not yet signed and the final price may settle lower — but $950B is the first number that would put Anthropic ahead of OpenAI's $825B. For the first time in four years, the #1 spot in AI labs is up for grabs.

The real story is the revenue curve. ARR went from ~$9B at the end of 2025 to $30B by April 7 (Dario Amodei publicly confirmed that number himself), and per the May Bloomberg piece, $44B in May. A 4.9× jump in five months. Inference gross margins improved from 38% a year ago to 70%+. Both top line and unit economics are accelerating at the same time.

For comparison, OpenAI reported ARR of $20B+ in Q1 2026. The gap had been wide. But if Anthropic really printed $44B in May, OpenAI's revenue lead is gone. ChatGPT still wins on user counts (~800M weekly vs. Claude's estimated 100-150M), but on revenue, Anthropic's heavy focus on enterprise and API channels has flipped the script.

The timing is deliberate. Same week as the Bloomberg story: PwC alliance expansion (May 14), Gates Foundation partnership (May 14), Claude for Small Business launch (May 13). The whole package reads like a deck slide for LP meetings: "We're advancing on revenue + enterprise + government + social impact, all at once." A momentum kit to show prospective investors before the term sheet is signed.

The Players — Anthropic, OpenAI, Lab Revenue Curves, and the LPs

Anthropic. Founded 2021 by ex-OpenAI siblings Dario and Daniela Amodei. Constitutional AI + safety-first. ~1,500 employees. ARR ~$44B as of May. Revenue mix — roughly (a) API/direct ~50%, (b) AWS Bedrock integration ~25%, (c) enterprise direct ~25%. Claude Opus 4.7 is the flagship; Cowork and Claude Code are the agentic-workflow SKUs.

OpenAI. Founded 2015. ~4,500 employees. ARR $20-25B as of April-May. Same May 11 launched the $4B Deployment Company (DeployCo) to accelerate enterprise penetration. May 14-15, "Codex for (almost) everything" — mobile app + 90+ plugins + GPT-5.2-Codex. On the funding side, the SoftBank-led round priced the company at $825B. User counts: 800M+ weekly on ChatGPT, the global #1.

Dario Amodei (Anthropic CEO). In an April 7 media round he confirmed "$30B ARR; inference gross margins moved from 38% a year ago to 70%+." Also forecast the "five-person, $1B-revenue company" era (referencing Cursor/Anysphere). On May 12 funding-talk reports, no official comment — Bloomberg notes "Anthropic declined to comment."

Claude. May lineup: Opus 4.7 (top, coding/reasoning/vision), Sonnet 4.7 (mid, cost-optimized), Haiku 4.5 (cheap, agent loops). Heavily skewed toward enterprise vs. OpenAI — API + enterprise is the dominant share. Cowork (launched Q1 2026) and Claude Code (2024+) are the growth SKUs.

The LPs (potential round participants). Undisclosed, but market chatter includes SoftBank Vision Fund, Saudi PIF, Norway's NBIM, Abu Dhabi ADIA/MGX, Canada's CPP, U.S. VCs (Sequoia, Lightspeed, Andreessen Horowitz), and possible follow-ons from Google and Amazon. A single LP writing $5B+ solo is unlikely; expect a consortium structure.

Inside the Numbers — Revenue Curve, Margins, Infrastructure

The revenue curve. ARR trajectory: end-2025 ~$9B → April 7, 2026 $30B (per Amodei) → May 2026 $44B (per Bloomberg). That's 4.9× in five months. Quarterly, that implies ~$11B in revenue. Compare with OpenAI Q1 2026 estimated at ~$5B. Anthropic may now lead OpenAI on quarterly revenue too. ChatGPT's 800M weekly users still leave OpenAI dominant on the consumer side, so the picture is becoming "users to OpenAI, revenue to Anthropic."

Inference gross-margin improvement. Amodei on April 7: "Inference gross margins moved from 38% a year ago to 70%+." Two reads. First, Anthropic has succeeded at model-side efficiency improvements. Second, increased share of non-Nvidia accelerators (AWS Trainium2, Google TPU v6) has cut per-token compute costs. 70%+ is SaaS-tier margins — the "GPU burn" concern dogging labs is, at least at Anthropic, dampened.

Infrastructure capacity. May 7 — reports of a 300MW+ compute deal with SpaceX's Memphis Colossus 1 data center. SpaceX is scaling Colossus 1 to 1GW by end-2025, and Anthropic just locked up roughly a third. That puts Anthropic in direct competition for the same site as xAI's Grok lineup. Anthropic is also expanding direct use of AWS Project Rainier (Trainium2 clusters) and Google TPUs.

The management narrative. In April interviews, Amodei predicted the era of "five-person companies with $1B in revenue." His own company's framing: (a) AI revenue is growing faster than we expected, (b) infrastructure capacity is now the bottleneck, (c) we need to lock in talent + compute simultaneously. The pitch for $950B: "We can accelerate ARR past $100B in the next 12-24 months."

Metric	Anthropic (May 2026)	OpenAI (Apr-May 2026)	Lead
ARR	$44B	$20-25B (est.)	Anthropic
Valuation	$950B (in talks)	$825B	Anthropic (if it closes)
Inference gross margin	70%+	undisclosed	—
Weekly consumer users	100-150M (est.)	800M+	OpenAI dominant
Employees	~1,500	~4,500	OpenAI larger
Revenue per employee (annualized)	$29M	$5M	Anthropic dominant
Enterprise + API share	75%+	<50%	Anthropic heavy on enterprise

Term-sheet risk. Bloomberg explicitly notes the term sheet is not yet signed and the price could come down. $950B is the top of the range; final pricing might settle in the $750-900B band. Other variables: (a) LP demand validation, (b) macro froth concerns around AI labs, (c) the May 12-16 chip selloff (Qualcomm -11%, Nvidia and AMD soft), (d) potential Trump-administration AI-lab regulation.

What Each Side Gets

Anthropic's wins. First, symbolism of overtaking OpenAI. The AI lab #1 spot changes hands for the first time in four years. Leading on both revenue and valuation is a strong message for talent, LPs, and revenue acceleration. Second, infrastructure war chest. $30-50B at $950B funds simultaneous expansion across SpaceX Colossus, AWS Project Rainier, and Google TPUs — doubling compute capacity. Third, talent. "Anthropic just passed OpenAI" is a powerful recruiting line for PhDs and senior engineers. Fourth, political capital. The U.S. government, EU, Gates Foundation, PwC — once they see Anthropic as #1, the company gains a louder voice in policy conversations.

LP wins and risks. Wins — if Anthropic really hits $100B in ARR within 12-24 months, an IPO at $2T+ becomes plausible. Doubling money in a year is on the table. Risks — (a) AI revenue growth may decelerate, (b) GPU/power infrastructure bottlenecks, (c) AI froth concerns, (d) Anthropic's model edge could erode within 6-12 months as OpenAI, Google, xAI, and Chinese labs catch up.

OpenAI's losses. The #1 AI lab title may be slipping. Lose the revenue lead and the "OpenAI premium valuation" narrative weakens. OpenAI did counter with the $4B Deployment Company launch on May 11, but Anthropic's May 14 PwC + Gates double-punch took the momentum.

Wins for Google and Amazon. Google was an early Anthropic investor ($3B+ total). Amazon is a deep partner via AWS Bedrock integration and Trainium2 infrastructure ($8B+ total). Both see portfolio uplift from a valuation jump. Amazon especially benefits — Trainium2 revenue is directly contributing to Anthropic's 70%+ margin story.

Nvidia in a tricky spot. Anthropic is increasing AWS Trainium2 and Google TPU share on its compute side. Nvidia's direct compute revenue share from Anthropic is shrinking. The flip side: as Anthropic raises the bar, Recursive Superintelligence, OpenAI, and xAI all expand their compute footprints (much of which is still Nvidia), so Nvidia's overall AI lab revenue is fine.

SoftBank, PIF, and other LP candidates. SoftBank just closed a major OpenAI follow-on. PIF is funding Saudi-domestic labs (Humain etc.). Participation in the Anthropic round shapes lab-funding dynamics for the next 12 months. A single LP backing "Anthropic + OpenAI + Saudi labs" simultaneously is plausible.

Historical Parallels — Wins and Losses

Win: Facebook pre-IPO rounds (2010-2012). Facebook's Goldman-led pre-IPO round priced at $50B; the IPO closed at $104B, a 2× pop. Anthropic at $950B is similar in structure — a premium pre-IPO mark. But Anthropic's IPO timing is undecided, more likely 2027-2028.

Win: WhatsApp acquisition (2014). Facebook bought WhatsApp for $19B; the market called it expensive. Ten years later WhatsApp is estimated at $100B+. Anthropic at $950B may look like a steal looking back from 2031.

Partial loss: WeWork (2018-2019). SoftBank invested $10B+ in WeWork at a $47B mark. A year later, the IPO failed and the valuation crashed to $8B. Failure causes — (a) the revenue curve didn't match the pitch, (b) negative margins, (c) governance issues. Anthropic differs decisively on all three. The comparison is just a reminder that "froth concerns" exist.

Partial loss: Sequoia's FTX investment (2021-2022). Sequoia invested at a $32B mark; one year later, zero. Big rounds = big risk. FTX was a fraud/governance case; Anthropic is the opposite on both. But "LPs always do due diligence" applies regardless.

Competitor Counter-Plays

OpenAI. Likely counters: (a) a new round at $1T+, (b) accelerate monetization of ChatGPT's 800M free users (ads, e-commerce, subscriptions), (c) accelerate GPT-5.5 → GPT-6 release, (d) deeper integrations with Apple, Google, Microsoft. The May 14-15 Codex blast already constitutes part of the counter. Expect a new funding announcement within 6 months.

Google DeepMind. As an early Anthropic investor, it's partly an ally. But on the Gemini lineup, direct competition. Watch Google I/O 2026 (May 19) for Gemini Ultra Pro plus a possible "Gemini for Enterprise" SKU. Google.org-level "AI for Good" expansion to mirror Gates-style partnerships.

xAI (Elon Musk). Anthropic taking a big share of Colossus 1 compute is a direct hit on xAI. Likely counter: a Grok 4 round at $100B+. Tighter Twitter/X integration for "real-time context" differentiation. Politically, xAI plays the Trump-administration-friendly card for government revenue.

Microsoft. Largest OpenAI investor, but also integrating Anthropic models on Azure. Multi-lab hedging strategy is intentional. Anthropic's valuation jump = renewed pressure on Microsoft's own AI revenue.

Chinese labs. DeepSeek, Moonshot, Qwen all accelerating model releases. Direct Chinese government backing (funding + data + GPU allocation) is forging a "U.S.-China two-bloc" AI market. Anthropic's valuation jump translates into uplift pressure on Chinese-lab valuations.

So What Actually Changes — by Persona

AI developers and engineers. Anthropic's revenue explosion may enable Claude API price cuts. Today Claude Opus 4.7 is $15/$75 per 1M input/output tokens — could drop toward $10/$50 over the next 6-12 months. OpenAI's GPT-5.5 faces similar pressure. Developers benefit directly. Cowork/Claude Code users may see free-tier expansion or higher usage caps.

Enterprise IT and digital-transformation executives. Lab revenue blowing up signals "your AI adoption is too slow." Easier to justify accelerated AI adoption to the CFO and CEO. Recommendation — multi-vendor strategy. Run OpenAI + Anthropic + Google Gemini + your own open-source models in parallel to avoid lock-in.

Investors, LPs, VCs. A $1T+ AI lab era is materializing. Fund-level pressure to "get AI lab exposure" intensifies. But entry pricing is high. Alternative exposure routes: (a) AI infrastructure (Cerebras, CoreWeave, Nvidia adjacencies), (b) AI application startups (Cursor, Anysphere, Perplexity, Notion AI), (c) AI data and labeling (Scale AI successors).

Consumers (U.S., Korea, OECD). Direct effect — Anthropic's revenue → infrastructure expansion → model performance acceleration. Claude Opus 5 (working name) plausibly ships late-2026 or early-2027. Free tiers likely expand. ChatGPT vs. Claude becomes a "both are good free options" choice for everyday users.

Big-tech employees (Google, Microsoft, Meta). Anthropic's revenue/employee ratio ($29M) is 10-20× big tech ($1-3M). Internal pressure mounts: "why aren't we this efficient?" Expect big tech to recruit ex-lab talent and tighten revenue-per-employee discipline.

Engineers (career-wise). Anthropic's valuation jump = stock-option uplift for 1,500 employees. Lower likelihood of talent attrition over the next 12 months, even as inbound recruiting accelerates. "Joining Anthropic" is one of the most competitive AI lab job markets in the next 6-12 months.

References

• Bloomberg — Anthropic in Talks to Raise $30 Billion at $900 Billion Valuation
• Sherwood News — Anthropic in talks at up to $950B
• tbreak — Anthropic seeks $50bn at $950bn valuation in 2026
• Reuters — Anthropic ARR hits $30B
• Dario Amodei — Machines of Loving Grace (2024)
• spoonai 2026-05-17: Anthropic × Gates Foundation $200M partnership (today)
• spoonai 2026-05-17: PwC × Anthropic alliance expansion (today)
• spoonai 2026-05-12: OpenAI Deployment Company $4B launch