Anthropic Reaches $380 Billion Valuation: What the Number Means for AI’s Investment Landscape

Photo: Solen Feyissa / Unsplash

A $380 Billion Company That’s Never Turned a Profit

Enterprise customers now number 500+ at $1M+ annually, signaling decisive market share capture while competitors race to scale.

Anthropic has never had a profitable year. It burns roughly $3 billion in cash annually, expects to lose money through 2027, and projects breaking even sometime in 2028. Its investors just valued it at $380 billion.

That number needs unpacking — not because it’s wrong, but because it tells you something specific about how investors are pricing the AI race right now.

The share sale announced Monday is a tender offer: outside investors, not Anthropic, are buying up to $6 billion worth of shares from current and former employees who’ve been at the company for at least 12 months. The sale price implies a $350 billion pre-money valuation — slightly below the $380 billion post-money figure from the February 12 Series G close, which is standard for secondary transactions. Employees get liquidity. New investors get exposure. Anthropic gets nothing directly from this particular transaction except retained talent.

What Vesting Actually Means — and Why This Sale Exists

Here’s the educational part that most headlines skip.

When Anthropic hires an engineer, researcher, or executive, part of their compensation is equity — shares in the company. But that equity doesn’t land in their account on day one. It vests, meaning it becomes theirs gradually over time. A typical tech startup equity package vests over four years with a one-year cliff: nothing until month 12, then 25% unlocks at once, then the remaining 75% unlocks monthly over the following three years.

The problem for employees at private companies is that vested equity is technically yours — but you can’t sell it. There’s no stock exchange for Anthropic shares. Your $500,000 in vested equity might as well be a number on a spreadsheet until one of three things happens: the company goes public (IPO), gets acquired, or runs a tender offer like this one.

That’s the specific mechanism here. A tender offer is a structured, company-facilitated process where outside investors offer to buy employee shares at a set price. Anthropic organizes it, sets the valuation, finds the buyers, and eligible employees decide how much (if any) of their vested shares to sell. The company itself isn’t the buyer — it doesn’t spend cash on this. But it controls the process, which gives it enormous leverage over who gets liquidity and when.

The math on Monday’s sale is striking. An employee who joined Anthropic in 2022 at a modest equity grant of $200,000 (pre-money valuation was around $4 billion then) now holds shares worth roughly 5-10x that amount at current pricing. Someone with $500,000 in vested equity from a 2023 grant could walk away from this tender offer with over $2 million in cash. That’s life-changing money, and it explains why these programs matter enormously for talent retention in a market where Google, Meta, and Microsoft are waving eight-figure packages at top AI researchers.

OpenAI has run the same playbook repeatedly, including a $6.6 billion secondary at a $500 billion valuation last year. Stripe and SpaceX have both used tender offers for years. At this point it’s less a perk and more a standard retention mechanism for elite private tech companies.

The Revenue Reality Behind the Valuation

Here’s where the numbers get genuinely interesting.

Anthropic grew from $1 billion in annualized revenue in December 2024 to $4 billion by July 2025, $9 billion by December 2025, and an estimated $14 billion as of February 2026. That trajectory — 14x in 14 months — has no precedent in enterprise technology history at this scale.

Sacra estimates that over 300,000 businesses are now Anthropic customers, accounting for roughly 80% of revenue. More than 500 companies spend over $1 million annually on Claude, up from roughly a dozen two years ago. Eight of the Fortune 10 are customers. Claude Code, the agentic coding product, hit $2.5 billion in annualized revenue in February 2026 alone — more than doubling since January 1.

At a $380 billion valuation with ~$14 billion in annualized revenue, Anthropic is trading at roughly 27x revenue. That’s not cheap. For context, Microsoft trades at about 13x revenue, Salesforce at roughly 8x. But Anthropic is growing at 7-10x annually where those companies grow at 10-15% annually. Whether 27x is reasonable depends entirely on whether you believe the revenue trajectory sustains. The company’s own projections — $18 billion in 2026, $55 billion in 2027, $70+ billion by 2028 — are the bet investors are making.

One more number: Anthropic projects positive cash flow by 2028. That’s not profit. It means more cash coming in than going out from operations. The path there depends on gross margins expanding from roughly 40% today to the projected 77% by 2028, as inference costs decline and enterprise contracts scale. If that margin expansion happens on schedule, the valuation is arguably conservative. If it doesn’t, the model is structurally challenged.

The Pentagon Subplot: Safety as a Business Risk

Quietly running alongside Monday’s share sale announcement is a story with real strategic implications for Anthropic.

xAI — Elon Musk’s AI company — this week signed an agreement with the Pentagon to deploy Grok in classified military systems. Until now, Anthropic’s Claude has been the only AI model operating at that classification level. The Pentagon is moving to replace or supplement Claude because Anthropic has refused to agree to “all lawful purposes” terms — specifically, they’ve drawn hard lines against using Claude for mass surveillance of Americans and the development of fully autonomous weapons.

Anthropic CEO Dario Amodei met with Defense Secretary Pete Hegseth Tuesday. Sources described it as a tense meeting with an ultimatum on the table.

The irony is sharp: Anthropic’s founding thesis is AI safety. Its “Constitutional AI” approach and its refusal to remove safeguards is literally the product. But the Pentagon’s position is that those safety constraints are contractual blockers. The government’s budget for AI is enormous — the existing contracts alone were $200 million per company across xAI, OpenAI, Google, and Anthropic — and the classified systems business is strategically valuable beyond the dollar amounts.

xAI agreed to the “all lawful use” standard that Anthropic rejected. Google and OpenAI are reportedly in talks to expand their own classified access. Anthropic may be the only frontier AI company willing to lose government business over safety principles — which is either its greatest brand asset or its most expensive policy decision, depending on your perspective.

The Competitive Map: Who’s Who and What They’re Worth

Anthropic at $380 billion is the second-most-valuable pure-play AI company in the world. Here’s the full landscape as it stands in February 2026:

OpenAI — The current leader by valuation. A $500 billion secondary sale valuation in October 2025, with an IPO being prepared that could target $750 billion to $1 trillion. OpenAI generated approximately $14.2 billion in revenue in 2025 and projects $20+ billion in 2026. The caveat: it’s burning cash aggressively, projecting cumulative losses of $115 billion through 2029 as it builds infrastructure. ChatGPT has 800 million weekly users. The consumer moat is real.

xAI (Grok) — Elon Musk’s company closed a $20 billion funding round in January 2026 at roughly $230 billion valuation. Revenue details are limited, but the Pentagon classified contract and the integration into X (formerly Twitter)’s 500+ million user base are distribution advantages no other AI lab has. The government deal this week is a meaningful escalation of xAI’s ambitions beyond consumer chatbot.

Google Gemini — Google is the sleeping giant of this race. Gemini 3.1 Pro is the current flagship. Google has 750 million weekly Gemini users, custom TPU infrastructure (no NVIDIA dependency), distribution inside Search, Chrome, Android, and Workspace, and a profitable parent company funding the build-out. Google isn’t valued separately from Alphabet, but the AI business embedded in a $2+ trillion market cap company means it’s effectively competing with unlimited runway.

Meta AI (Llama) — Meta takes a different path entirely: open source. Llama 4 is available for anyone to download, run, and fine-tune. Meta doesn’t need AI revenue from models — it monetizes through advertising on Facebook and Instagram, using AI to improve targeting. The open-source strategy is arguably more disruptive to the paid API model than any closed competitor.

DeepSeek — The Chinese wildcard. A hedge fund offshoot (controlled by founder Liang Wenfeng, not VC-backed) that published models matching GPT-quality at a fraction of the cost. Not publicly valued. DeepSeek V3.2 reportedly costs roughly 30x less to run than comparable U.S. models. DeepSeek V4 is expected imminently — new architecture code (“MODEL1”) appeared in its GitHub repo, with sources suggesting a late-February release. If V4 delivers on the V3 trajectory, it will again force the entire industry to reprice inference economics.

ByteDance / Doubao — TikTok’s parent company has 155 million weekly active Doubao users — the most of any Chinese AI app. ByteDance launched Seed 2.0 and Seedance 2.0 (video generation) in February 2026, with its new Dola-Seed-2.0 model ranking ahead of Grok 4.1 and Gemini 3 on user preference benchmarks. ByteDance is privately held with no disclosed AI-specific valuation, but the parent company was valued at approximately $300+ billion in recent secondary markets.

Kimi / Moonshot AI — Beijing-based startup, currently seeking a valuation around $10 billion (up from $4.3 billion in December 2025, a 132% increase in two months). Kimi K2.5 launched January 26 with agentic capabilities and a 96.1% score on AIME 2026 math benchmarks. Its challenge: ByteDance’s Doubao has distribution through TikTok/Douyin; DeepSeek has cost leadership. Moonshot is technically impressive but squeezed from both sides.

The Question Nobody’s Answering

If you put all these numbers next to each other — OpenAI at $500 billion burning $115 billion through 2029, Anthropic at $380 billion projecting profitability in 2028, xAI at $230 billion with limited disclosed revenue, Chinese models matching capabilities at 30x lower cost — the honest question is: how many of these companies need to exist?

The AI model business has genuine commoditization pressure. Anthropic’s own gross margins are expected to expand as inference costs fall industrywide. The technology that makes Claude valuable today will be replicated (or approximated) by DeepSeek or Doubao or Qwen faster than most Western investors are pricing. Meanwhile, 79% of Anthropic’s customers are already OpenAI customers. Churn rates are nearly identical at 4%. The market isn’t winner-take-all — but it’s not infinitely expansive either.

What Anthropic has that most competitors don’t is an enterprise trust brand, a safety reputation that matters to regulated industries (finance, healthcare, legal), and Claude Code’s hold on developer workflows. Those are real moats. Whether they’re worth the 13-point premium over OpenAI’s revenue multiple is the genuine debate.

Dario Amodei said in a recent presentation that he believes we’re approaching “a country of geniuses in a datacenter.” If he’s right, $380 billion is cheap. If the timeline slips by two years, it’s very expensive.

What to Watch This Week

Wednesday, February 25 — NVIDIA Q4 FY2026 Earnings: The most important AI earnings call of 2026. Watch Q1 FY2027 guidance specifically — consensus is ~$70–75B. Anything above that breaks the three-month stock consolidation. NVIDIA’s commentary on enterprise AI adoption (beyond hyperscalers) will directly affect sentiment around Anthropic and the entire AI buildout thesis.

Tuesday ongoing — Anthropic Pentagon meeting fallout: Dario Amodei met with Pete Hegseth Tuesday. Watch for any public statement on whether Anthropic will modify its safeguards to retain classified Pentagon access, or hold firm and cede that market to xAI and others.

Imminent — DeepSeek V4: Sources suggest a late-February release. If V4 repeats the V3 pattern — matching frontier performance at dramatically lower cost — it will again put pressure on the pricing models that underpin Anthropic’s and OpenAI’s revenue trajectories.

Friday, February 27 — PPI data: January Producer Price Index drops Friday. A cool reading starts reopening the second-half rate-cut debate. A hot reading keeps the Fed on hold through mid-2026. Either way, it sets the tone for the FOMC meeting on March 18–19.

—