On April 17, 2026, a screenshot went viral in the AI investing world: Anthropic's implied valuation on secondary markets like Caplight and Ventuals briefly crossed the $1 trillion mark — surpassing OpenAI for a moment. No official announcement, no press release, just the pre-IPO market casting its vote.
Investors cheered the revenue trajectory: Anthropic's annualized revenue surged from $9 billion at end-2025 to $30 billion in four months, a 233% jump. The key driver was Claude Code, whose annualized revenue exceeded $2.5 billion, with enterprise subscription users quadrupling.
The G-round investor roster as a geopolitical alignment
Anthropic closed its Series G on February 13, 2026, raising $30 billion at a $380 billion post-money valuation. Lead investors included GIC and Coatue, joined by Blackstone, Goldman Sachs, JPMorgan, Qatar Investment Authority, Temasek, with Nvidia committing up to $10 billion and Microsoft up to $5 billion. This list represents global capital lining up behind the thesis that AI leadership should stay in the U.S.
Yet profitability remains distant — not expected until 2027. The $30 billion revenue against a $380 billion valuation means relentless cash burn and pressure to raise again. That's Anthropic's invisible wall.
OpenAI's counterpunch: disputed revenue inflation
Days after Anthropic's claimed lead, an internal memo from OpenAI's chief revenue officer Denise Dresser leaked, accusing Anthropic of inflating revenue by roughly $8 billion using a "gross method" — counting the full amount clients paid via AWS, Google Cloud, etc., without deducting the cloud providers' cut. Stripping that out, Anthropic's real revenue would be around $22 billion, below OpenAI's $25 billion. The memo was as much a defense of OpenAI's own market position as an attack on its rival.
DeepSeek's first funding: cracking the door after three years
Around the same time, word came from Hangzhou: DeepSeek was planning its first external fundraising since inception, targeting a valuation north of $10 billion and raising at least $300 million. After three years of refusing money.
Founder Liang Wenfeng, born 1985, a Zhejiang University graduate in information engineering, started as a quantitative trader. He founded High-Flyer in 2015, invested nearly $30 million in 2019 to build the "Firefly 1" cluster with 1,100 GPUs, secured A100s early to become one of the first in Asia Pacific, spent another $140 million on "Firefly 2" with ~10,000 A100s in 2021, and pivoted to LLMs in 2023 to create DeepSeek.
He kept investors outside for three years, saying: "Our problem has never been money, but the ban on high-end chips." Outside investors might also interfere with decisions. But that tool is now failing: without external funding, there's no market valuation, no stock option value. Rivals Zhipu listed in Hong Kong, MiniMax followed — option liquidity puts talent retention pressure on DeepSeek. Internally, the company debates shifting from "pure research" to "building a revenue-generating business."
The first raise targets over $10 billion valuation (up from ~$3.4 billion in 2025) with a dilution of less than 3%. It's as restrained as someone touching the door handle to feel the temperature before pushing open.
Two paths, one question
Anthropic is closed-source, monetizing enterprise trust — its monthly active users generate $211 in revenue. Liang says "open source is more a culture than a business action." Two different political visions for AI's future. But both face the same test: when you grow this big, how do you prove the valuation? Anthropic points to revenue growth and a 2027 profitability target; DeepSeek's answer is still being written.
Capital markets don't buy faith — they buy income statements. Both paths remain unfinished.