Anthropic’s one trillion, and DeepSeek’s ten billion

Original author: Lin Wanwan

On April 17, 2026, the AI financing circle got excited again.

A screenshot has been wildly circulating among investors. Anthropic’s implied valuation—on secondary-market and derivatives platforms such as Caplight and Ventuals—quietly crossed a line: $1 trillion.

Brief, but real—surpassing OpenAI.

No official announcement, no press release, and no CEO Dario Amodei stepping out to say anything—only the pre-IPO market voting for itself.

Investors watched the revenue curve and got excited. Anthropic’s annualized revenue rose from $9 billion at the end of 2025 to $30 billion. In four months, it grew 233%. Then they rushed to spread the news: the AI boss has changed.

First, let’s get one thing clear. Anthropic’s latest official post-money valuation is $380 billion after completing its Series G in February 2026. After that, several venture capital firms quoted $800 billion or higher, but Anthropic has not accepted.

That $1 trillion is an implied figure on secondary-market platforms.

Almost the same day, another piece of news came from Hangzhou.

DeepSeek is planning its first external financing since its founding. The target valuation is over $10 billion, and it plans to raise at least $300 million—three years in, for the first time.

One is being chased by capital and reaching right up to the doorstep of a trillion. The other kept capital out for three years, then chose a moment it believed was right, cracking the door open just a little.

Read these two pieces of news together and what you get is the same story: this spring, two countries, two of the most important AI companies in each country, have reached the boundary of their own paths.

Anthropic’s lineup of backers

First, let’s talk about Anthropic.

On February 13, 2026, Anthropic completed its Series G financing, raising a total of $30 billion, with a post-money valuation of $380 billion. The lead investors were Singapore’s government investment company GIC and the hedge fund Coatue. Co-investors included Blackstone, Goldman Sachs, JPMorgan Chase, the Qatar Investment Authority, Temasek, and Nvidia, which committed up to $10 billion. Microsoft committed up to $5 billion.

Say this list aloud: Singapore’s sovereign fund, Qatar’s sovereign fund, the largest American investment bank, Nvidia, Microsoft.

This is a lineup of backers. Global capital is voting with real money: the voice of AI—should stay in the United States, and be held in the hands of this company.

Two months later, the results came in.

According to monitoring data from Ramp, an enterprise spend management platform, in March 2026, among incremental funds used for enterprises’ first purchases of AI services, as much as 73% flowed to Anthropic, while OpenAI’s share fell to 27%. Only 10 weeks earlier, the two sides had been in a 50:50 balance.

Their core weapon is Claude Code. Annualized revenue exceeds $2.5 billion—more than doubled since early 2026—and the number of enterprise subscription users has grown fourfold.

You can understand this reversal like this. OpenAI is building a consumer Disney, charging tickets based on foot traffic. Anthropic is building a toll highway to the core enterprise systems. The toll is far higher than the ticket—once the car is on the road, it won’t easily change lanes.

Just days after Anthropic announced it had overtaken, an internal memo authored by Denise Dresser, OpenAI’s Chief Revenue Officer, was leaked. It accused Anthropic of inflating about $8 billion of revenue using a “total amount method.”

When customers buy services via platforms such as AWS and Google Cloud, Anthropic counts the entire amount paid by the customer as revenue, including the portion that must be shared with cloud service providers. If that portion is removed, Anthropic’s true revenue is about $22 billion, which does not exceed OpenAI’s $25 billion.

The wording of this document feels more like two former colleagues exposing each other.

To understand this memo, you need some background. Anthropic’s valuation in the private market is about $600 billion—at a significant premium compared with the prior funding round. OpenAI’s valuation in the secondary market is about $765 billion—about 10% below its prior funding round. The old boss is starting to come under pressure in the capital market. Issuing this document is both a way to take aim at the opponent and to steady its own footing.

Then there’s that figure that doesn’t fit in with the celebratory noise. Anthropic expects to turn a profit only in 2027. $30 billion in annualized revenue, a $380 billion valuation—each funding round breaks records, but profitability is still “the day after tomorrow.” The higher the valuation, the greater investors’ expectations, the faster the cash burn, and the more urgent the next financing becomes. Anthropic has no way to break this cycle on its own. It can only keep up by running fast enough to sustain it. This is its invisible wall.

And DeepSeek—he put the entire investment circle on hold for three years

Now, let’s talk about Liang Wenfeng.

After R1 went viral, the entire Chinese investment scene was thrown into uproar. Zhu Xiaohu—the person who had just said “I don’t think startups should build large models”—publicly stated that price was no longer that important; the key was participation. Tencent executives went. Alibaba executives went. VCs from every direction came one after another to pay visits.

Rumors emerged of Alibaba investing $1 billion, and then of a $700 million Series C—one after another, one after another, each time being debunked.

Liang Wenfeng simply kept the entire investment circle waiting outside the door, holding it off for three years.

His reason was one sentence: “In the short term, we have no plans for financing. The problem we face has never been money—it’s that high-end chips are banned.”

For DeepSeek’s first phase of R&D, Huanfang Quantitative funded it entirely with its own capital of 3 billion RMB, supported solely by the profits from quantitative private funds. He really isn’t short of money. What he lacks is chips—and financing can’t solve the chip problem.

As for why he wouldn’t accept outside investment, he had another concern: once external investors come in, they might interfere with company decision-making.

When you read Liang Wenfeng’s story, you feel a kind of consistency throughout. Born in 1985 in Zhanjiang, Guangdong, he graduated from Zhejiang University’s School of Information and Electronic Engineering. After graduating, he didn’t look for a job and instead went straight into quantitative investing. In 2015, he founded Huanfang Quantitative. In 2019, he invested nearly 200 million RMB to build a computing cluster called “Firefly No.1,” equipped with 1,100 GPUs.

When the A100 was launched, he grabbed chips ahead of many companies, becoming one of the first recipients in the Asia-Pacific region. In 2021, he poured another $1 billion into building “Firefly No.2,” with about 10,000 A100s. In 2023, he shifted computing power to build large models and founded DeepSeek.

For every thing he does, there is an engineer’s kind of pre-judgment: prepare the tools first, then start the work. Refusing financing is one of his tools.

But now, this tool is starting to malfunction.

DeepSeek’s compensation levels are absolutely not low, but they cannot match the equity incentives and valuation premium offered by market giants like ByteDance, Alibaba, and Tencent. Liang Wenfeng has begun pushing the company’s valuation work—clearly defining option pricing—to give the team more certainty.

Without external financing, there is no market-based valuation, and therefore no option value. For a top engineer, working at DeepSeek might mean you’re changing the world—but you can’t produce a piece of equity evidence that can calculate wealth.

In January 2026, Zhipu rang the bell on the Hong Kong Stock Exchange. MiniMax followed shortly after with its listing. Peers’ options are being realized, and the talent pressure at DeepSeek is becoming more and more real.

Another issue is being forced to come to the surface. DeepSeek and Huanfang’s executives are still discussing whether the company should shift from “primarily focused on research” to “building a business that generates substantial income and eventually profitability.” The discussion itself is a crack in the door.

This first external financing round targets a valuation of over $10 billion. In 2025, the company’s valuation was about $3.4 billion. If the financing is completed, the valuation will jump several times. Raising $300 million at a $10 billion valuation means the dilution ratio is less than 3%. This figure is extremely restrained—like someone placing a hand on the doorknob before opening the door, feeling the temperature, confirming there’s no danger, and only then pushing it open gently.

Liang Wenfeng used three years of independence to win the biggest bargaining chip for himself. He opened the door when he was at his most confident.

Two civilizations at the AI table

Putting these two stories side by side reveals a hidden thread.

Anthropic’s Series G backers: GIC (Singapore), the Qatar Investment Authority, Blackstone, Goldman Sachs, Nvidia, and Microsoft.

Behind that list is a complete logic: the voice of AI should stay in the United States. “Safe and trustworthy” AI is the next infrastructure. Every dollar that enters is a bet on this judgment.

DeepSeek’s first external financing, with potential investors including Alibaba and state funds and other top domestic institutions, is the first time China’s capital has publicly priced a top-tier AI research institution. It bets on a different logic: technological self-reliance, an open-source ecosystem, and locally built computing power.

These two lists, placed on the same table, represent two civilizations placing their bets on their respective paths.

Closed source and open source are also two choices of power structure in this game.

Anthropic stays fully closed. It relies on enterprise trust premiums. Monthly active users generate $211 in revenue. What it sells is not only model capability, but a reassuring feeling vouched for by experts—you don’t need to understand it; you just need to trust it.

Liang Wenfeng said that open source is “more a culture than a commercial behavior. Contributing to open source will earn us respect.” The former keeps the definition of “what is good AI” in the hands of a few. The latter hands it over to developers worldwide to discuss.

These are two political positions about AI’s future.

But in fact, both companies are facing the same problem: when you become big enough, how do you prove you’re worth that price?

Anthropic’s answer is revenue growth and enterprise customers, but profitability will only arrive in 2027—while the old boss keeps picking at it. DeepSeek’s answer is still taking shape.

Epilogue

This competition doesn’t have a referee yet.

Anthropic’s valuation is surging toward $1 trillion, but profits will likely have to wait until 2027. How long are the world’s most astute sovereign funds and top investment banks willing to wait? The history of AI is still too short. No one has seen how a company of this scale makes a soft landing, and no one has seen how it crashes hard. Everyone is feeling around in the dark—only the way they feel their way differs.

DeepSeek’s problem is the cost of choosing. After fundraising, external shareholders come in. How much longer can Liang Wenfeng’s independence—something he has guarded for a long time—hold? Once the door is opened, there is not a single founder in the world who can fully control what comes in after the door opens.

Dario Amodei positions himself as “an explorer searching for a third path between the narrow roads to heaven and the plunge into hell.” People around Liang Wenfeng say the ultimate goal is AGI, and money and commercialization are not high priorities.

The two of them each believe they are doing something more important than fundraising.

Capital markets don’t trust faith—they only trust the profit-and-loss statement.

Three years from now, or five years from now, when we come back to flip through this ledger: has the company whose valuation once surged toward $1 trillion proven it was worth that price? Has the company that used three years of independence to earn respect—then decided to take the first step—stayed true to its original intention?

Both paths are still not finished.