Anthropic's one trillion, and DeepSeek's 10 billion

Original Title: One Trillion for Anthropic, 10 Billion for DeepSeek

Original Author: Dongcha Beating

Original Source:

Repost: Mars Finance

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

A screenshot was circulating wildly among investors, showing Anthropic’s implied valuation quietly crossing a line: one trillion dollars on secondary market and derivative platforms like Caplight, Ventuals.

Brief, but real — surpassing OpenAI.

No official announcement, no press release, no CEO Dario Amodei stepping out to say anything, just the market voting itself.

Investors watched the revenue curve excitedly, with Anthropic’s annualized revenue rising from $9 billion at the end of 2025 to $30 billion, a 233% increase in four months, then spreading the word: the AI leader has changed.

Let’s clarify one thing. Anthropic’s latest official post-IPO valuation was $380 billion after completing Series G in February 2026, with several venture capital firms later quoting $800 billion or higher, but Anthropic has not accepted those offers.

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

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

DeepSeek is planning its first external financing since its founding, targeting a valuation of over $10 billion, with at least $300 million to be raised. The first time in three years.

One is being chased by capital, reaching the trillion-dollar threshold. The other has kept capital at bay for three years, then chose a moment they believed was right, and slightly cracked open the door.

Reading these two pieces of news together reveals the same story: this spring, two of the most important AI companies in their respective countries have both reached the edge of their own paths.

Anthropic’s List of Backers

First, 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 GIC and hedge fund Coatue, with co-investors including Blackstone, Goldman Sachs, JPMorgan Chase, Qatar Investment Authority, Temasek. Nvidia pledged up to $10 billion, Microsoft up to $5 billion.

Read this list: Singapore sovereign fund, Qatar sovereign fund, the largest American investment bank, Nvidia, Microsoft.

This is a list of alliances. Global capital is voting with real money: the voice of AI should stay in the US, in this company’s hands.

Two months later, the results came.

According to data from Ramp, an enterprise expense management platform, in March 2026, up to 73% of new funds flowing into enterprise AI services went to Anthropic, with OpenAI’s share dropping to 27%. Just ten weeks earlier, the two were evenly split at 50:50.

Their core weapon is Claude Code, with annual revenue exceeding $2.5 billion, more than doubling since early 2026, and enterprise subscription users quadrupling.

This reversal can be understood as follows: OpenAI is building a consumer-facing Disney, relying on foot traffic to sell tickets. Anthropic is building a toll road to core enterprise systems, with tolls much higher than tickets, and once on the road, it’s not easy to switch lanes.

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

When customers buy services via AWS, Google Cloud, and other platforms, Anthropic counts the entire amount paid by the customer as revenue, including the portion paid to cloud providers. Excluding this part, Anthropic’s true revenue is about $22 billion, not surpassing OpenAI’s $25 billion.

The tone of this document resembles two former colleagues exposing each other’s shortcomings.

Understanding this memo requires some background. Anthropic’s private market valuation is about $600 billion, a significant premium over the previous round, while OpenAI’s secondary market valuation is about $765 billion, roughly 10% below its last funding round. The old boss is starting to face pressure in the capital market, and releasing this document is both a way to hit the opponent and to stabilize their own footing.

Then there’s that number that seems out of place in the celebration: Anthropic expects to be profitable only by 2027. $30 billion in annualized revenue, a $380 billion valuation, each funding round breaking records, but profit still is in the future. The higher the valuation, the greater the investor expectations, the faster the burn, and the more urgent the next round of financing. This cycle is something Anthropic cannot actively break; it can only run fast enough to sustain itself. That’s the invisible wall it faces.

Meanwhile, DeepSeek has kept the entire investment circle waiting for three years.

Next, about Liang Wenfeng.

After R1 exploded, the entire Chinese investment scene was thrown into chaos. Zhu Xiaohu, who just said “not optimistic about startups building large models,” publicly stated that price was no longer the main concern, participation mattered more. Tencent executives went, Alibaba executives went, various VCs came knocking.

Rumors of Alibaba investing $1 billion, C-round $700 million, surfaced one after another, only to be denied each time.

Liang Wenfeng kept the entire investment circle outside, waiting for three years.

His reason was a simple sentence: “No financing plans in the short term. Our problem has never been money, but the ban on high-end chips.”

Huanfang Quantitative invested 3 billion RMB in DeepSeek’s first phase of R&D, entirely supported by profits from quantitative private funds. He truly isn’t short of money; his problem is chips, and financing can’t solve that.

As for why he refuses external investment, he has another concern: external investors might interfere with company decisions.

Reading Liang Wenfeng’s background, one feels a consistent thread. Born in 1985 in Zhanjiang, Guangdong, graduated from Zhejiang University’s School of Information and Electronic Engineering, he went straight into quantitative investing without looking for a job. In 2015, he founded Huanfang Quantitative, and in 2019, invested nearly 200 million yuan to build a computing cluster “Firefly No. 1,” equipped with 1,100 GPUs.

When A100 launched, he was among the first in Asia-Pacific to get the chips, beating many companies. In 2021, he invested another 1 billion yuan to build “Firefly No. 2,” with about 10k A100s. In 2023, he shifted focus to large models, founding DeepSeek.

Every step he takes is guided by an engineer’s pre-judgment: prepare the tools first, then do the work. Refusing funding is one of his tools.

But now, this tool is starting to fail.

DeepSeek’s absolute salary levels are not low, but they can’t match the equity incentives and valuation premiums of giants like ByteDance, Alibaba, Tencent. Liang Wenfeng has begun pushing for valuation work, clarifying option pricing, giving the team more certainty.

Without external funding, there’s no market-based valuation, and no option value. For a top engineer, working at DeepSeek might mean changing the world, but you won’t get a piece of equity that can be calculated as wealth.

In January 2026, Zhipu listed on the Hong Kong Stock Exchange, followed by MiniMax’s listing. Peer companies’ options are being cashed out, and talent pressure at DeepSeek is becoming more real.

Another issue is emerging: whether DeepSeek and Huanfang should shift from “mainly focused on research” to “building a business that generates substantial income and eventually profits.” This discussion itself is a crack in the door.

The first financing target valuation exceeds $10 billion, while the company’s valuation in 2025 was about $3.4 billion. If the financing completes, the valuation will jump several times. Raising $300 million on a $10 billion valuation results in less than 3% dilution. This number is very restrained, like someone feeling the temperature at the door handle before gently pushing it open.

Liang Wenfeng’s three years of independence have earned him the greatest bargaining chip. He opened the door when he was most confident.

Two Civilizations at the AI Table

Putting these two stories together reveals a hidden thread.

Anthropic’s Series G investors include GIC (Singapore), Qatar Investment Authority, Blackstone, Goldman Sachs, Nvidia, Microsoft.

Behind this list is a complete logic: AI discourse power should stay in the US. “Safe, trustworthy” AI is the next infrastructure, and every dollar invested is a bet on this judgment.

DeepSeek’s initial funding, with potential investors including Alibaba, state funds, and other top domestic institutions, marks China’s capital’s first public valuation of a top AI research institution. It bets on a different logic: technological independence, open-source ecosystems, and domestic computing power.

These two lists, placed at the same table, represent two civilizations betting on their respective paths.

Closed source and open source are also two power structures in this game.

Anthropic remains fully closed source, relying on enterprise trust premiums. Its monthly active users generate $211 in revenue, selling not just model capabilities but a sense of reassurance endorsed by experts—you don’t need to understand it, just trust it.

Liang Wenfeng says open source “is more a culture than a business strategy; contributing to open source earns us respect.” The former consolidates the definition of “what is good AI” in the hands of a few, the latter leaves it to global developers to discuss.

These are two political stances on the future of AI.

But both companies face the same fundamental question: when you grow big enough, what do you use to prove your worth?

Anthropic’s answer is revenue growth and enterprise clients, but profitability is only expected in 2027. The old boss keeps poking holes. DeepSeek’s answer is still forming.

Epilogue

This race has no referee.

Anthropic’s valuation surged toward a trillion, but profit is expected only by 2027. How long are the world’s smartest sovereign funds and top investment banks willing to wait? The history of AI is short; no one has seen how such a company lands softly or crashes hard. Everyone is groping in the dark, each with their own approach.

DeepSeek’s problem is the cost of choice. After financing, external shareholders come in, and Liang Wenfeng’s long-guarded independence might not last long. Once the door opens, no founder can fully control what comes in afterward.

Dario Amodei describes himself as “an explorer seeking a third way between the narrow paths of rushing to heaven and falling into hell.” People around Liang Wenfeng say that AGI is his ultimate goal, with money and commercialization not a priority.

Both believe they are doing something more important than fundraising.

Capital markets don’t believe in faith, only in profit and loss statements.

Three or five years from now, when we revisit this ledger: will the company that once surged toward a trillion valuation prove it was worth it? Will the company that used three years of independence to earn respect and then take the first step stay true to its original intention?

Both paths are still unfinished.