DeepSeek Doubled to $45B in Weeks — Beijing Funded It

A Chinese AI lab that has never taken outside venture capital just became one of the most valuable private AI companies on the planet. DeepSeek is in talks to close its first ever funding round at a valuation of around $45 billion — more than double the $20 billion figure that was floating only weeks ago, according to reporting from the Financial Times and Bloomberg surfaced by TechCrunch on May 6, 2026. The round is being led by a Chinese state investment vehicle, with Tencent and Alibaba reportedly looking to participate. For a company whose founder owns close to 90% of the equity and whose models live on consumer-grade Hugging Face downloads rather than locked-down APIs, this is not a normal Series A. It is the moment a sovereign AI strategy stops pretending to be a startup story.

From Zero VC to $45B in Weeks

The Math Behind the Mark

Until this round, DeepSeek had taken zero outside capital. The lab was built and funded almost entirely by Liang Wenfeng’s hedge fund High-Flyer, which is why he still owns roughly 90% of the company. According to the FT and Bloomberg reports, the talks moved from a $20 billion valuation to as high as $45 billion in only a few weeks — a reset that puts DeepSeek in the same neighborhood as Anthropic and xAI by paper value. The lead investor, China Integrated Circuit Industry Investment Fund (commonly called the “Big Fund”), is the same state vehicle Beijing has used for years to backstop domestic semiconductor champions like SMIC and YMTC. Tencent and Alibaba — both of whom have their own foundation models — are reportedly in talks to take checks alongside it, which means strategic capital is stacking on top of state capital.

Why a 90% Owner Suddenly Wants Investors

The Talent War Forced His Hand

According to the Financial Times, Liang did not pursue this round because he needed cash. He pursued it because his researchers were being poached. With nearly 90% of the equity sitting on his cap table, DeepSeek had almost nothing to offer engineers in stock — the single most powerful retention tool that OpenAI, Anthropic, and Chinese rivals like Alibaba’s Qwen team use to lock in talent. Raising at a $45B mark gives Liang a defensible currency to grant employee equity, while still keeping voting control. The implicit benchmark is brutal: every senior researcher at DeepSeek is now being valued in the same hiring market as their counterparts in San Francisco, where individual offers have crossed eight figures.

DeepSeek + Huawei: A Parallel AI Stack

Why Beijing Wants This Pair, Specifically

The detail buried in the TechCrunch summary is the most strategically loaded one: DeepSeek has been optimized to run on Huawei silicon. That makes the “Big Fund + DeepSeek + Huawei” combination an end-to-end Chinese AI stack — model weights, training and inference compute, and now state capital — that can in principle operate without any U.S. accelerator in the loop. For Beijing, this is the exact decoupling pattern it has been building since the 2022 export controls: vertically integrate, fund the weakest link, then let the market commercialize. The fact that DeepSeek keeps its model weights freely available on Hugging Face is part of the same playbook — open weights are how China exports influence at the model layer while the U.S. closes APIs.

What It Signals for OpenAI and Anthropic

Open Weights Become a Wedge

The American frontier labs are now competing against an open-weight competitor with sovereign-grade balance sheet support and a hardware partner that does not need ASML or NVIDIA to ship. DeepSeek’s edge has always been the cost narrative — train on a fraction of the compute, deploy on cheaper chips, give the weights away. A $45B funding round buys two things on top of that: enough runway to stop optimizing for survival, and enough equity to retain the team that built the cost advantage in the first place. For OpenAI and Anthropic, the implication is not that DeepSeek wins the U.S. enterprise market; it is that Chinese, Southeast Asian, Middle Eastern, and African deployments increasingly default to a non-Western stack — and that the global AI map starts looking less like one league and more like two.

Related

• OpenAI Chair’s Side Project Is Now Worth $15.8B
• The Pentagon Just Picked 8 AI Giants — Not Anthropic
• Anthropic’s CEO Just Issued a 6-Month Warning
• SAP Bets $1.16B on an 18-Month-Old German AI Lab
• Anthropic and OpenAI Both Launch Wall Street JVs

Sources

1. TechCrunch — DeepSeek could hit $45B valuation from its first investment round (May 6, 2026)
2. Financial Times — DeepSeek funding talks at $45B valuation (original reporting cited by TechCrunch)
3. Bloomberg — China Big Fund leads DeepSeek round; Tencent and Alibaba in talks