DeepSeek has operated for years as an outlier. The Chinese AI lab built powerful models on modest budgets. It released them openly. And it largely rejected outside capital. That chapter ends now.
The startup stands on the verge of raising about $7.4 billion in its first external funding round. The deal would value the company at up to $59 billion post-money. Investors include Tencent Holdings and CATL. Founder Liang Wenfeng commits roughly $2.8 billion of his own money. Reuters first detailed the terms on June 3.
But the numbers only begin to tell the story. This round marks a sharp turn. DeepSeek moves from a research outfit bankrolled by its founder’s quant hedge fund to a commercial player with boards, expectations and a mandate to generate revenue. The shift arrives at a moment when Chinese AI faces hardware limits, talent wars and intensifying pressure to deliver agents that act with minimal human oversight.
Liang, who co-founded High-Flyer Capital Management after earning degrees from Zhejiang University, launched DeepSeek in 2023. Profits from the hedge fund covered early costs. The arrangement let the company stay independent. No venture terms. No dilution pressure. For 18 months it remained the most discussed AI lab that almost nobody could back. The Next Web reported the investor breakdown and Liang’s controlling stake on the same day.
Everything changed when DeepSeek released its V3 and R1 models early last year. They trained cheaply. They excelled at reasoning. They shipped with open weights. Silicon Valley took notice. American labs suddenly confronted the possibility that frontier performance did not require frontier spending. DeepSeek became China’s national AI champion almost overnight.
Yet praise brought new problems. Rivals poached researchers. Compute demands grew. The industry pivoted toward AI agents capable of complex, multi-step tasks. DeepSeek’s April announcement of its V4 model claimed new open-source benchmarks. Independent tests, however, showed it trailing top entries from both U.S. and domestic competitors. The gap underscored a hard truth. Staying competitive now requires scale.
So Liang reversed course. He will inject 20 billion yuan personally, about 40 percent of the round. That preserves his dominance. His post-money ownership reportedly climbs toward 34 percent directly and over 80 percent including indirect holdings. Tencent weighs in with roughly 10 billion yuan. CATL, the battery giant expanding into data-center power infrastructure, eyes 5 billion yuan. Talks continue with China’s national AI fund, NetEase, JD.com, IDG Capital and Monolith Capital. Fewer than 10 investors total. The round could close within weeks.
Commercial pressure meets efficiency legacy.
This capital infusion buys time and hardware. It funds the shift toward revenue-generating products rather than pure research papers. Yet it also tests DeepSeek’s core identity. The company earned respect by doing more with less. Open models. Transparent methods. Modest training runs that outperformed expectations. Will new investors demand closed weights, higher spending and tighter secrecy? The question hangs over every term sheet.
Geopolitics shapes the answer. Western export controls block access to the latest American chips. DeepSeek cannot match the multi-billion-dollar compute budgets of OpenAI or Anthropic even if it wanted to. “Western export bans mean DeepSeek cannot access frontier American silicon,” said Alfredo Montufar-Helu, Beijing-based managing director at Ankura China Advisors, in the Reuters report. “Without the ability to buy that hardware, they have no reason to match the multi-billion-dollar computing budgets of their U.S. rivals.”
The comparison flatters DeepSeek on efficiency but highlights the valuation gap. OpenAI and Anthropic have drawn Western capital at hundreds of billions. DeepSeek’s $52 billion to $59 billion post-money figure, while enormous for a first round, remains a fraction of those sums. Still, the deal ranks among the largest private tech fundraisings in Chinese history.
Recent reporting shows the valuation climbed fast. Early April discussions hovered near $10 billion. By late April the figure doubled. May brought leaks of $45 billion. Now the ceiling sits near $59 billion. Founder capital anchors the structure and signals confidence. The Information noted in May that DeepSeek is accelerating revenue plans alongside the fundraising.
CATL’s participation reveals another layer. The company supplies batteries for electric vehicles. It now eyes AI data centers hungry for reliable power and energy storage. Tencent brings cloud infrastructure and distribution. Its own Hunyuan model trails leaders such as ByteDance’s Doubao and DeepSeek itself. A deeper tie could help Tencent close that gap while DeepSeek gains enterprise reach.
NetEase and JD.com add gaming and e-commerce muscle. The national AI fund ties the effort directly to Beijing’s self-sufficiency drive. Domestic strategic capital fills the space left by restricted foreign money. The lineup reflects necessity as much as choice.
DeepSeek has not signaled IPO plans. By contrast, OpenAI and Anthropic eye public listings. The Chinese lab instead focuses on closing this round and translating technical admiration into sustainable income. Chatbots alone won’t suffice. Agents that plan, execute and iterate with limited supervision represent the next battleground. Training and inference costs rise accordingly.
And here lies the tension. DeepSeek built its reputation on frugality and openness. The V3 and R1 models forced a global recalibration of assumptions about Chinese capabilities. They proved high performance need not require massive closed-door budgets. Now the company must scale without losing that edge. Investors will want returns. Talent demands competitive pay. Compute still costs money even when optimized.
Liang’s personal commitment buys breathing room. It keeps control inside the founding team. It signals to employees and partners that the mission remains intact. But boards introduce new voices. Commercial targets introduce new metrics. The next models will reveal whether the discipline survives.
Industry watchers note the speed of escalation. Valuation talk jumped from $10 billion to nearly $60 billion in weeks. Such momentum reflects genuine excitement around DeepSeek’s technical output. It also captures the froth in China’s AI race. Talent retention drove early fundraising interest. Compute expansion and agent development drive the current size.
Whatever the final terms, the round formalizes DeepSeek’s place at the center of China’s AI push. The lab no longer operates as a side project of a hedge fund. It stands as a national champion with a war chest, strategic backers and a mandate to deliver products that matter in the agent era. The cheques are nearly signed. The real test begins once the money clears.
WebProNews is an iEntry Publication