
DeepSeek is in talks with Tencent Holdings and Alibaba Group for its debut DeepSeek funding round at a valuation above $20 billion, according to The Information, doubling an earlier $10 billion target in days amid intense investor demand.
The round marks a sharp reversal for a company that built its brand on self-sufficiency, and signals that China's open-source AI leaders can no longer fund frontier model development from internal balance sheets alone. It deepens the big-tech consolidation already visible across Chinese AI, where Tencent and Alibaba have backed rivals Zhipu, MiniMax and Moonshot.
DeepSeek valuation doubles in days on investor demand
DeepSeek originally targeted roughly $300 million from the raise at a $10 billion valuation, per earlier reporting from The Information. The target moved above $20 billion within days as interest intensified, and people familiar with the talks told the publication the final figure may move higher before the round closes. Both the valuation and the amount being raised remain subject to change.
Even at $20 billion-plus, DeepSeek sits well below the top tier of Western AI labs — Anthropic closed a $30 billion round at a $380 billion valuation in February and is fielding VC offers at $800 billion, while xAI is valued north of $200 billion. The gap underscores how much cheaper the Chinese AI stack remains, even as the capital race widens.
Alibaba Group Holding (NYSE: BABA) shares rose about 2% on Wednesday following the report, according to Investing.com. None of the companies commented to The Information.
Tencent-Alibaba AI investment signals shift for the self-funded startup
Founder Liang Wenfeng, whose hedge fund High-Flyer (Zhejiang High-Flyer Asset Management) has bankrolled the company since 2023, previously declined multiple offers from Chinese VCs and reportedly trained DeepSeek's flagship model for about $5.6 million — the figure that upended assumptions about AI training costs when the R1 model launched in early 2025.
Tencent told investors in March it plans to lift AI spending again in 2026 after chip export restrictions capped 2025 capex at roughly 79 billion yuan ($11.5 billion), per Reuters. S&P Global Ratings has forecast the company's AI-driven capex could push above 80 billion yuan in 2026 as it procures more chips for internal and external demand. Alibaba has moved in parallel, pouring investment into its Qwen model family and cloud infrastructure.Costs shift from training to inference as V4 model nears launch
Costs shift from training to inference as V4 model nears launch
The funding pivot reflects a broader shift in DeepSeek's cost base: from one-off training runs to sustained inference — the compute cost of actually running models for users — and product buildout. That burden has been compounded by researcher departures, a delayed V4 next-generation model, and months spent porting the company's architecture away from export-restricted Nvidia chips toward Huawei Ascend and Cambricon alternatives.
V4 is now slated for release in late April, founder Liang told Chinese outlets earlier this month. A deal at this level would give DeepSeek fresh runway to close the gap with U.S. frontier labs while defending its position against domestic rivals racing to launch agentic AI products — software that can carry out multi-step tasks without human intervention — including Tencent's own OpenClaw suite.
The Daily Crypto Integrated Newsletter
Stay updated on the latest crypto news across Ethereum, Solana, AI and Macro, distilled into a 2-minute read.
Delivered via Substack.


