DeepSeek Raises $50B in Record First Round! Liang Wenfeng Invests $20B, DeepSeek V4.1 Set for June
Valuation quintupled in 21 days. DeepSeek’s first funding round hit $50B. Founder Liang Wenfeng personally invested $20B. V4.1 is scheduled for June.
In May 2026, one of the most notable pieces of news in the AI industry arrived: DeepSeek officially launched its first funding round, with a target of up to RMB 50 billion. If finalized, this will be the largest single funding round ever for a Chinese large-model company.
Funding Details: Liang Wenfeng Personally Invests $20B
According to The Information, the largest check in this round did not come from traditional VCs or internet giants, but from founder Liang Wenfeng himself — personally investing up to RMB 20 billion, accounting for 40% of the planned round.
The overall target for this round is up to RMB 50 billion. This means DeepSeek’s pre-money valuation nearly quintupled in 21 days, reaching approximately RMB 350 billion (about $50 billion).
Valuation Timeline
- Early April 2026: ~$10B, DeepSeek initiates first funding round
- April 22, 2026: Over $20B, Tencent and Alibaba begin investment discussions
- May 6, 2026: ~$45B, National IC Fund in talks to lead
- Early May 2026: Some reports suggest final valuation could reach $50B
A company once best known for its “no fundraising, no commercialization, no roadshows” stance is now preparing to raise a potentially record-breaking amount. Behind this shift are three pressing realities.
Three Pressures: Compute, Talent, Productization
Surging Compute Demand
Competition at the frontier is no longer about publishing a paper and training a model. Reasoning capabilities, Agent capabilities, ultra-long context, enterprise-grade stability — each pushes compute costs higher.
DeepSeek’s official V4 series, released in April, already pushed context length to 1M and began testing vision modes. These capabilities are great for developers, but they require sustained large-scale compute investment. If the company continues to deepen its enterprise services, compute will become not just a one-time training cost, but a recurring cost for inference and stable delivery.
Intensifying Talent Competition
DeepSeek has already lost some star researchers. Competition for top AI talent has reached a stage where idealism alone is no longer enough. Research culture attracts people, but compensation, equity, and future returns are equally critical for retaining core teams.
One direct effect of funding is to price employee stock options, making “growing with the company” more concrete and attractive.
Productization Is Urgent
The Information reports that DeepSeek employees have already begun promoting models to enterprises across various industries, hoping to turn technology into billable products and services.
A lab can focus solely on model metrics. But a heavy-asset AI company must care about customers, revenue, delivery, costs, and talent structure. DeepSeek is clearly moving from “strong models” to “products that sell.”
V4.1 Scheduled for June: Commercial Foundation Ready
The answer is already visible in the V4 series.
The V4 series launched on April 24, with two models: deepseek-v4-pro and deepseek-v4-flash, both supporting 1M context length. For enterprises, 1M context means long-document processing, large code-base analysis, multi-turn complex tasks, and complex workflow orchestration — all in one go. Combined with tool calling, JSON output, and other capabilities, the barriers to integrating into real business processes have been largely cleared.
This is the technical foundation for commercialization.
V4.1 is planned for June. Based on current disclosures, the new version will give enterprise users more tools, better support for the industry-standard MCP protocol, and is planned to handle images and audio simultaneously — true multimodal capabilities.
Funding is now driving the release pace. DeepSeek recently told some investors that the company plans to accelerate its model release cadence to something closer to industry norms.
From Idealism to Heavy-Asset Company
Looking back, DeepSeek’s most unique quality was its slowness. No external investors pushing for growth, no commercial team enforcing KPIs, and no premature revenue model constraints. Many believe this is precisely why DeepSeek was able to achieve technical breakthroughs.
But now, that model has reached a new stage.
DeepSeek is no longer just “a model team.” It is starting to look like a heavy-asset AI company. The keywords for such a company are compute, data centers, product teams, enterprise customers, stock options, and release cadence.
This will give DeepSeek more firepower, but also clearer commercial pressure.
From the perspective of the AI industry landscape, DeepSeek’s positioning has changed. Models, chips, domestic compute, and enterprise applications are no longer parallel tracks — they are beginning to intersect at DeepSeek.
Of course, nothing is final yet. Funding is still being negotiated, and DeepSeek has not officially commented on the reports. But the direction is clear enough: the DeepSeek that once refused fundraising and commercialization is now learning to do AI as a company.