梁文锋出资200亿！DeepSeek首轮创纪录融资500亿，V4.1定档6月

量子位

量子位2026年5月9日

Liang Wenfeng Invests 20 Billion! DeepSeek Raises Record-Breaking 50 Billion in First Round, V4.1 Scheduled for June

8.5Score

TL;DR · AI Summary

DeepSeek secured its largest funding round ever—up to 50 billion RMB (approx. $7B), with founder Liang Wenfeng personally contributing 20 billion (40%). V4.1 is set for release in June, signaling a shift from idealistic lab to capital-intensive AI company.

Key Takeaways

Liang Wenfeng personally invested up to 20 billion RMB, accounting for 40% of th
DeepSeek's valuation surged fivefold in just three weeks—from $10B to $50B.
V4.1 will support image and audio processing, enabling enterprise adoption throu

Outline

Jump quickly between sections.

§Introduction: Record-breaking funding sparks industry attention
DeepSeek’s first-round funding aims at 50 billion RMB, potentially becoming the largest ever for a Chinese LLM company if finalized.
·Core mechanism: Valuation surge and Liang Wenfeng’s investment
DeepSeek’s valuation increased fivefold in three weeks—from $10B to $50B—with Liang Wenfeng’s personal contribution as the key driver.
·Transition drivers: Pressure from compute, talent, and productization
Rising compute demands, top researcher attrition, and commercialization needs pushed DeepSeek to accelerate fundraising and timelines.
·V4.1 roadmap: Multimodal capabilities enable commercialization
V4.1 will support image and audio inputs, enhancing enterprise toolchains and laying technical groundwork for real-world deployment.
·Future challenges: From lab to capital-intensive AI firm
Post-funding, DeepSeek must balance R&D with revenue goals, facing clearer commercial pressures and operational complexity.

Mindmap

See how the topics connect at a glance.

查看大纲文本（无障碍 / 无 JS 友好）

DeepSeek融资与转型
- 融资规模
  - 500亿人民币目标
  - 梁文锋出资200亿（40%）
- 估值变化
  - 3周内翻5倍
  - 从100亿→500亿美元
- 产品演进
  - V4系列支持1M上下文
  - V4.1支持图像+音频
- 战略转型
  - 从实验室转向重资产公司
  - 商业化节奏加快

Highlights

Key sentences worth saving and sharing.

Liang Wenfeng personally invested up to 20 billion RMB, accounting for 40% of the total fundraising target.
— Paragraph 2
⬇︎ 下载 PNG 𝕏 分享到 X
DeepSeek’s valuation surged fivefold in just three weeks—from $10B to $50B.
— Paragraph 3
⬇︎ 下载 PNG 𝕏 分享到 X
V4.1 will support image and audio processing, while optimizing MCP protocol for industry-wide compatibility.
— Paragraph 5
⬇︎ 下载 PNG 𝕏 分享到 X

#DeepSeek#AI Funding#Large Language Models#Commercialization#China Tech

Open original article

Liang Wenfeng Invests 20 Billion! DeepSeek Raises Record-Breaking $70 Billion in First Round, V4.1 Scheduled for June – QbitAI

[](https://www.qbitai.com/)

[](javascript:void(0))

Scan to Follow QbitAI

[](https://weibo.com/qbitai?is_all=1)

< img id="wx_img" src="https://www.qbitai.com/wp-content/uploads/imgs/qbitai-logo-1.png" width="400" height="400">

Liang Wenfeng Invests 20 Billion! DeepSeek Raises Record-Breaking $70 Billion in First Round, V4.1 Scheduled for June

_[Meng Chen](https://www.qbitai.com/author/mengchen "Posted by Meng Chen")_ 2026-05-09 10:08:59 Source: QbitAI

If finalized, this will be the largest funding round ever for a Chinese large model company.

By Meng Chen from Aofei Temple
QbitAI | Public Account QbitAI

In just 21 days, DeepSeek’s valuation surged to $500 billion!

DeepSeek V4.1 is reportedly scheduled for June, accelerating everything.

The biggest check in this round may not come from VCs or internet giants—but from Liang Wenfeng himself.

According to The Information, Liang Wenfeng personally invested up to 20 billion RMB, accounting for 40% of the total planned fundraising of 50 billion RMB.

If completed, this would become the largest single funding round in the history of Chinese large model companies.

This company was once most famously known for being “no financing, no commercialization, no roadshows.”

An AI lab that always prioritized research suddenly seeks a record-breaking sum of money.

Within just three weeks, DeepSeek’s valuation has been revised four times:

Early April 2026: ~$14 billion, DeepSeek launched its first funding round;
April 22, 2026: >$28 billion, talks with Tencent, Alibaba, and others;
May 6, 2026: ~$63 billion, National Integrated Circuit Industry Investment Fund expressed interest in leading;
Early May 2026: Reports suggest final valuation could reach $70 billion.

The shift has begun.

When Idealists Pay the Price

DeepSeek’s past story carries a strong tinge of technical idealism.

A team incubated from a quant fund, focused on building models without rushing into commercialization or investor meetings—quietly pushing forward.

This stance itself became a talking point.

But by 2026, this romanticism collided with three harsh realities.

First, compute power.

Frontier models are no longer as simple as “writing a paper and training one.” Capabilities like reasoning, agent functionality, ultra-long context, and enterprise-grade stability continue to push compute demands upward.

DeepSeek’s official V4 series released in April already extended context length to 1 million tokens, and began testing multimodal capabilities.

These features are developer-friendly but require substantial compute behind them.

If the company continues toward enterprise services, compute won’t just be a high cost during training—it’ll also become part of ongoing inference and stable delivery.

Second, talent.

DeepSeek has already lost several star researchers—Guo Daya, Wang Bingxuan, Wei Haoran—who all moved to positions offering higher compensation.

At this stage, top AI researchers can’t be retained solely through idealism.

Research culture attracts people, but salaries, equity, and future returns must also be considered.

Funding here helps price employee stock options, making “growing together with the company” more concrete.

Third, productization.

DeepSeek is now emphasizing it can’t stay forever at “just having powerful models.”

According to The Information, DeepSeek employees have started promoting models to enterprises across industries, aiming to turn technology into paid products and services.

A lab can focus only on model metrics.

A heavy-asset AI company must care about customers, revenue, delivery, costs, and talent structure.

Funding solves part of the money problem.

But once capital arrives, the next question becomes: Where will DeepSeek invest these funds?

After V4, Commercialization Gets Real

The answer is already emerging in DeepSeek V4 and V4.1.

V4, launched on April 24, includes two models: deepseek-v4-pro and deepseek-v4-flash, both supporting 1 million token context.

For enterprises, 1 million tokens means long documents, codebases, multi-turn tasks, and complex workflows. Combined with tool calling and JSON output, barriers to integrating real business processes are now cleared.

This is the foundation for commercialization.

According to The Information, DeepSeek plans to release V4.1 in June.

The new version will offer more tools for enterprise users, better support for industry-standard MCP protocols, and the ability to process both images and audio simultaneously.

Funding is speeding up release schedules.

Previously, slower releases were acceptable—a purely idealistic lab could wait until fully satisfied before launching.

But recently, DeepSeek told some investors it plans to accelerate model releases closer to industry norms.

Looking back, this funding round appears to be about money—but fundamentally, it's about a transformation in company structure.

DeepSeek previously had no external investors pushing growth, no commercial teams setting KPIs, and wasn’t constrained by early revenue models.

Many believe this freedom was one reason DeepSeek achieved breakthroughs.

Now, this model has reached a new phase.

DeepSeek is no longer just “a model team”—it’s becoming a heavy-asset AI company.

Key traits of such companies: compute, data centers, product teams, enterprise clients, equity, and release cadence.

These elements give DeepSeek stronger firepower.

They also bring clearer commercial pressure.

From an industry perspective, DeepSeek’s positioning has changed. Models, chips, domestic compute, and enterprise applications are no longer parallel lines—they’re intersecting within DeepSeek.

That’s why this funding round is receiving so much attention.

Large model competition is shifting from lightweight model races to a stage where compute, talent, capital, and commercialization all come to the table together.

Of course, nothing is certain yet.

The funding is still under negotiation, and DeepSeek hasn’t commented on related reports.

But the direction is clear enough.

Once a company that refused to raise funds or commercialize, DeepSeek is now learning how to do AI like a proper business.

References:

[1]

https://www.theinformation.com/articles/deepseek-raise-7-billion-startup-plots-revenue-efforts

_© Copyright. No part of this article may be reproduced or used in any form without prior written permission. Violators will be prosecuted._

Deepseek

![Image 6[Meng Chen](https://www.qbitai.com/author/mengchen "Posted by Meng Chen")](https://www.qbitai.com/2026/05/414432.html#)

[NVIDIA Reconsiders AI TCO: Why Cost Per Token Is the Only Important Metric](https://www.qbitai.com/2026/05/413604.html "NVIDIA Reconsiders AI TCO: Why Cost Per Token Is the Only Important Metric")_2026-05-07_
[Musk Sells All 220,000 GPUs to Claude: 5-Hour Limit Doubled, Joint Space Compute Project Announced](https://www.qbitai.com/2026/05/413569.html "Musk Sells All 220,000 GPUs to Claude: 5-Hour Limit Doubled, Joint Space Compute Project Announced")_2026-05-07_
[ChatGPT Free Model Just Upgraded: Hallucinations Cut in Half / Better Memory / More Concise Answers](https://www.qbitai.com/2026/05/412995.html "ChatGPT Free Model Just Upgraded: Hallucinations Cut in Half / Better Memory / More Concise Answers")_2026-05-06_
[Apple’s Official App Accidentally Bundled Claude.md—Even Big Companies Do Vibe Coding?](https://www.qbitai.com/2026/05/412713.html "Apple’s Official App Accidentally Bundled Claude.md—Even Big Companies Do Vibe Coding?")_2026-05-02_

Scan to Share on Moments

[](https://service.weibo.com/share/share.php?url=https://www.qbitai.com/2026/05/414432.html&title=Li%C3%A4ng+Wenfeng+Invests+20+Billion%EF%BC%81DeepSeek+First+Round+Breaks+Record+Financing+50+Billion%EF%BC%8CV4.1+Scheduled+for+June&appkey=4017757111&searchPic=true&ralateUid=6105753431 "Share on Weibo")[](https://www.qbitai.com/2026/05/414432.html)