DeepSeek reportedly secured ¥800 billion in new funding and is preparing for an IPO, marking a strategic shift toward capital mobilization for frontier AI model development.

On October 6, Bloomberg reported, citing sources familiar with the matter, that DeepSeek is close to locking in at least 80 billion yuan (approximately 12 billion dollars) in its latest funding round—far exceeding the company's initial target of roughly 50 billion yuan—paving the way for a potentially significant IPO in early 2027.

What has caught market attention even more is that, according to signed investment term sheets, the final funding round could approach 100 billion yuan, while the company's initial target valuation stood at approximately 500 billion yuan.

Informed sources reveal that Contemporary Amperex Technology Co. Limited (CATL) and Tencent have committed the highest investment amounts. The financing process is drawing to a close, though the relevant transaction has not been finalized and specific terms may still change.

This is not the first time DeepSeek has rewritten its financing scale. To understand the origins and context of this 80 billion yuan round, we need to go back to April of this year.

**From Never Needing Capital to Embracing It**

DeepSeek was founded on July 17, 2023, by Hexagon Quantitative, a renowned quantitative asset management giant. From its first day, founder Liang Wenfeng established an iron rule: no external financing, no equity dilution, and no being bound by anyone else's commercialization timeline.

This confidence stemmed from Hexagon Quantitative. In 2025, Hexagon achieved an average return rate of 56.6 percent, managed assets exceeding 700 billion yuan, and has supported DeepSeek's R&D spending with tens of millions in annual profits.

However, multiple pressures converged in 2026, forcing Liang Wenfeng to make a strategic adjustment.

The first pressure came from competitors. Stanford University's 2026 AI Index Report showed that the performance gap between American top-tier models and China's strongest competitors had narrowed to just 2.7 percentage points, with the marginal cost of catching up growing steeper. Competitor OpenAI had just completed 400 billion dollars in financing, further widening the capital gap.

The second pressure came from talent attrition. According to public information, at least five core research personnel from DeepSeek left the company in succession. The "AI prodigy" Luo Fuli was recruited away by Xiaomi to lead AI operations; 95-generation researcher Guo Daya switched to ByteDance; DeepSeek LLM core author Wang Bingxuan was recruited by Tencent. These departures covered four core technical areas: foundational models, inference, OCR, and multimodality.

The third pressure came from R&D of the next-generation flagship model V4. Model parameters were leaping from the hundred-billion scale to the trillion scale, with costs for computing expansion and data annotation growing exponentially. Hexagon Quantitative's internal funding alone could no longer support this.

Under the weight of multiple pressures, in April 2026, DeepSeek formally initiated its first external financing since its founding, targeting a valuation of no less than 10 billion dollars and planning to raise at least 3 billion dollars.

On April 28, DeepSeek completed business registration changes, increasing registered capital from 10 million to 15 million yuan. Liang Wenfeng's registered capital commitment increased from 100,000 yuan to 5.1 million yuan, raising his direct equity stake from 1 percent to 34 percent, with ultimate beneficial ownership of 84.29 percent and voting rights of 100 percent.

By June, the first round of financing was formally completed, raising a total of 51 billion yuan with a post-investment valuation of nearly 400 billion yuan, setting the record at that time for the largest single-round financing in China's AI industry.

The investor consortium for this round was exceptionally strong:

Founder Liang Wenfeng personally contributed approximately 20 billion yuan, making him the largest single investor at 40 percent of total funds raised;

Tencent contributed approximately 10 billion yuan;

The CATL ecosystem contributed approximately 5 billion yuan, including CATL and Puquan Capital;

NetEase, JD.com, Monolith Capital, and IDG Capital each contributed approximately 3 billion yuan;

Zhengxin Valley Investments and Shixiang Technology each contributed approximately 1.5 billion yuan;

The National Artificial Intelligence Industry Investment Fund participated strategically, holding approximately 0.28 percent equity.

The transaction employed a limited partnership structure, with external investors generally lacking voting rights and their shares locked for five years. Liang Wenfeng retained nearly 100 percent voting control.

Notably, after the first round of financing closed, Liang Wenfeng's net worth more than doubled to 36 billion dollars, making him the AI model developer founder with the highest personal wealth globally.

**The Leap from 50 Billion to Nearly 100 Billion**

Following the completion of the first round, DeepSeek rapidly initiated its second round in mid-July while simultaneously beginning IPO preparations.

This round's initial target was approximately 50 billion yuan in financing with a target valuation of approximately 500 billion yuan.

However, the successful release of the latest AI models significantly boosted investor sentiment, causing the fundraising scale to be repeatedly raised.

Product-side breakthroughs are the core foundation for this capital oversubscription. On April 24, DeepSeek released and open-sourced the V4 series preview under the MIT license; by late July to early August, V4-Flash officially launched; on September 10, DeepSeek released the all-new V4.1-Flash architecture with native multimodal vision understanding capabilities, achieving a GPQA Diamond score of 90.9.

Bloomberg Industry Research reported on October 5 that since the V4.1 Flash launch, the benchmark testing score gap between top-tier U.S. and Chinese models had narrowed to approximately 3 percent, compared to approximately 9 percent in May this year and approximately 15 percent at the start of the year.

In September, V4.1 Flash ranked sixth on the LiveBench leaderboard with a score of 81.1, becoming the highest-ranked Chinese model since DeepSeek's breakthrough with R1 in 2025, with performance already comparable to leading systems from Anthropic and OpenAI, while at significantly lower calling prices.

In the second round, Tencent and CATL continued to increase their investment as lead investors. Tencent aims to integrate AI capabilities across its various platforms; CATL sees the increasingly profitable business of supplying components to the AI industry and delivering to data centers.

As of now, representatives from CATL declined to comment, Tencent has not responded, and DeepSeek has not replied to requests for comment.

However, financing is not the endpoint—an IPO is.

According to previous Reuters reporting, DeepSeek has engaged China International Capital Corporation Limited (CICC) as its sponsor institution and is preparing for listing on the Shanghai Stock Exchange's STAR Market, with CICC having engaged with DeepSeek and entered the due diligence stage.

The company targets submitting its IPO application by the end of 2026 at the earliest, with formal listing in early 2027.

The opening of a policy window has provided critical support for DeepSeek's IPO. On June 17, 2026, the Shanghai Stock Exchange released guidance on "Application of the Fifth Set of STAR Market Listing Standards for AI Large Language Model Enterprises," specifically targeting companies with self-developed large models and unprofitable AI enterprises.

This policy has opened an A-share IPO channel for AI enterprises like DeepSeek that have not yet achieved stable profitability.

However, DeepSeek's commercialization still faces challenges. In the first seven months of 2026, DeepSeek generated 475 million yuan in revenue, achieving a tenfold increase compared to full-year 2025 revenue, but during the same period still incurred a net loss of 715 million yuan.

Industry analysts in the capital markets believe that based on DeepSeek's current 500 billion yuan valuation, the overall market capitalization following listing is expected to fall between 1.5 trillion and 2.5 trillion yuan.

Following the completion of financing, one of DeepSeek's primary tasks is team expansion.

During the National Day period, DeepSeek Harness head Cui Tianyi posted on social media that a substantial portion of his daily work time is devoted to recruiting and interviewing for DeepSeek, with target positions concentrated in server-side development and Agent elastic computing engineering directions, planning to recruit approximately 150 people globally.

**Betting on the Huawei Ecosystem**

Notably, DeepSeek's financing capital will be heavily invested in artificial intelligence infrastructure construction.

According to reports, DeepSeek plans to build large-scale data centers in Inner Mongolia, deploying at least 160,000 units of Huawei high-end AI accelerators, which, if fully scaled, would become one of the largest known Huawei AI chip clusters.

Meanwhile, DeepSeek has formally open-sourced infrastructure components for Huawei's Ascend computing platform, including compilation tools, computing libraries, and distributed communication libraries, further deepening its binding with the Huawei ecosystem.

From "rejecting external financing" to billion-yuan financing sprinting toward IPO, DeepSeek's transformation is undeniably significant.

The AGI track is a typical "capital bottomless pit," where one must continuously pour money into computing power, chips, data centers, and talent.

As the U.S.-China AI competition enters a fever pitch, DeepSeek's choice to embrace capital may not be a compromise, but rather another form of acceleration.