DeepSeek reportedly prepares for IPO, engages CITIC Securities as sponsor

Chinese AI startup DeepSeek (深度求索) is reportedly preparing for an initial public offering (IPO), potentially as early as 2027, as the company transitions from a technology-driven model to a capital-driven growth strategy.

IPO and Financing Progress

DeepSeek has reportedly engaged CITIC Securities for IPO preparations, though no formal advisory agreement has been signed. The company's listing preparations appear to be moving in parallel with a new funding round, which targets a valuation of 500 billion yuan (approximately $68.5 billion). Investors in this round are expected to include existing shareholders from the first financing round and waitlisted investors who missed the earlier opportunity.

In June, DeepSeek completed its first financing round, with total market estimates ranging from 50 billion to 51 billion yuan. Key contributors include:

• Liang Wenfeng (founder): approximately 20 billion yuan, the largest single investor
• Tencent: approximately 10 billion yuan
• CATL and Puquan Capital
• NetEase and JD.com

Reports from late August suggested DeepSeek is preparing for a potential IPO, with the earliest filing possible by the end of 2026 and a target listing in 2027.

Policy Tailwinds

Analysts believe DeepSeek's IPO preparations are closely tied to favorable policy developments. On June 17, 2026, the Shanghai Stock Exchange issued guidelines supporting AI companies that have not yet achieved stable profitability to list under the STAR Market's fifth set of standards. DeepSeek began IPO preparations less than three months after the guidelines were released, reflecting strong alignment between the company and regulatory policy.

Team Expansion and Engineering Focus

On September 7, 2026, Cui Tianyi, technical lead of DeepSeek's Harness team, posted recruitment information on social media, seeking approximately 150 global hires. Notably, these positions focus on server-side development and Agent elastic computing, rather than deep learning research.

Cui explained that as data volume, machine and container counts, training and evaluation tasks, Agent environments, user numbers, and request volumes all grow rapidly, the existing backend system faces increasing complexity. The team expansion aims to upgrade, maintain, and rewrite the backend system.

Commercialization Challenges

DeepSeek's reported revenue for the first seven months of 2026 was approximately 475 million yuan, about 10 times its full-year 2025 revenue. Gross margin for the period was 44.6%, with API business gross margin at 82.9%. Sources caution that these figures are not from official financial reports and should be verified against future audits or prospectuses.

Industry observers note that DeepSeek's core advantage lies in architectural innovation that lowers inference costs. However, low costs in technical reports do not automatically translate to high margins in financial statements. R&D trial-and-error, peak computing capacity redundancy, and enterprise services all impact the profit and loss statement.

Key commercialization challenges include:

• Pricing pressure: Lower prices expand the market but reduce revenue per token
• Revenue sustainability: Whether API calls reflect genuine production demand or subsidy-driven traffic
• Customer retention: Improving call volume, repurchase rates, and computing utilization to build a durable moat

Talent Retention Issues

Between the second half of 2025 and early 2026, several key technical staff members left DeepSeek. Competition from major tech companies for top AI researchers has pushed talent costs to new heights. The company may need to offer more attractive equity incentives post-IPO to retain talent.

Market and Valuation Concerns

While DeepSeek's pre-money valuation in the primary market has reached 500 billion yuan, its revenue for the first seven months was only about 475 million yuan, implying a price-to-sales ratio exceeding 1,000x. Whether this valuation can be accepted by the secondary market remains uncertain.

There are also reports of speculative investments through multi-layer SPV structures. Liang Wenfeng is personally reviewing the shareholder list to prevent complex ownership structures from hindering the IPO process.

Industry Implications

Analysts believe DeepSeek's IPO will reshape valuation logic across the AI industry:

• Capital concentration: Funds will flow to leading players
• Infrastructure layer: Must combine model efficiency with commercialization capability
• Application layer: Needs industry data and customer access
• Mid-tier companies: Those without technological differentiation will face the most pressure

Computing procurement evaluation will shift from celebrating "10,000-GPU clusters" to focusing on GPU utilization, unit inference costs, and real business output. Talent pricing will also change, with greater value placed on those who can connect models, inference systems, and customer needs on a single profit-and-loss statement.

Timeline and Next Steps

CITIC Securities has entered the due diligence phase, but no formal advisory agreement has been signed. No listing filing information for DeepSeek has been found with the Zhejiang Securities Regulatory Bureau. Given that the辅导备案 (guidance filing) process typically takes at least three months, filing within the year would be a tight timeline.

Overall, DeepSeek's IPO represents a milestone for the industry—moving from "proving we can build it" to "answering whether we can make money long-term." Continued attention to its commercialization data and regulatory progress is warranted.

(Source: China Business Journal)