DeepSeek hires first CFO as Chinese AI startup prepares for possible STAR Market IPO

DeepSeek is preparing to appoint Yan Wentao, a partner at GL Ventures, as its first chief financial officer, adding a senior dealmaker as the Chinese AI company moves toward a possible listing on Shanghai’s technology-focused STAR Market.

The appointment has not yet been formally announced by DeepSeek, so the exact start date and employment terms remain unconfirmed.

The broader direction is becoming clearer because DeepSeek has brought in CITIC Securities to prepare for a domestic IPO, with due diligence already underway even though a formal listing-advisory agreement has reportedly not yet been signed.

The company is targeting the beginning of the IPO process within 2026, while simultaneously pursuing another private financing round that could value it at approximately 500 billion yuan, or about $74 billion.

That represents another structural change for a company that until this year had largely been financed by founder Liang Wenfeng through High-Flyer, his quantitative investment group.

DeepSeek completed its first major external capital raise in June, raising approximately 50 billion yuan, or $7.4 billion, at a post-money valuation above $50 billion.

The proposed CFO appointment therefore connects three developments that have unfolded rapidly: external fundraising, institutional finance infrastructure and preparation for public markets.

ITEM	CURRENT POSITION
First CFO	Yan Wentao reportedly expected to join
Current role	Partner at GL Ventures
Previous DeepSeek CFO	None
Formal investor-relations organization	Previously limited
Potential exchange	Shanghai STAR Market
IPO adviser	CITIC Securities
Due diligence	Underway
Formal listing-advisory agreement	Reportedly not yet signed
IPO process target	Potentially during 2026
June external financing	Approximately 50B yuan / $7.4B
June valuation	More than $50B
Current valuation under discussion	Approximately 500B yuan / $74B
IPO valuation	Not determined
IPO size	Not determined

··········

DEEPSEEK IS BUILDING THE FINANCIAL INFRASTRUCTURE OF A PUBLIC COMPANY

Hiring a first CFO is considerably more consequential for DeepSeek than filling a routine executive position.

The company developed into one of China’s most important AI laboratories while operating with an unusually research-heavy organizational structure and relying heavily on financing connected to Liang Wenfeng.

A public company requires a much more developed financial architecture, and the CFO becomes responsible for financial reporting, internal controls, budgeting, capital allocation, audit preparation, investor communication, financing strategy and relationships with regulators and investment banks.

Those functions become particularly demanding during an IPO because an AI laboratory must convert highly technical expenditure into financial information that institutional investors can evaluate.

Training clusters become capital commitments.

Model development becomes research and development expenditure.

Inference capacity becomes an operating-cost question.

Employee equity becomes dilution.

Long-term semiconductor and cloud commitments become balance-sheet and liquidity considerations.

A dedicated CFO gives DeepSeek an executive responsible for connecting those operational realities to capital markets.

··········

YAN WENTAO BRINGS DEAL EXPERIENCE RATHER THAN A TRADITIONAL INTERNAL-FINANCE PROFILE

Yan Wentao’s background makes the reported appointment particularly relevant to the IPO strategy.

He has worked as a technology investor at GL Ventures, the venture-capital operation associated with Hillhouse, with exposure to major technology companies and AI businesses.

His investment experience has reportedly included companies such as ByteDance, MiniMax and Zhipu AI.

That profile is different from appointing a CFO primarily known for accounting or internal corporate finance.

DeepSeek appears to be adding someone familiar with technology valuations, private financing, investor negotiations and large capital-market transactions precisely as the company transitions between financing regimes.

GL Ventures itself is not reported to be an investor in DeepSeek.

The appointment therefore appears more closely connected to Yan’s financial and deal experience than to an existing shareholder relationship.

For DeepSeek, the immediate requirements extend beyond bookkeeping.

The company needs to establish how investors should value a frontier-model developer whose infrastructure requirements can rise by billions of dollars while its underlying products and pricing change considerably faster than those of conventional software companies.

··········

THE STAR MARKET WOULD KEEP DEEPSEEK'S IPO INSIDE CHINA

The Science and Technology Innovation Board, commonly known as the STAR Market, is part of the Shanghai Stock Exchange and was designed specifically for technologically advanced companies.

A domestic listing would give DeepSeek access to Chinese public equity without requiring the company to pursue a New York or Hong Kong listing.

That choice has strategic significance.

DeepSeek operates in an industry affected directly by US semiconductor export controls, Chinese technology policy, data regulation and growing geopolitical competition around advanced AI.

A STAR Market listing keeps the company's primary capital-market relationship inside mainland China.

It could also give domestic investors unusually direct exposure to one of the country's leading frontier-model developers.

The regulatory path remains incomplete because, although CITIC Securities is reportedly conducting preparatory work, DeepSeek has not publicly filed a prospectus establishing an IPO price, number of shares, offering size or definitive timetable.

IPO STEP	STATUS
Decide whether to explore listing	Yes
Select likely domestic market	STAR Market
Engage investment-bank support	CITIC Securities involved
Preliminary due diligence	Underway
Build finance / IR organization	First CFO reportedly being added
Formal listing preparation	Developing
Public prospectus	Not yet available
Final valuation	Not established
Share count / offer size	Not established
Completed IPO	No

DeepSeek remains in a preparatory phase: it is working toward a possible IPO, but it has not completed the regulatory process required to guarantee one.preparing for a possible IPO, rather than having completed the regulatory process required to guarantee one.

··········

DATA STUDIOS CALCULATES A ROUGHLY 42.5% VALUATION INCREASE FROM JULY'S IMPLIED LEVEL

DeepSeek’s evolving private valuation provides one way to measure how quickly its financing environment is changing.

A Chinese public filing examined in July implied a DeepSeek valuation of approximately 350.88 billion yuan.

The valuation now being discussed for the company’s latest financing is approximately 500 billion yuan.

Holding everything else constant, the increase would be:

(500B / 350.88B) − 1 = approximately 42.5%

VALUATION METRIC	VALUE
July implied valuation	350.88B yuan
Current reported valuation scenario	500B yuan
Absolute increase	149.12B yuan
Implied valuation increase	~42.5%
Valuation multiple	~1.43×

This is a Data Studios calculation using reported valuation figures, not an official DeepSeek IPO valuation.

The company has not announced what investors would be asked to pay in a public offering.

It nevertheless illustrates the speed of the revaluation.

A roughly 42.5% increase in private-market value over a period of approximately two months would mean that an eventual IPO starts from a substantially higher valuation base than DeepSeek’s first external financing.

That raises the financial threshold the company must eventually justify with growth, revenue, margins and technological leadership.

··········

DEEPSEEK HAS MOVED FROM FOUNDER FINANCING TO MULTIBILLION-DOLLAR EXTERNAL CAPITAL IN MONTHS

DeepSeek’s financing history is unusual for a frontier AI company of its scale.

For much of its development, the company relied on resources connected to Liang Wenfeng and High-Flyer.

That gave DeepSeek more freedom from conventional venture-capital fundraising than many competing laboratories.

The economics of frontier AI make that structure increasingly difficult to maintain indefinitely.

Training and serving advanced models requires investment in accelerators, memory, storage, networking, data centers, engineering and highly compensated research talent.

Agentic systems can increase inference demand further because a single user task may involve repeated model calls, tool executions and extended reasoning.

DeepSeek therefore began accepting significant external capital this year.

Its June financing of approximately 50 billion yuan represented a major transition from founder-backed research company toward institutionally financed technology group.

The reported IPO preparations extend that transition.

FINANCING STAGE	CAPITAL STRUCTURE
Early development	Primarily founder / High-Flyer financed
Expansion of frontier models	Increasing compute and talent requirements
June 2026	First major external financing, roughly 50B yuan
Current stage	Additional private fundraising at potentially ~500B yuan valuation
Next potential stage	STAR Market IPO
Long-term effect	Permanent access to broader equity capital

The CFO appointment therefore arrives at the exact point where DeepSeek’s capital structure is becoming materially more complicated.

··········

DEEPSEEK-V4.1-FLASH SHOWS WHY THE CAPITAL REQUIREMENTS ARE STILL EXPANDING

DeepSeek is pursuing its financial transformation while continuing an aggressive model-development cycle.

On September 10, the company released DeepSeek-V4.1-Flash, the newest member of its architecture family.

The model uses a 552-billion-parameter mixture-of-experts architecture, with only 8 billion parameters active during input processing and 16 billion during output generation.

DeepSeek also says the architecture substantially reduces KV-cache requirements, targeting one-quarter of the HBM and one-eighth of the SSD storage required by the previous generation.

Those optimizations illustrate an important part of DeepSeek’s economic strategy.

The company competes partly by finding architectural techniques that reduce the amount of hardware required for each unit of useful intelligence.

Lower inference cost can improve commercial economics.

It does not eliminate the need for large-scale capital.

Researching new architectures, training frontier models and building infrastructure capable of serving increasing usage still requires substantial investment.

A company can simultaneously become more efficient per token and more capital intensive in aggregate as the number of users, models and agent workloads expands.

That is one reason public-market access becomes attractive even for laboratories whose technical strategy emphasizes efficiency.

··········

AN IPO WOULD FORCE DEEPSEEK TO REVEAL ECONOMICS THAT ARE CURRENTLY LARGELY PRIVATE

DeepSeek currently publishes substantial information about its models.

Its public materials include model architectures, technical reports, pricing and performance information.

Its corporate financials remain much less visible.

An IPO would materially change that information gap because a public prospectus would be expected to provide investors with significantly more detail about the business behind the models.

The most important disclosures would likely include revenue and revenue growth, operating losses or profitability, cash balances and cash consumption, research and development expenditure, infrastructure and semiconductor commitments, customer concentration, API versus consumer revenue, employee compensation and equity, related-party relationships, ownership and voting control, risk from export restrictions, dependence on domestic semiconductor supply and regulatory exposure.

Those numbers would allow investors to compare DeepSeek with US frontier laboratories on economic rather than purely technical terms.

At present, model benchmarks can be compared relatively easily.

Capital efficiency is much harder to compare because the underlying corporate accounts remain private.

··········

THE IPO COULD TEST WHETHER DEEPSEEK'S EFFICIENCY ADVANTAGE TRANSLATES INTO BETTER FINANCIAL ECONOMICS

DeepSeek became globally significant partly because it challenged the assumption that top-tier AI performance required spending on the same scale as the largest US laboratories.

Its architecture and pricing strategy have repeatedly emphasized efficiency.

Public markets would eventually test whether that engineering advantage produces a durable financial advantage.

Lower computational cost can influence several variables:

EFFICIENCY DRIVER	FINANCIAL EFFECT
Lower active parameter count	Less compute per inference workload
Smaller KV cache	Lower memory and storage requirements
Higher throughput	More requests served per unit of infrastructure
Lower API pricing	Potentially greater usage but lower revenue per token
Efficient training	Lower marginal cost of model iteration
Open-weight availability	Wider adoption but different monetization model

Efficiency can lower DeepSeek's cost base, but the company also uses low pricing as a competitive weapon, which means part of the infrastructure saving can be transferred directly to customers instead of being retained as margin.

The public-market question is therefore not simply whether DeepSeek can run models cheaply.

It is whether it can convert that efficiency into sustainable gross margins, revenue growth and free cash flow while continuing to finance frontier research.

··········

A $74 BILLION PRIVATE VALUATION WOULD ALREADY PLACE A HIGH BURDEN ON FUTURE GROWTH

A 500 billion yuan valuation would make DeepSeek one of China’s most valuable private technology companies before an IPO.

That scale changes investor expectations.

At earlier stages, a frontier AI laboratory can be valued primarily around technology, talent and strategic optionality.

At tens of billions of dollars, the valuation increasingly needs a path toward very large commercial cash flows.

Consider a simple illustrative framework.

If a company valued at $74 billion eventually traded at:

REVENUE MULTIPLE	REVENUE IMPLIED BY $74B VALUE
30× revenue	~$2.47B
20× revenue	~$3.70B
15× revenue	~$4.93B
10× revenue	~$7.40B

These are Data Studios hypothetical calculations, not estimates of DeepSeek revenue or forecasts of its eventual market multiple.

They show how valuation and required commercial scale interact.

A higher valuation today places more future performance into the price investors are already paying.

If DeepSeek ultimately reaches the STAR Market at a materially higher valuation than the current private-market level, the revenue and cash-flow expectations embedded in the listing would rise again.

··········

FOUNDER CONTROL WILL BECOME ONE OF THE MOST IMPORTANT IPO QUESTIONS

DeepSeek’s transition to public markets also raises a governance issue.

Liang Wenfeng has historically exercised unusually strong control over the company and financed much of its early development himself.

The first external financing was reportedly structured in a way designed to preserve significant founder control.

An IPO introduces a much larger shareholder base.

That does not necessarily require Liang to surrender strategic authority.

Companies can preserve founder influence through ownership structure, voting arrangements, board composition and limitations on the percentage of equity sold to public investors.

The eventual prospectus should therefore be examined closely for Liang’s post-IPO ownership, voting rights, board appointment power, related-party structures and the rights granted to earlier investors.

Those details will matter because DeepSeek must balance two potentially competing objectives.

It wants access to much larger amounts of institutional capital.

It also appears determined to preserve the long-term technical direction that allowed the laboratory to develop differently from many venture-funded competitors.

··········

STAR MARKET CAPITAL COULD HELP DEEPSEEK COMPETE FOR CHIPS AND TALENT

Additional capital would primarily strengthen DeepSeek in the two scarce resources frontier laboratories compete for most aggressively: compute and specialized research talent.

DeepSeek faces both constraints intensely.

Access to the highest-end US accelerators is limited by export restrictions, increasing the strategic importance of domestic hardware optimization and alternative chip architectures.

At the same time, China’s major technology companies are competing aggressively for researchers capable of building frontier models.

Public equity could strengthen both sides of that competition because IPO proceeds can finance infrastructure while listed shares can also support employee compensation.

A publicly traded share can also become part of employee compensation.

That distinction becomes especially relevant when competitors can offer researchers large packages combining salary, bonuses and equity, because publicly traded shares provide a transparent market price and a clearer liquidity path than private-company stock.

For a talent-intensive AI company, equity liquidity can become part of the recruiting infrastructure.

··········

DEEPSEEK AND ANTHROPIC ARE MOVING TOWARD PUBLIC MARKETS THROUGH VERY DIFFERENT SYSTEMS

DeepSeek’s IPO preparation arrives as Anthropic is also moving toward a potential listing.

The comparison illustrates how frontier AI is becoming a capital-markets sector across geopolitical blocs.

Anthropic is preparing for a possible US listing and has reportedly selected Nasdaq.

DeepSeek is pursuing a domestic path through Shanghai’s STAR Market.

DIMENSION	DEEPSEEK	ANTHROPIC
Home market	China	United States
Likely exchange	Shanghai STAR Market	Nasdaq reportedly selected
Current IPO stage	Preparatory / due diligence	More advanced listing preparation
Model business	Frontier models, APIs, open models	Frontier Claude models, APIs, enterprise AI
Capital challenge	Compute, chips, talent	Compute, infrastructure, talent
Geopolitical constraint	US chip restrictions	US frontier-AI regulation and competition
Public-market significance	Major Chinese frontier-AI benchmark	Major US frontier-AI benchmark

The eventual valuations will not be directly comparable without adjusting for revenue, margins, ownership, regulation and market structure.

They would nevertheless give investors something that has historically been missing from frontier AI: public market prices for companies whose principal economic assets are the models themselves.

··········

THE CFO APPOINTMENT MAY BE THE CLEAREST SIGNAL YET THAT DEEPSEEK IS BECOMING A CAPITAL-MARKETS COMPANY

A model release can change DeepSeek’s technical position within days.

Hiring a first CFO changes something slower but equally important: the institutional structure surrounding the laboratory.

DeepSeek has moved in a short period from founder financing to a multibillion-dollar external capital raise, another potential valuation step-up, investment-bank due diligence and preparation for a domestic IPO.

The company still lacks a public prospectus, confirmed offering size and final listing timetable.

A STAR Market IPO therefore remains a possibility rather than a completed transaction.

The direction is becoming increasingly concrete as DeepSeek builds the financial organization required to support a business whose compute requirements, model ambitions and valuation have expanded beyond the structure of a privately financed research laboratory.

If Yan Wentao formally joins as CFO and the STAR Market process proceeds, DeepSeek’s next major transformation may occur in its capital structure as much as in its models.

··········

FOLLOW US FOR MORE.

DATA STUDIOS

datastudios.org