DeepSeek's Bigger Funding Target Tests Its Revenue Story

DeepSeek is considering raising as much as 100 billion yuan ($14.9 billion) in its latest funding round, twice its initial target, people familiar with the talks told CNBC. They said the company is seeking a valuation of about 500 billion yuan ($75 billion). Those are negotiating figures, not the proceeds and price of a closed deal. The round was initially expected to close by the end of August; one person told CNBC some prospective investors had wavered over the price.

For a founder, the first number buys time and options. For an employee holding equity, the second number can shape expectations about what that equity might someday pay. Neither tells either person how much cash DeepSeek keeps after paying to build and serve its models. The useful question is what buyers of this round expect to receive for the price: a share of growing model revenue, a claim on future technology, or a position ahead of a possible public listing.

The Earlier Round is the Useful Baseline

DeepSeek began taking outside capital this year. Its first external round closed in June and raised about 50 billion yuan; an investor filing related to that round implied a valuation of roughly 350 billion yuan, CNBC said. The new round began with a target equal to June's proceeds, while the valuation now being sought is higher than the earlier implied figure. The basis of the filing's implied valuation and that of the current asking price are unspecified, so those figures cannot establish a like-for-like rise in the value of existing shares.

DeepSeek has already completed a large outside raise. Considering a bigger second one suggests that investors and the company see a use for substantially more capital, but the size of a proposed round says nothing by itself about spending, cash balance or profitability. Rising AI capital needs and pressure to retain talent were context for opening up to external funding, CNBC said. Founder Liang Wenfeng told investors in July that he prioritized progress toward artificial general intelligence over maximizing profit, CNBC said. That may be an appealing mission to a patient investor; it also makes the terms of patience worth examining.

DeepSeek is reportedly vetting who gets to provide that capital. It has turned away private funds raised from individual investors and limited participation largely to government and corporate funds, people familiar with the talks told CNBC. CATL, Geely Auto, Monolith Management and Loyal Valley Capital had committed capital, one person told CNBC; the companies named in that account did not respond to its requests for comment. Reported commitments, access to the round and money received at closing are different stages of a transaction.

A large corporate or state-backed investor might value a strategic relationship alongside a financial return. That is an inference from the reported investor mix, not a description of anyone's negotiated rights. If such buyers accept a price that a return-focused fund rejects, the headline valuation may say as much about who is allowed into the round as it does about a common view of DeepSeek's earnings power. Conversely, the reported interest offers a strong case for DeepSeek's bargaining position: it can seek capital while choosing its counterparties, even as some prospects question the price.

The choice of backers could also affect how outsiders interpret the eventual round. An investor buying access to technology or a long-term relationship might assess the deal differently from a fund seeking a financial exit. A founder choosing among those offers would have to weigh the price against any obligations that came with it. The reported investor mix makes that trade-off a question to ask, rather than an answer about DeepSeek's contracts.

Revenue Has a Clock Attached

DeepSeek's annualized revenue run rate reached $1 billion, up from less than $500 million a few months earlier, two people with direct knowledge told The Information. Liang shared the latest figure with investors. The increase reflected a recent price hike as well as continued model popularity, those people said. DeepSeek was aiming to finish a 50-billion-yuan round by the end of October when The Information published its account in September. That target and the October discussion of a round as large as 100 billion yuan describe negotiations at different points, rather than two completed raises.

A run rate extends a recent pace over a year. It gives an investor a way to ask whether current sales could support a much larger business. It is not twelve months of realized revenue. Customer retention and the cost of serving customers would help test whether the business could sustain its economics.

The commercial and research cases could coexist. Higher revenue could finance more development, while an ambitious research program could consume cash faster than the new revenue arrives. The reported run rate reflects a higher revenue pace; the figures do not separate the effects of price increases and usage or establish whether demand is durably expanding. A larger funding target could reflect a desire to finance research ahead of future receipts, but the target alone cannot identify DeepSeek's spending or its investors' motives. The eventual round terms, if agreed, will show what investors were willing to pay to take those risks.

A valuation headline also hides the cap table. An employee trying to estimate an equity payout would need a prospective sale price, their ownership after dilution, and any contractual claims that take priority over their shares. Liquidation-preference math depends on the actual share-class rights and amounts in those agreements; multiplying a grant's ownership percentage by the reported company valuation cannot supply a payout. DeepSeek's proposed price does not specify those terms, so an employee weighing an equity grant would need to see them rather than treat the fundraising figure as a cash-out forecast.

For example, if preferred shareholders have a right to recover invested capital before ordinary shareholders receive sale proceeds, that claim comes out before an employee's common shares participate. The order of claims, the amount invested and any right to share in remaining proceeds would determine the outcome. A further raise could also dilute an existing grant. A proposed valuation answers none of those questions, even if an employee can see a percentage ownership figure on a grant document.

What a Listing Would Put on the Page

MiniMax offers a bounded comparison. The Chinese model developer listed in Hong Kong in January. Its prospectus figures put revenue at $53.4 million and its net loss at $512 million for the nine months ended September 30, 2025, CNBC said. MiniMax said most of its revenue came from subscriptions and in-app purchases and described monetization as nascent. Those are figures for a defined accounting period alongside a loss, the sort of paired measures that let readers ask what growth costs.

MiniMax's products, revenue mix and accounting period differ from DeepSeek's reported run-rate measure. Its loss cannot be used to estimate DeepSeek's, and DeepSeek's annualized pace cannot be used to rank the businesses. The useful parallel is disclosure: a listed company's reported sales and net result can be read together, while a private-round price offers neither measure.

DeepSeek has tapped CITIC Securities to prepare for a possible Shanghai STAR Market IPO, Reuters said in September. Preparation leaves both the timing and the decision to list open. If DeepSeek does reach public markets, revenue for a defined period alongside its costs and share terms would let employees and investors test the story now being negotiated privately. Until then, the proposed 100-billion-yuan raise measures financing ambition, while the reported revenue pace measures recent sales. The space between them is where the investment case has to do its work.