DeepSeek’s Reported $74 Billion Funding Round: What the Numbers Actually Show

DeepSeek’s reported second funding round would be one of the most ambitious private financings in China’s AI sector: about 50 billion yuan, or nearly $8 billion, at a pre-money valuation close to 500 billion yuan ($74 billion). Reuters and Bloomberg reported that the round had resumed, with Monolith Management in talks to participate. 13

The central caveat is that these remain reports based on people familiar with confidential discussions. DeepSeek has not publicly confirmed that the financing has closed, that every named investor has committed, or that the reported financial figures are audited.

The reported financing terms

The proposed round is reported to target:

• Raise: about 50 billion yuan, roughly $7.4 billion to $8 billion depending on the exchange-rate conversion used in each report
• Pre-money valuation: close to 500 billion yuan, or approximately $74 billion
• Timing: investment agreements expected to be signed by the end of August 2026
• Purpose: funding infrastructure, chips and research ahead of a possible public listing, according to reporting about the company’s plans 1214

“Pre-money” means the valuation before the new investment is added. If the full 50 billion yuan were raised at that price, the post-money valuation would be approximately 550 billion yuan, assuming a straightforward equity transaction and no adjustments for deal structure.

Who is reportedly involved?

The investor picture is not fully settled and differs across reports. Reuters, citing Bloomberg, described Monolith Management as being in talks to participate; that wording does not establish Monolith as a confirmed investor in the reopened round. 13

Other reporting attributed to the South China Morning Post names Monolith, Shixiang Capital and CATL among investors associated with the earlier financing, while identifying CPE, Legend Capital, Stony Creek Capital, GigaDevice-backed funds and Hefei state-backed investment vehicles as potential participants in the new round. Those names should be treated as reported negotiations or deal possibilities, not confirmed commitments. 12

That distinction matters in a private financing. Investor discussions can change before definitive agreements are signed, and a reported target is not the same as capital already received.

What is known about DeepSeek’s earlier valuation?

There is limited public evidence about DeepSeek’s first external financing. A filing by Chinese luggage maker Anhui Korrun said a fund it backed deployed 2.9 billion yuan for an indirect 0.8265% stake in DeepSeek. The transaction implied a company valuation of about 350.88 billion yuan, or $51.82 billion. 4

Reuters said DeepSeek had not publicly announced the details of its maiden external round. As a result, reports describing an earlier 50 billion yuan fundraise or assigning a precise completed-round valuation should be handled cautiously. 4

The reported move from an implied valuation of about 350.88 billion yuan to a proposed 500 billion yuan pre-money valuation would represent a substantial increase, but the two figures may not be perfectly comparable. Deal terms, ownership calculations and the timing of each transaction can affect private-market valuations.

Revenue grew sharply—but losses continued

Reports attributed to people familiar with DeepSeek’s financial data put its January–July 2026 revenue at 475 million yuan, or about $70.7 million. That was roughly ten times the company’s full-year 2025 revenue, according to the same reporting.

The company reportedly recorded a 715 million yuan net loss during those seven months, compared with a 935 million yuan loss for all of 2025. In other words, revenue growth was rapid, but DeepSeek was still loss-making before the proposed financing.

The reported figures also put DeepSeek’s overall gross margin at 44.6% and its API business gross margin at 82.9%. The API figure may indicate strong direct economics for model access, but it should not be read as a company-wide profitability measure. Gross margin can exclude substantial costs such as research, infrastructure, sales, administration and other expenses.

Most importantly, these are private-company figures reported by people familiar with the business—not audited public-company disclosures. They may be useful indicators of momentum, but they are not equivalent to results filed with a stock exchange.

How DeepSeek compares with OpenAI and Anthropic

On reported revenue, DeepSeek remains much smaller than the leading U.S. AI companies. Its $70.7 million in revenue for seven months is not directly comparable with OpenAI’s reported $5.7 billion in first-quarter 2026 revenue or Anthropic’s reported $11.5 billion in second-quarter revenue. The periods, accounting bases and disclosure standards differ.

A simple annualization of DeepSeek’s January–July figure would produce roughly $121 million, but that is only a mechanical calculation—not management guidance or a forecast. It would still be far below the quarterly revenue figures reported for OpenAI and Anthropic.

The margin comparison is even less straightforward. Reports cited a 39% gross margin for OpenAI and a projected 63% full-year gross margin for Anthropic, compared with DeepSeek’s reported 82.9% API gross margin. These figures may measure different businesses and cost bases. DeepSeek’s API margin is a product-level metric, while the other figures may reflect broader company or forecast measures. They should not be treated as a like-for-like ranking of profitability.

OpenAI’s reported revenue later rose to $6.7 billion in the second quarter, while Anthropic was also reported to have exceeded $11.5 billion in that period. Those figures underline the enormous gap in current monetized scale—even though DeepSeek’s growth rate has reportedly been much faster from a smaller base.

What the $74 billion valuation implies

Using the reported figures, a $74 billion valuation would be very high relative to DeepSeek’s current revenue:

• About 1,047 times its reported seven-month revenue of $70.7 million
• About 612 times a simple annualization of that seven-month revenue

These are rough, non-standard multiples based on reported private-company data. They are not a substitute for a conventional valuation analysis, especially because AI companies can be valued on future contracts, strategic assets, model capabilities, ecosystem position and expected growth rather than current earnings alone.

Still, the basic conclusion is clear: the proposed valuation appears to depend heavily on expectations of exceptional future expansion and strategic importance. It is not supported by current net profitability, and it is difficult to justify using present revenue alone.

The possible Shanghai IPO

DeepSeek is reportedly preparing for a potential listing on Shanghai’s STAR Market. Reports say the company may file an IPO application as early as late 2026 and is targeting a 2027 market debut, with investment banks selected for the preparations. 12

That is a reported timetable, not a filed prospectus or an approved listing. The financing may be intended to strengthen DeepSeek’s balance sheet and fund the computing, chip and research requirements associated with a public-market strategy, but the eventual timing and terms remain uncertain.

Bottom line

The strongest source-backed version of the DeepSeek story is a fast-growing but still loss-making AI startup reportedly reopening a 50 billion yuan funding round at a 500 billion yuan pre-money valuation. Revenue reportedly reached 475 million yuan in the first seven months of 2026, and API economics appear strong on the reported figures. 112

What remains unproven is just as important: the round has not been publicly confirmed as closed, the full investor lineup is unsettled, the financial data is unaudited, and the IPO is still only a reported preparation. DeepSeek may be gaining the capital and strategic support to compete at a much larger scale—but the $74 billion valuation is primarily a bet on what the company could become, not a reflection of what its reported financials already demonstrate.