#DeepSeek将完成1000亿元融资 DeepSeek’s ¥100 billion financing: AI’s “arms race” has reached the eve of its IPO—but the fervor in private markets may not be embraced by public markets

DeepSeek’s second financing round is approaching ¥100 billion (original target: ¥50 billion, led by Tencent and CATL), with a pre-money valuation of around ¥500 billion (approximately $74 billion)—up roughly sevenfold in less than four months from its initial valuation of about $10 billion in June; it has also hired CITIC Securities to prepare for a STAR Market IPO, targeting a filing in late 2026 or early 2027 and a listing in 2027.
My view: This financing is a landmark event in the “AI capital arms race”—more money = more computing power = a bigger story, and the logic holds; but the combination of “a sevenfold valuation increase in seven months + the eve of an IPO” is precisely the prelude to a collision between “private-market sentiment” and “public-market reality.” For DeepSeek, it is ammunition; for the industry, a signal; for ordinary investors, this is someone else’s private-market carnival, while your exam begins on the day of the IPO.

I. What is this ¥100 billion buying
① Buying time for “computing-power autonomy.” Liang Wenfeng himself has said that Chinese AI remains heavily dependent on NVIDIA chips—the most practical uses of the financing are stockpiling chips (NVIDIA) and betting on domestic computing power. 21st Century Business Herald directly called this a “high-stakes bet on domestic computing-power pretraining.” In the end, the AI battle is a war of computing-power attrition; whoever has more money survives longer.
② Buying an “IPO pass.” This round is essentially pre-IPO bridge financing—allowing the company to knock on the STAR Market’s door with “¥100 billion in cash reserves + backing from major corporations.” Tencent and CATL’s participation is not just about money; it is strategic credibility.
③ Buying the “gap narrative.” Bloomberg Intelligence estimates that DeepSeek V4.1 Flash has narrowed the gap between China’s and the United States’ top models to around 3%—the story that “Chinese AI is catching up” is the fuel for the sevenfold valuation increase.

II. But the market requires extreme caution
① A sevenfold increase in seven months = sentiment-driven pricing, not performance-driven pricing. The valuation rose from $10 billion to $74 billion on the back of scarcity—“you can’t get a stake”—and AI fervor. Private-market valuations are set by “those who want to get in”; public-market prices are set by “those who want to get out.”
② A cautionary example: MiniMax. Another leading large-model company, MiniMax saw its share price plunge 80% after listing—“Who will take over in the large-model IPO wave?” is already a question the market is asking. Whether the market will accept DeepSeek’s $74 billion valuation at listing is another matter.
③ The dilution and profitability math. New shareholders in this round will receive 17–20% (compared with only 13% in the first round)—diluting the founding team; meanwhile, DeepSeek’s business model—low-priced APIs—is known for its “value for money.” Whether its revenue can support a $74 billion valuation will depend on the prospectus.

DeepSeek’s ¥100 billion financing proves that the “Chinese AI arms race” has entered the “eve of the IPO”—with ample ammunition, backing from major corporations, and a complete gap narrative; but the sevenfold valuation increase in seven months must ultimately pass the “public-market” test. For DeepSeek, it is a “high-stakes gamble”; for the industry, a “signal”; for retail investors, “wait until the IPO and see.”
The rules of the AI world are much like those of the crypto world—private markets are always feverish, while public markets are always clear-eyed. Raising ¥100 billion is an achievement; avoiding a post-listing drop below the issue price is the ultimate achievement. No matter how loudly the story is told, it only counts if someone is willing to take over.