DeepSeek nears $12B round with Tencent, CATL before IPO

DeepSeek is closing in on one of the largest private funding rounds in AI history — and it's not even close to done.

The Chinese AI startup is set to raise at least 80 billion yuan ($12 billion) in a new round, with Tencent and battery giant CATL among the largest backers, according to a Bloomberg report published on October 6, 2026 (UTC). Based on signed term sheets, the final tally could approach 100 billion yuan ($15 billion). The round has blown past DeepSeek's original target of roughly 50 billion yuan ($7.5 billion), a sign of just how hot investor demand has become for the company behind some of the most cost-efficient frontier models on the market.

The capital is positioned as pre-IPO funding. DeepSeek is targeting a public listing in early 2027, with the proceeds earmarked for compute, model development, and hiring. The round is expected to close in October 2026 (UTC), according to people familiar with the matter.

What changed

DeepSeek originally sought about 50 billion yuan at a valuation of roughly $75 billion. That number kept moving up as more investors piled in. The strong interest follows the success of DeepSeek's recent model releases — particularly V4-Flash, which has gained traction for delivering near-frontier performance at a fraction of the cost of Western rivals.

Tencent and CATL are contributing the largest shares. The pairing is unusual: one is China's dominant social media and gaming conglomerate, the other is the world's largest electric vehicle battery maker. Their joint backing signals that DeepSeek is seen as strategic infrastructure rather than just another AI lab.

Why it matters

A $12–15 billion round would put DeepSeek in rare company. For context, OpenAI's most recent completed round was $12.2 billion in March 2026 (UTC) at a $85.2 billion post-money valuation. Anthropic's last major private round was $6.5 billion. DeepSeek's raise, at a $75 billion valuation, is comparable in size to OpenAI's — and that's before the IPO.

The numbers tell a story about cost efficiency. DeepSeek has built its reputation on doing more with less. Its models run on older, less expensive chips and still compete with frontier systems that require the latest Nvidia hardware. That efficiency is exactly why investors are willing to write nine-figure checks: if DeepSeek can maintain its cost advantage as it scales, the unit economics of its API business could be dramatically better than Western competitors.

Tencent's involvement is the most strategically interesting piece. Tencent already has its own AI efforts through its Hunyuan model line. Backing DeepSeek — a competitor — suggests Tencent sees DeepSeek as infrastructure it wants access to, similar to how Microsoft's investment in OpenAI gave it privileged access to frontier models. It also gives Tencent a hedge: if DeepSeek becomes the dominant open-weight model provider, Tencent wants a seat at the table.

CATL's participation is harder to parse but no less significant. The battery maker has been expanding into energy storage and AI infrastructure, and data center power is becoming a bottleneck for AI training. CATL's backing could give DeepSeek an edge in energy-efficient compute — a real advantage as training runs grow larger and power costs become a bigger share of total expense.

The competitive picture

Sources: Bloomberg, company filings, prior reporting

DeepSeek's valuation of $75 billion is roughly 88% of OpenAI's $85.2 billion — despite DeepSeek having a fraction of the revenue and brand recognition. That gap reflects both the promise investors see in DeepSeek's efficiency and the premium Chinese tech assets command when they demonstrate global competitiveness.

Critical lens

There are real risks beneath the hype. DeepSeek operates under U.S. export controls that restrict its access to the most advanced Nvidia chips. While the company has proven it can work around these constraints with older hardware and clever optimization, there's a ceiling to how far that approach can scale. If the next generation of frontier models requires compute that DeepSeek simply cannot access, its cost advantage could narrow quickly.

The IPO timeline is also ambitious. A 2027 listing would require DeepSeek to demonstrate sustainable revenue growth, clear governance, and a path to profitability — all while navigating geopolitical tensions that could affect where it lists and who can invest. Hong Kong is the most likely venue, but even there, Chinese AI companies have faced volatile receptions.

The round's ballooning size — from $7.5B to potentially $15B — is itself a warning sign. When investor demand far outstrips a company's original ask, valuations can detach from fundamentals. DeepSeek's $75B valuation is already pricing in a lot of future success.

What to watch

• IPO filing: Watch for DeepSeek's formal IPO prospectus in late 2026 or early 2027. The revenue numbers and growth rates disclosed there will tell us whether the $75B valuation is justified.
• Next model release: DeepSeek's V4 line has been impressive, but the real test is whether it can close the gap with GPT-6 and Claude Fable on raw capability while maintaining its cost edge.
• Tencent integration: If Tencent starts integrating DeepSeek models into WeChat or its cloud services, that would be a major distribution win — and a signal that Tencent's own Hunyuan line is being deprioritized.
• Export control escalation: Any new U.S. restrictions on chip access or cloud services for Chinese AI firms could directly impact DeepSeek's roadmap.

DeepSeek's funding round is more than a capital raise. It's a bet that the future of AI isn't just about who has the most compute — it's about who can do the most with what they have. If that bet pays off, the $75 billion valuation will look cheap. If it doesn't, the investors who piled in at the top will be left holding a very expensive lesson.