DeepSeek nears a record $12 billion round from Tencent and CATL ahead of a 2027 IPO

DeepSeek, the Chinese lab that shocked Silicon Valley in 2025 with a cheap model that kept up with the best American systems, is about to collect one of the largest funding rounds any AI company outside the United States has ever raised. According to Bloomberg, which cited people familiar with the matter on Tuesday, 6 October 2026, DeepSeek is close to securing at least 80 billion yuan, roughly 12 billion US dollars. Battery giant CATL and WeChat owner Tencent are among the biggest backers. Once the round closes, DeepSeek plans to restructure itself for an initial public offering in early 2027.

The headline number is big on its own. What makes it more interesting is how far it overshot the plan. DeepSeek went into this round hoping for about 50 billion yuan. Investor demand was so strong that the round is already at 80 billion yuan, and based on the term sheets that have been signed, the final total could come close to 100 billion yuan. In plain terms, DeepSeek could end up with almost twice the money it asked for.

What Bloomberg reported

The report comes from Bloomberg News and was picked up on Tuesday by outlets such as The Straits Times, CNBC TV18 and The Decoder. Here are the points that are on the record so far.

• Size. DeepSeek is close to raising at least 80 billion yuan, about 12 billion US dollars. The final figure could approach 100 billion yuan.
• Original target. The company first sought about 50 billion yuan. Interest grew after the successful release of its latest model, V4 Flash.
• Investors. CATL, the world's largest maker of electric car batteries, and Tencent, the company behind WeChat, have committed some of the largest amounts.
• Valuation. DeepSeek had originally targeted a valuation of about 500 billion yuan for this financing, which is roughly 75 billion US dollars.
• Next step. After the round closes, DeepSeek is expected to restructure the company to prepare for a stock market listing in early 2027.
• Compute. DeepSeek plans to deploy at least 160,000 of Huawei's top AI accelerators in a huge data center it is building in Inner Mongolia.

None of this has been announced by DeepSeek itself. The people who spoke to Bloomberg said details could still change in the final talks. A CATL representative declined to comment, Tencent had no immediate comment, and DeepSeek did not answer an email because of a public holiday in China. So treat the exact numbers as a strong report, not a signed and closed deal.

Why the round grew so fast

The short answer is V4 Flash. According to the Bloomberg report, the release of DeepSeek's latest model reset expectations for what you can get for your money compared with the leaders, Anthropic and OpenAI. That is exactly the story that made DeepSeek famous in the first place. In 2025 it released a model that matched the abilities of leading Silicon Valley systems at a fraction of the cost, and the market reaction was so strong that it briefly shook confidence in the whole American AI trade.

Investors have learned since then that DeepSeek is one of a small group of companies at the center of China's effort to compete with the United States on AI. When such a company opens its books, demand tends to outrun supply. The jump from a 50 billion yuan target to at least 80 billion yuan shows that the demand is still there, even after a year in which money poured into AI everywhere.

There is also a simple practical reason. Frontier AI is extremely expensive. Training and serving top models needs huge clusters of chips, power and data centers, and every lab that wants to stay near the front needs fresh capital again and again. DeepSeek is no exception, however efficient its models are.

Who is paying, and why it makes sense for them

The two largest backers say a lot about where Chinese AI money comes from.

Tencent already knows DeepSeek well. In the previous round it made the biggest single investment, 10 billion yuan. Tencent wants to build AI into its platforms, and WeChat alone reaches a huge share of China's population. Having a close relationship with one of the country's strongest model makers is valuable for a company like that, whether it uses DeepSeek models directly or simply wants a seat at the table.

CATL looks like the more surprising name at first. It is a battery company, not a software firm. But data centers are an increasingly lucrative market for energy storage and other components, and the AI boom has turned electricity into one of the main limits on growth. A battery maker that invests in an AI lab with a giant data center plan is effectively buying insight into one of its future biggest customer groups.

The numbers in context

It helps to put the round next to DeepSeek's own history. The company grew out of High-Flyer, the hedge fund run by founder Liang Wenfeng, which paid for most of its early work. This summer DeepSeek closed its first round of external funding. In that previous round it raised about 50 billion yuan at a valuation of roughly 350 billion yuan, and Liang put money in again himself. According to Bloomberg, his personal net worth more than doubled to around 36 billion US dollars when that round closed.

Now, only a few months later, the company is raising more money again, with an original valuation target of about 500 billion yuan. That would be a step up of roughly 40 percent in valuation in a matter of months, and the amount raised would grow from about 50 billion yuan to somewhere between 80 and 100 billion yuan.

DeepSeek is not alone. Its rival Moonshot AI, the maker of the Kimi models, has closed its final private round at a valuation of about 50 billion US dollars, according to a separate Bloomberg report on Tuesday that was summarised by The Standard in Hong Kong. Moonshot is considering raising up to 5 billion US dollars in a Hong Kong listing in the first quarter of 2027. Its valuation in the round before this one, earlier in the summer, was 31.5 billion US dollars, and its annual recurring revenue is expected to grow from about 1 billion to 2 billion US dollars by December. Two of China's best known AI labs heading for the stock market at the same time is a clear sign that this market is entering a new phase.

160,000 Huawei chips in Inner Mongolia

The most important detail for the technology race may be the hardware plan. DeepSeek wants to deploy at least 160,000 of Huawei's top accelerators at a large data center it is building in Inner Mongolia. If that happens, it could become one of the largest known clusters of Huawei AI chips anywhere.

This matters because of US export controls. American rules limit which Nvidia chips can be sold to Chinese companies, and Chinese labs have been pushed to rely more on domestic hardware. DeepSeek has already released software, built together with Huawei, that is used to program AI chips. That is a move into a new area for the company and a sign of a close partnership with the hardware maker. According to The Decoder, DeepSeek is also working on its own inference chip to reduce its dependence on both Nvidia and Huawei.

The topic is sensitive for DeepSeek. The company briefly paused this very funding round after comments widely attributed to Liang went viral on social media. Chinese outlet Yicai reported in July that he had talked about a reliance on Nvidia chips and about China still lagging behind the United States in AI sophistication. Raising tens of billions of yuan and building one of the biggest Huawei clusters in the world is, among other things, a very concrete answer to that worry.

Open models and the IPO question

One thing to watch closely is whether DeepSeek stays as open as it has been. Its models are known for being released with open weights, which means anyone can download and run them. According to Bloomberg, Liang promised in at least one meeting with investors this year that he will keep developing open AI models while pursuing the larger goal of artificial general intelligence. He also made it clear that his main aim is to push the technology forward, not to maximise income.

That promise will be tested. A public company has shareholders, quarterly expectations and pressure to show revenue. Many AI labs that started out very open became more closed as the money grew. DeepSeek has so far been the clearest example of a leading lab that gives its best work away, and that is a big part of why developers around the world like it. If the company keeps that approach after an IPO, it would be a real exception in the industry.

There is also a regulatory question. The Information reported earlier that Chinese regulators had opened an investigation into DeepSeek and Moonshot AI over data security, as The Standard noted on Tuesday. It is not clear yet whether that will affect the valuations or the timing of either listing.

Why it matters

For people who use AI tools, the main effect is indirect but real. A DeepSeek with 12 billion dollars or more in fresh capital can train bigger models, serve them more cheaply and keep releasing open weights that other companies and hobbyists build on. Cheap, strong open models put pressure on prices everywhere, including at OpenAI, Anthropic and Google.

For the global AI race, the round shows that China's top labs can raise money at a scale that used to be reserved for American companies. It also shows that the hardware story is shifting. A cluster of 160,000 Huawei accelerators would be a major test of whether Chinese chips can carry frontier AI work at large scale.

For investors, a DeepSeek IPO in 2027 would be one of the most watched Chinese market debuts in years. Together with Moonshot's planned listing, it gives public market investors their first real chance to buy into China's leading AI labs directly.

Dany's take

I find the gap between target and result the most telling part. Asking for 50 billion yuan and ending up near 100 billion means investors are not just interested, they are worried about missing out. That is good news for DeepSeek and for everyone who uses open models, because more money usually means more and better releases. But it also raises the stakes. The real question for me is not whether DeepSeek can raise money, it clearly can. It is whether it will still publish its best models openly once it has shareholders to answer to. I hope it does, because DeepSeek staying open keeps the whole market honest on price.

Sources

• Bloomberg: DeepSeek to raise at least 12 billion in Tencent backed funding
• The Straits Times (Bloomberg text): DeepSeek set to raise at least 15 billion Singapore dollars in Tencent backed funding
• CNBC TV18: China's DeepSeek to raise 12 billion dollars in Tencent backed funding
• The Decoder: CATL and Tencent back DeepSeek's ballooning funding round
• The Standard: Moonshot AI mulls Hong Kong IPO in Q1 2027 after topping 50 billion dollar valuation