Report: DeepSeek Banked $15B, Battery Maker Doubled Down

By Daniel Rao · October 22, 2026

Fifteen billion dollars, raised in the dark, confirmed by no one. According to Bloomberg — the only newsroom with this story, citing unnamed people familiar with the matter — DeepSeek has closed a roughly 100 billion yuan (about $15 billion) funding round that was supposed to be half that size, then doubled, then doubled again as investors fought for allocation. And DeepSeek, Tencent and CATL — the last two reportedly the round's biggest backers — all declined to comment. Bloomberg says they didn't even answer the emails.

That is the state of the biggest AI funding story out of China this year: enormous numbers, an impeccable sourcing pedigree, and not a single named human willing to stand behind any of it. Which is not a reason to ignore it. It is a reason to read it the way it deserves to be read — as Bloomberg reported it, hedges and all — and to pay attention to the strangest detail in the whole package: the company writing the largest cheques to China's most important AI lab makes batteries.

The $15 billion nobody will confirm

The report landed at 02:17 UTC on October 22, with bylines for Lulu Yilun Chen and Haze Fan of Bloomberg Technology. The claim, via Bloomberg's unnamed sources: DeepSeek has raised about 100 billion yuan in its latest round — $14.98 billion at the exchange rate Reuters used for its relay ($1 = 6.6749 yuan), which Bloomberg itself rounds to an even $15 billion.

Here is the attribution chain in full, because it matters more than usual. Bloomberg is the sole origin. Every other outlet — Reuters, CNA, the Straits Times, the Edge — is a relay of the Bloomberg report. Reuters, in its relay (reporting by Devika Nair in Bengaluru), was blunt in the way financial wires rarely are: "Reuters could not immediately verify the report." The sources "asked not to be named as the details are not public," Bloomberg reported. And, per the report: "DeepSeek, Tencent and CATL representatives did not respond to e-mailed requests for comment."

Even the question of whether the round has formally closed is murky. Bloomberg's October 6 piece had described the round as closing soon, with the final tally set to "approach" 100 billion yuan. The October 22 piece uses the past tense — "has raised" — still on anonymous sourcing, with no company statement and no regulatory filing cited. The honest phrasing, and the one this article will use throughout: Bloomberg reports that DeepSeek has raised the money. Nothing more is confirmed.

We checked every image URL and every internal link in this piece with fresh HTTP requests before filing — all returned 200, and both images are used here under their stated licences.

Doubled. Then doubled again.

Per Bloomberg's sources, the round's target was raised at least twice during negotiations: 50 billion yuan, then 80 billion, then a final tally of roughly 100 billion. Mainland financial media, citing unnamed investment-banking sources, tied the doubling to investor demand that kept outrunning supply — though with every name in the chain anonymous, treat the motive as colour, not fact.

Bloomberg itself named only two of the biggest participants: Contemporary Amperex Technology — CATL, the battery giant — and Tencent, the WeChat operator. The fuller roster comes from Chinese financial media (Caijing, via relays): new investors said to include Stony Creek Capital, China Fortune-Tech Capital, CPE Asset Management, Legend Capital and Geely, alongside existing backers CATL, Tencent, IDG Capital, Lux Capital and Silicon 100 — with new and existing shareholders each expected to take roughly half the round. One honest caveat: the English renderings wander between outlets. Some transliterate the Chinese names as "Shixi Capital" and "SMIC Capital" instead of Stony Creek and China Fortune-Tech. They appear to be the same entities, but nobody on the record has said so.

What is consistent across the reporting is CATL's appetite. The battery maker reportedly put roughly 10 billion yuan into this round — about $1.5 billion — double the roughly 5 billion yuan it committed in DeepSeek's first round. When your battery supplier doubles its bet on your AI lab within six months, that is not diversification. That is a thesis.

On valuation: the October 22 report states none. Earlier reporting — Bloomberg's October 6 piece and mainland financial media — put the round's pre-money valuation at roughly 500 billion yuan (about $75 billion), launched in July, with the mainland reports describing the valuation as holding steady into the close. DeepSeek's first round, in April, raised over 50 billion yuan at a valuation above 350 billion yuan, per Caijing, and closed in late June — though other summaries of that April round run as high as 400 billion yuan post-investment. The sources don't agree with each other, so this article won't pretend they do.

Photo: Rows of server racks in a generic data centre (file photo — not a DeepSeek facility). The reported ~100 billion yuan raise is expected to fund DeepSeek's AI data-centre build-out, per Bloomberg's earlier reporting. Photo: Carl Lender / Wikimedia Commons (CC BY 2.0).

What fifteen billion buys in Hangzhou

Money this large has a destination, and Bloomberg's earlier reporting sketches it: the Hangzhou startup has begun spending more aggressively to build out AI data centres, including a facility under construction in Inner Mongolia where DeepSeek plans to deploy at least 160,000 Huawei Ascend AI chips. In a year when Washington has spent enormous energy trying to starve Chinese labs of advanced compute, 160,000 domestic accelerators going into the ground in Inner Mongolia is not a footnote. It is the whole strategy in one number.

Photo: Huawei Atlas AI systems on display at WAIC 2025 in Shanghai (file photo — not DeepSeek's own data centre). DeepSeek plans to deploy at least 160,000 Huawei Ascend accelerators at a data centre in Inner Mongolia, Bloomberg reported earlier this month. Photo: Xuthoria / Wikimedia Commons (CC BY-SA 4.0).

The technology the money is meant to scale is not theoretical. DeepSeek released V4.1 Flash in September 2026, and Bloomberg reports the release promptly knocked the share prices of local rivals MiniMax Group and Z.AI — while Bloomberg's earlier coverage said the model "reset expectations for cost versus performance against leaders Anthropic and OpenAI." A day after Washington's purge order (more on that below), DeepSeek open-sourced its DeepInfer inference engine under Apache 2.0 — a move we covered as a direct answer to the US pressure campaign. And The Information reported in late September, via relay coverage, that DeepSeek's annualised revenue run rate had reached roughly $1 billion — more than double the sub-$500 million level of a few months earlier. Revenue doubling, models shipping, data centres rising: that is the package investors were reportedly fighting to buy into.

China's biggest AI raise — maybe ever

Bloomberg's report put it bluntly: the round is "one of the country's largest ever and a sign of the startup's ambitions to compete globally with the likes of Anthropic PBC and OpenAI."

Mainland financial media went further, calculating that the two 2026 rounds together total roughly 150 billion yuan within six months — which they called the largest primary-market financing in China this year. That is a mainland-media claim, not a Bloomberg one, and no independent outlet has verified the arithmetic. But the direction is not in doubt: no Chinese AI startup has ever raised at this pace.

The money also sets up what Bloomberg said "sets the stage for one of the most highly anticipated Chinese market debuts in years." Earlier Bloomberg reporting said DeepSeek hired CITIC Securities to prepare a STAR Market listing in Shanghai — a "landmark initial public offering in early 2027." The pattern is becoming a template: rival Moonshot AI has reportedly closed its final private round at a roughly $50 billion valuation and is pushing toward a Hong Kong IPO in the first quarter of 2027. China's frontier labs are queuing for public markets on a strikingly similar timetable.

Still pocket change next to the Americans

And now the cold water, also courtesy of Bloomberg: "The latest fundraising shows that imbalance is beginning to moderate, although it still has far less financial firepower than Anthropic and OpenAI, which have each raised more than US$100 billion."

Read that again. Fifteen billion dollars — possibly the largest AI raise in Chinese history — and it still leaves DeepSeek an order of magnitude behind the two US labs it is supposed to be challenging. Anthropic, for context, has locked in $518 billion of compute commitments it cannot cancel, per its IPO filing — more than thirty times the size of DeepSeek's reported round. That is the actual shape of the race: not a funding gap closing, but a funding gap that has barely dented.

DeepSeek's counter-argument has always been efficiency, and it has teeth. Bloomberg's 2025 framing still echoes: the lab "triggered a US$1 trillion meltdown in global markets in 2025" by proving it could rival Silicon Valley's leaders at a fraction of the model cost. An Artificial Analysis benchmark chart sketched what Bloomberg called the "DeepSeek death zone" — the price-performance band where competitors "will struggle to survive if they cannot beat the firm's performance or they cannot charge lower prices." Separate data we have covered shows Chinese models already accounting for 57–67% of developer token traffic on OpenRouter. The money may be American-scale lopsided. The usage is not.

Why is a battery company bankrolling China's AI challenger?

This is the detail that deserves more attention than the headline number. CATL is the world's largest EV battery manufacturer. Its reported ~10 billion yuan commitment makes it one of the two largest investors in China's most important AI lab — a lab whose founder, mainland outlets report, once insisted he would never raise financing at all. Sit with that image for a moment: somewhere in this story is a battery executive who went from that to a ten-billion-yuan AI bet in about a year. The industrialist who decided the future is models, not just cells.

Mainland reporting adds that founder Liang Wenfeng has long preferred industrial investors — companies that build things over funds that flip stakes. Geely, the automaker, appears on the Caijing roster too. Whether or not the full list holds up, the shape is consistent: this is industrial capital, not venture capital. CATL and Geely do not need a 10x return in seven years; they need AI capability inside their supply chains, their factories, their cars. That makes the money stickier than VC money — and, from Beijing's perspective, far more useful.

It also reframes the "challenge to the leading US labs" narrative. Anthropic and OpenAI are being bankrolled by cloud hyperscalers and sovereign wealth. DeepSeek is being bankrolled by the company that powers the world's electric vehicles and the company that runs China's super-app. Different cap tables, different theories of what AI is for.

Washington already purged it

There is a second storyline running in parallel, and it is American. Earlier this month, the Department of Homeland Security ordered civilian federal agencies to purge DeepSeek products — and third-party services built on DeepSeek — from agency IT systems, with a contractor bar following in the FY2026 defence authorisation. We covered the purge order as the moment Washington stopped treating DeepSeek as a curiosity and started treating it as infrastructure risk.

So consider the split screen. In Washington: banned from federal systems, named in defence legislation, treated as a security problem. In Hangzhou: a $15 billion reported raise, a STAR Market IPO on the calendar, 160,000 domestic chips going into the ground in Inner Mongolia. The two-stack world is no longer a forecast. It is a funding round and a blacklist arriving in the same month.

Our take: a $15 billion rumour with excellent sourcing

Here is our stance, stated plainly. A mega-round reported by one newsroom, relayed by two wires, and confirmed by zero named humans is not a fact — it is a rumour with excellent sourcing, and it should be handled like one. Bloomberg's technology desk has earned the benefit of the doubt; "Reuters could not immediately verify the report" is the sentence that keeps everyone honest. Treating anonymous-source fundraising as a done deal would be colossally sloppy journalism, and the shoddiest thing in this story is how little anyone involved will say on the record about the largest AI financing in their country's history.

But the shape of the story rings true even through the hedges, because every surrounding fact checks out: the July launch at ~500 billion yuan, the April round, the September model release that rattled local rivals, the October open-sourcing, the data-centre build-out, the IPO preparations. Anonymous sources lie about valuations; they rarely invent entire financing architectures that cohere this tightly with six months of public reporting.

And the battery maker is the tell. When CATL doubles its commitment to an AI lab — from ~5 billion to ~10 billion yuan in six months — it is not making a venture bet. It is buying a stake in the intelligence layer of its own industry. The US labs are racing to become the everything-company; DeepSeek's backers seem content for it to become the everything-factory's brain. Whether $15 billion is enough to close a gap measured in hundreds of billions is the wrong question. The right one is whether DeepSeek even needs to close it — or whether the death zone, the open weights, and 160,000 Ascend chips are a different game entirely, played on a board Washington can't quite see.

Until a filing or a company statement lands, this stays a reported raise, not a closed one. We'll update this story when — if — anyone involved finds their voice.

Sources: Bloomberg (via the Bloomberg Law mirror and Straits Times/Edge syndication of the October 22 report, plus Bloomberg's October 6 reporting via syndication); Reuters (via CNA relay); Caijing (via The Standard and BigGo relays); 36kr; The Information (via relay coverage); TechPulse USA's own earlier reporting. Quotes are Bloomberg's and Reuters' published wording as carried by those outlets. Analysis is the author's own.

Daniel Rao covers AI, autonomous agents and the business of technology for TechPulse USA.

Update, Oct. 22: the round is wrapping up

Reuters now reports the round is wrapping up, with about 100 billion yuan ($14.98B) generated from existing and new investors, and completion expected in the coming days — though one source cautions the financial details are "not ironclad and could still change." The original 50-billion-yuan target was raised at least twice, to 80 billion and then about 100 billion (Bloomberg), with many investors backed by the state and big corporations. Reuters also reports DeepSeek hired CITIC Securities in September to prepare for a potential STAR Market IPO — a public-markets tap is getting closer.

The $1B revenue line nobody headlined

The Information reported in September that DeepSeek's annualised revenue run rate hit $1B — more than doubled in a few months. CEO Liang Wenfeng disclosed the figure at an investor meeting. Higher API prices didn't drive customers away, and most of the revenue comes from API access for businesses and developers. The raise, in other words, is a bet on a business that's already printing money.

DeepSeek's $15B round: your questions, answered

Who are the biggest investors?

Battery maker CATL and Tencent, per two Reuters sources — many of the round's backers are state-backed funds and big corporations.

Is the round actually closed?

Not yet. Completion is expected in the coming days, and one source cautions the financial details are "not ironclad and could still change."

What does 100 billion yuan buy?

Compute, talent, and a serious run at OpenAI and Anthropic — the war chest to compete at the frontier.

Is a STAR Market IPO next?

DeepSeek hired CITIC Securities in September to prepare for a potential STAR Market IPO — reported preparation, not a decision.

Search This Blog

TechPulse USA

DeepSeek Nears $15B Close as CATL, Tencent Double Down