If you have been watching Chinese AI and wondered why a lab famous for refusing outside money suddenly has Tencent and Alibaba at the table, this roundup pulls the DeepSeek funding news into one place. For two years, the Hangzhou-based company ran on its hedge-fund parent’s balance sheet. In April 2026 — the same week it shipped the V4 Preview — that changed. The initial target was $300 million at a $10 billion valuation. Within 48 hours of investor interest, the benchmark had moved past $20 billion, with one source floating north of $100 billion. Below: who is bidding, how much equity is on the table, how it ties to the V4 launch, and what it means for anyone building on the API.

What the DeepSeek funding news actually says

The short version: DeepSeek is raising external capital for the first time since Liang Wenfeng spun it out of quant hedge fund High-Flyer in July 2023. DeepSeek, the Hangzhou-based AI startup, is looking to raise at least $300 million at a valuation of over $10 billion, reports Reuters. This is DeepSeek’s first external funding round since its founding in 2023 by Liang Wenfeng.

The round was first reported by The Information on April 17, 2026. Within a week, Bloomberg and Reuters were reporting that the valuation benchmark had already moved. DeepSeek, a Chinese AI lab that gained attention in early 2025, is in talks for its first external funding round at a valuation of more than $20 billion. Investor interest pushed the valuation above $20 billion in just 48 hours, according to people familiar with the talks.

A separate South China Morning Post report quoted an investor suggesting the ceiling could go much higher. Large state-backed funds were expected to participate, for example, the AI-focused affiliates of China’s so-called “Big Fund III,” one of the investors said, adding that DeepSeek’s estimated valuation could, in fact, exceed US$100 billion. Treat that number with caution — it is one source on an un-closed deal — but it tells you how fast the market’s opinion is moving.

The numbers at a glance

Who is bidding — Tencent, Alibaba, and state funds

The two named strategic bidders are Chinese cloud incumbents. Tencent Holdings Ltd. and Alibaba Group Holding Ltd. are in discussions to join a maiden round of financing for Chinese AI pioneer DeepSeek, marking a milestone for the country’s artificial intelligence sector. Tencent has proposed the acquisition of as much as a 20% stake in the startup as part of the funding round, though DeepSeek isn’t keen on ceding such a large portion of control, a person familiar with the matter said.

The strategic logic is about more than money. Both Tencent and Alibaba are competing for leadership in Chinese AI alongside startups like Moonshot and MiniMax. They are also major providers of cloud computing services, similar to Microsoft Corp. and Amazon.com Inc., and are eager to supply computing power to emerging startups that require data centers to run their AI platforms. In other words: backing DeepSeek is a way to lock in the compute customer relationship.

State-backed capital is also circling. The company is being highly selective in the round, according to another investor based in Hangzhou, adding that DeepSeek is prioritising state-backed and industrial investors – including local government guidance funds and affiliated platforms – favouring those that can provide strategic resources such as AI infrastructure over purely financial capital, the investor said.

Alibaba’s terms are less clear than Tencent’s. Alibaba is also engaged in funding discussions, but it remains unclear what the e-commerce company is proposing. None of the three companies has commented on the record; the market reaction was still visible — their shares declined in Hong Kong on Thursday.

Why DeepSeek is raising now — after two years of saying no

DeepSeek has spent the entire post-R1 news cycle turning investors away. The structural reason: Liang Wenfeng’s hedge fund has been underwriting the lab. For two years, DeepSeek was funded entirely by its parent, hedge fund Zhejiang High-Flyer Asset Management, which posted a 56.6% return in 2025. That is a very profitable hedge fund — but it is still one hedge fund.

Three forces pushed the company to the table now.

1. Compute scale has outgrown what a single fund can cover

Reasoning models, agent systems, and the infrastructure require a lot of money, more than even a successful hedge fund can provide forever. This tracks with DeepSeek’s newly public infrastructure push. To support the full-scale deployment of its new generation model, DeepSeek is no longer relying solely on software-level optimization, but is expanding into the asset-heavy infrastructure sector. The company is planning to build a large physical data center in Ulanqab, Inner Mongolia, in northern China, to meet the massive and growing computing power demands of future models.

2. Talent retention

The research team is small and being poached. DeepSeek is not in urgent need of outside capital. Instead, the round was intended to establish a market benchmark for its valuation, giving employees greater clarity on the worth of their stock options – a move aimed at stemming departures amid aggressive poaching by well-funded rivals, the Hangzhou and Beijing-based investors said.

Reporting from LatePost (via BigGo) put numbers on the bleed: since the success of the R1 model, at least five core R&D members have left DeepSeek, covering key technical areas like base models, reasoning, OCR, and multimodality. These include key V3 model contributor Luo Fuli (who joined Xiaomi) and GRPO algorithm core author Guo Daya (who joined ByteDance). DeepSeek’s core team numbers only around a hundred people. Talent loss at this scale exposes the shortcomings of its extremely flat, free, yet immature equity incentive system. With competitors poaching talent with total compensation packages (including cash and options) 2-3 times higher, and the wealth effect created by the listings of companies like Zhipu AI and MiniMax, an option agreement that cannot be priced or liquidated is becoming increasingly less attractive to top talent. A priced round gives employees a number to put on those options.

3. Timing with the V4 Preview launch

The funding leak landed the same week DeepSeek shipped its next generation. The DeepSeek V4 family debuted on April 24, 2026, splitting into a frontier-tier DeepSeek V4-Pro (1.6T total, 49B active parameters) and a cost-efficient DeepSeek V4-Flash (284B / 13B). Both carry a 1M-token default context window and up to 384K-token output. Raising money on the eve of a launch is defensive: it validates the company’s story before the benchmarks get picked apart.

The control battle: 3% versus 20%

The tension in the deal is not the valuation — it is how much of the company Liang is willing to sell.

The original plan was narrow. The Hangzhou-based company, a spin-off from hedge fund High-Flyer, is seeking to sell no more than 3 per cent of its equity, said three investors with direct knowledge of the plans, who requested anonymity as the information is private. A 3% sale at $10 billion gets you to the $300 million target cleanly and keeps Liang in absolute control.

Tencent’s proposal upends that. A 20% stake would make Tencent the dominant outside shareholder, which cuts against everything DeepSeek’s founder has said about independence. The key risk is a loss of control and culture. Tencent has proposed acquiring as much as a 20% stake, a move DeepSeek is resistant to. This dynamic represents a direct conflict between securing deep-pocketed capital and maintaining the research-focused independence that produced its breakthrough models. The funding round will test whether DeepSeek can scale without ceding strategic control or its open-source ethos.

How the valuation compares to peers

The numbers start to make sense once you set them beside other Chinese AI firms. For comparison, Zhipu AI’s latest market capitalization after listing in Hong Kong is approximately $50.7 billion; MiniMax’s latest market cap is around $34.4 billion; fellow unlisted unicorn Moonshot AI’s valuation has risen from $4 billion in November 2025 to $18 billion.

At the $10 billion anchor, DeepSeek looked priced below its weight. At $20 billion+ it looks rational. If the reported ceiling of $100 billion held, it would put DeepSeek within shouting distance of frontier US labs — though still far behind OpenAI. OpenAI said in February that it was raising $110 billion in a funding round valuing the company at about $840 billion, with that figure later rising to roughly $852 billion in early April.

One complication for valuation models

Unlike OpenAI or Anthropic, which earn revenue from subscriptions and APIs, DeepSeek gives its models away for free, so it is harder to judge its value based on revenue. The API generates revenue, but the open-weight strategy and aggressive pricing mean traditional SaaS multiples do not map cleanly. V4-Flash lists at $0.14 per million cache-miss input tokens and $0.28 per million output; V4-Pro at $1.74 / $3.48. Both are well below frontier-US rates. Investors are paying for influence and talent density as much as a P&L.

What the raise does not change — and what it might

For developers and enterprise buyers watching this news, the operational story is mostly unchanged in the short term. DeepSeek’s API (POST /chat/completions, the OpenAI-compatible endpoint hosted at https://api.deepseek.com) continues to run against the V4 family. A minimal Python call using the OpenAI SDK still looks like this:

from openai import OpenAI
client = OpenAI(base_url="https://api.deepseek.com", api_key="...")
resp = client.chat.completions.create(
    model="deepseek-v4-flash",
    messages=[{"role": "user", "content": "Summarise the DeepSeek funding round."}],
)

Three points worth noting, since they come up in every funding-news comment thread:

• The API is stateless. Each request must resend the conversation history. The web chat and DeepSeek app persist session state; the API does not.
• Thinking mode is a parameter, not a model ID. Both V4 tiers accept reasoning_effort="high" with extra_body={"thinking": {"type": "enabled"}}, returning reasoning_content alongside the final content.
• Legacy IDs still work until 2026-07-24 15:59 UTC. deepseek-chat and deepseek-reasoner currently route to deepseek-v4-flash. After that cutoff, requests using those IDs will fail — see DeepSeek API documentation for the migration path.

What could shift longer-term: commercial posture. While unconfirmed, this move, coupled with recent hiring for “ToC cloud business” procurement roles, signals a shift from pure R&D towards productization and commercial operations. Analysis suggests the funding’s core purposes may include establishing a market-based price for employee stock options to retain talent, selecting cooperative investors via a relatively low valuation, and potentially paving the way for financial independence from its parent company, High-Flyer Quant. A financially independent DeepSeek with cloud-giant backers probably looks less like a research lab and more like a company that sells things — whether that is dedicated capacity, enterprise support, or regional deployments.

What to watch next

1. Final close size and valuation. The round is not closed. The size could still land anywhere from $300 million at $10 billion to a multiple of that.
2. Tencent’s final stake. If it settles below 10%, the independence narrative survives. If it lands at or near 20%, read it as a meaningful structural change.
3. State-fund participation. “Big Fund III” affiliates as anchor investors would strengthen DeepSeek’s domestic regulatory standing and complicate any future US engagement — see US restrictions on DeepSeek for context.
4. Hugging Face license on V4 checkpoints. V4-Pro and V4-Flash ship under MIT for both code and weights. Any post-funding drift away from that would be a material change.
5. API pricing durability. Today’s V4-Flash rates undercut V3.2. Investors with a 3–7× expected return on a $20 billion entry valuation may push for margin expansion.

For ongoing coverage, the DeepSeek news hub tracks each disclosure as it lands, and DeepSeek company news consolidates corporate-level updates including leadership, hiring, and regulatory items.

Bottom line

The DeepSeek funding news is not a rescue round; it is a benchmark round. The parent hedge fund is profitable, the V4 Preview shipped on schedule, and the company is selling a small slice of equity mainly to put a defensible number on employee options and to bring in compute-provider partners. What is genuinely new is that a lab which publicly disdained outside capital for two years just opened the door, and two of China’s biggest platforms walked through it. Whatever number the round closes at will set the reference price for every Chinese AI deal for the rest of 2026.

Last verified: 2026-04-24. DeepSeek AI Guide is an independent resource and is not affiliated with DeepSeek or its parent company. Model IDs, pricing and API behaviour change; check the official DeepSeek documentation and pricing page before committing to a production decision.