DeepSeek’s DeepSeek funding round is shaping up as more than a simple cash raise. It is the clearest sign yet that founder Liang Wenfeng wants the Chinese AI lab to scale up without backing away from its core pitch: pursue AGI, keep the models open-source, and resist the usual pressure to chase near-term revenue first.
That makes this moment unusually revealing. In meetings with prospective investors, Liang has said DeepSeek intends to pursue artificial general intelligence as a primary goal. He has also said the lab plans to keep releasing open-source models rather than prioritize short-term commercialisation, a stance that sets the company apart at a time when many foundation-model developers are under pressure to prove recurring revenue.
It is also a turning point in how DeepSeek finances itself. Until now, the company had been funded entirely by High-Flyer Quant, the quantitative trading firm Liang founded. This new raise would mark the first time he has accepted outside money.
Summary
DeepSeek’s first outside raise
The DeepSeek funding round is reportedly the company’s first external financing effort, formalizing a shift that had been building behind the scenes. For years, the lab’s independence was part of the story. That insulation helped it avoid investor demands around product timelines and monetization.
Now, however, the scale of the work appears to be forcing a new phase.
DeepSeek has told investors that AGI is a primary goal, according to the reported messaging around the round. Just as important, the lab is signaling that it does not want to trade away its research identity in exchange for faster commercial returns. DeepSeek intends to keep releasing open-source models, an approach that has become central to its reputation.
Why this matters is straightforward: outside capital usually comes with expectations. If investors buy into an AGI-first, open-source strategy at this stage, they are not just funding compute and talent. They are also backing a model-development path that is less tied to immediate enterprise sales and more tied to long-horizon research ambition.
What the DeepSeek funding round is seeking
The numbers attached to the DeepSeek funding round are already drawing attention.
Bloomberg described the fundraising as an ongoing 70 billion yuan raise, or roughly $10 billion in valuation terms. That figure is reported as the valuation DeepSeek is targeting, not the amount of money being raised. Separately, The Information reported that the company is seeking at least $300 million in external capital in the round.
So the picture, based on the reported details, looks like this:
- Target valuation: 70 billion yuan, roughly $10 billion
- External capital sought: at least $300 million
Those figures matter because they place DeepSeek in a very different category from a self-funded lab operating on internal support. A valuation target at that level suggests investors are no longer treating the company as a niche research shop. Instead, they are pricing in strategic relevance.
Why DeepSeek is changing now
One reason for the shift is scale. DeepSeek had previously relied on High-Flyer Quant, but training runs at its current size appear to have pushed beyond what even a profitable hedge fund can comfortably self-finance.
That helps explain why the first outside raise is happening now, not earlier. The lab’s ambitions have grown alongside the cost of staying at the frontier.
DeepSeek also kept building while much of the market focused on its earlier breakthroughs. In April, it released V4-Pro and V4-Flash, extending its model lineup with systems that reinforced its open approach. The company released both under permissive open-source licensing.
That is not a small detail. It shows the open-model strategy was not just branding around one high-profile release. It continued into the V4 family even as DeepSeek moved toward external fundraising.
How the V4 models fit DeepSeek open-source models strategy
The V4 family also sends a hardware message. These models are optimized for Huawei Ascend, Cambricon, and Nvidia silicon.
That broad compatibility matters in China’s AI market. It points to a lab trying to stay competitive across a changing chip environment, especially as access to the highest-end US accelerators has become more constrained. In practical terms, DeepSeek is not building only for one hardware stack. It is positioning itself for resilience.
What the move signals for investors and regulators
The strategic backdrop to the DeepSeek funding round is impossible to ignore. The lab’s R1 model had already captured global attention and rattled assumptions about how expensive frontier AI development had to be. According to the reported account, R1’s release erased roughly $600 billion from Nvidia’s market capitalisation in a single trading session.
The exact pricing debate around R1 later became contested. However, the broader message stuck: a Chinese lab could compete at the frontier, and do so in the open.
That is a big part of why investors are paying attention now. DeepSeek is no longer just an interesting technical project. It has become a test case for whether open-source AI can attract serious capital without giving up its identity.
There is another audience watching too. Chinese regulators may pay close attention to a funding round of this size and profile, especially involving a foundation-model developer with an explicit AGI framing. That does not mean a specific regulatory response is set. But it does underline the stakes. Once a lab moves from self-funded experimentation to outside-backed expansion, scrutiny tends to grow.
A new phase for DeepSeek
The DeepSeek funding round brings several threads together at once: bigger compute demands, stronger investor interest, an open-source commitment, and a public AGI ambition from Liang Wenfeng.
For a company backed until now by High-Flyer Quant, that is a major shift in posture. It suggests DeepSeek believes it can scale without abandoning the principles that made it stand out in the first place.
If that holds, the round will not just finance the next stage of model development. It will test whether investors are willing to fund frontier AI on terms defined less by immediate commercialisation and more by long-range research conviction.
