DeepSeek, the Chinese AI lab that rattled Western tech markets earlier this year with its surprisingly capable and cost-efficient models, is in talks to raise its first round of venture capital at a valuation of approximately $45 billion — more than double the $20 billion figure that was circulating just weeks ago.

China’s Big Fund, officially known as the China Integrated Circuit Industry Investment Fund, is in discussions to lead the round. Tencent is also reportedly in talks to participate. The company’s founder, Liang Wenfeng, who holds 89.5% of DeepSeek, may also invest personally.

The rapid valuation surge is notable because DeepSeek originally set out to raise only a modest amount — enough to establish a valuation for employee stock options and prevent competitors from poaching key staff. It has since become something significantly larger.

Why This Number Matters

A $45 billion valuation puts DeepSeek in the same stratosphere as some of the most well-funded AI labs in the Western world, despite raising essentially no outside capital until now. For context, that figure is larger than many late-stage enterprise software companies and reflects just how seriously investors are taking DeepSeek’s competitive position in the global AI race.

What makes this unusual is that DeepSeek built its reputation not by spending freely, but by being efficient. Its models have consistently matched or outperformed Western counterparts at a fraction of the compute cost. That narrative — doing more with less — is exactly what investors are betting on.

What DeepSeek Has Built

DeepSeek’s R1 and subsequent model releases demonstrated that large language models don’t require the kind of massive compute budgets that OpenAI or Anthropic have spent. The company achieved frontier-level performance using optimization techniques that allowed it to sidestep some of the US chip export restrictions that constrain Chinese AI development.

Its models are open-weight, which means businesses and developers can run them on their own infrastructure. That openness, combined with strong benchmark performance, has made DeepSeek’s models popular globally — including in markets outside China — and has pressured Western AI providers to justify their pricing.

The most recent DeepSeek V4 preview extended that trend, with an expanded context window designed specifically for enterprise workloads. The lab is not standing still.

State Capital at the Center

The involvement of China’s Big Fund is worth noting. This is the same state-backed vehicle that has poured capital into semiconductor development in China over the past decade. Its interest in DeepSeek is a signal that the Chinese government sees AI model development as strategic infrastructure — not just a commercial bet.

That framing matters for how Western businesses and governments interpret DeepSeek’s growth. A Chinese AI company backed by state chip capital, offering open-weight frontier models that outperform on cost, is a fundamentally different competitive dynamic than a US startup funded by VC.

What This Means for Business

For business leaders evaluating AI strategy, the DeepSeek story carries a few practical implications.

Cost benchmarks are shifting. If you’re paying premium rates for AI inference and assuming that’s the only option, DeepSeek’s growth puts pressure on that assumption. Open-weight models running on your own infrastructure can dramatically reduce per-token costs for high-volume use cases.

The AI market is not a duopoly. Many enterprise AI conversations default to OpenAI vs. Anthropic. DeepSeek’s trajectory — and its $45 billion valuation signal — suggests that a capable third force is real and growing. Your vendor evaluation should account for it.

Open-source AI has staying power. DeepSeek’s open-weight approach has attracted not just individual developers but enterprise teams who want control over their data and compute. The market is validating that model, not just as a niche alternative but as a serious competitive strategy.

China-US AI competition is intensifying. The involvement of state-backed capital on both sides (US CHIPS Act funding, China’s Big Fund) means the geopolitical dimensions of AI infrastructure will continue to shape what models are available, where they can run, and what compliance obligations apply to businesses using them.

The Broader Picture

In the span of a few months, DeepSeek went from being a Chinese lab most Western businesses had never heard of to being valued at $45 billion in its first funding round. That pace of market validation reflects genuine technical achievement, but it also reflects how quickly the AI landscape is evolving.

For any organisation still treating AI as a single-vendor, wait-and-see decision, the DeepSeek story is a useful forcing function. The model landscape is diversifying rapidly. The cost of capable AI is falling. And the question is no longer whether to adopt AI, but which models, on which infrastructure, with which vendors — and how to stay agile as the answer keeps changing.

If your organisation doesn’t yet have a structured way to evaluate and stay current on AI developments, that gap is worth closing. Enterprise DNA’s learning platform helps data and business teams build the literacy to make those calls confidently, rather than defaulting to whoever has the biggest marketing budget.