A company with roughly 160 employees just raised around $7 billion and is gunning for a valuation north of $70 billion. DeepSeek, the Chinese AI lab that sent shockwaves through global markets when it dropped its R1 model in January 2025, is no longer just a scrappy underdog. It’s becoming something that looks a lot more like a strategic national asset.
The Hangzhou-based company, founded in July 2023 as a spin-off from the High-Flyer quantitative hedge fund, is now in discussions for yet another funding round at a pre-money valuation between $71 billion and $74 billion. The endgame: a potential IPO on the Shanghai STAR Market, reportedly targeted for 2027.
From hedge fund side project to global AI contender
Liang Wenfeng, the company’s CEO, carved the AI lab out of his quant trading firm and staffed it with a skeleton crew. The team then built the V3 model for an estimated $6 million in training costs. Leading Western frontier models have been developed with budgets measured in hundreds of millions, sometimes billions. DeepSeek accomplished comparable benchmark performance at a fraction of the cost, partly through innovations like Mixture-of-Experts architecture and the creative use of older, less restricted hardware.
The R1 model, launched in January 2025, matched the performance benchmarks of top Western competitors. That release was widely described as a “Sputnik moment” for AI. The V4 model preview, unveiled in April 2026, focused on enhanced coding abilities and agent capabilities.
The money keeps flowing
DeepSeek closed its first major external funding round at approximately $52 billion in valuation around late May or early June 2026, pulling in roughly $7 billion to $7.4 billion. Prior to that, the company was primarily bankrolled by the High-Flyer hedge fund itself.
Now the company is already exploring additional capital raises, pushing toward that $71 billion to $74 billion pre-money valuation. If the IPO materializes in 2027, it would be one of the most closely watched public offerings in the AI sector. DeepSeek operates under the shadow of US chip export restrictions designed specifically to slow down Chinese AI development.
Why crypto and tech investors should care
When DeepSeek’s R1 model dropped in early 2025, AI-related digital asset tokens took marked hits, revealing an underappreciated correlation between frontier AI developments and crypto market performance.
No major crypto tokens are directly linked to DeepSeek. But the company’s collaboration with Mind Network to integrate fully homomorphic encryption tools for secure AI inference, announced around February 2025, highlights where AI and crypto-native technology increasingly overlap. Fully homomorphic encryption allows computations on encrypted data without ever decrypting it, meaning AI models can process sensitive information without actually seeing it.
DeepSeek’s cost-efficiency threatens to compress margins and valuations across chip manufacturing and AI infrastructure firms. US companies that have commanded premium valuations based on the assumption that frontier AI requires frontier hardware budgets now face a competitor proving otherwise.
DeepSeek has released models under open-weight licenses, a strategic choice that accelerates adoption and ecosystem development while potentially undermining the closed-model business strategies of Western competitors.