Picture a company that gives its most valuable product away for free, refuses outside investment for years, and then closes one of the largest private AI funding rounds in Chinese history, all while making sure investors have no voting rights and cannot touch their money for five years. That is not a paradox. That is a deliberate product strategy, and understanding it is worth your time whether you build AI products, invest in them, or just use them. ## The Numbers, and the Structure Behind Them DeepSeek, the Chinese open-source foundational model company, raised $7.5 billion at a valuation north of $50 billion, according to Axios. The round is the company's first external fundraise ever. As recently as April, the numbers being discussed publicly were at least $300 million at a $10 billion valuation, according to The Decoder, which means the final figure landed at five times that valuation and roughly 25 times that capital target. That kind of gap between early whisper numbers and a closing print usually signals either enormous competitive demand for the deal or a deliberate decision to time the raise for maximum leverage. Given what we know about the deal structure, the answer is probably both. The structure itself is what product people should study. According to The Decoder and CityBiz, investors were required to put their money into a limited partnership managed by founder and CEO Liang Wenfeng rather than investing directly into DeepSeek. They receive no voting rights. They face a five-year lock-up period. The only participant exempt from those terms was China's National Artificial Intelligence Industry Investment Fund, which invested directly and retained both voting rights and liquidity flexibility. Liang himself contributed approximately 20 billion yuan to the round, according to Reuters. Tencent and battery manufacturer CATL were among the largest outside participants, per Axios. ## What "Open-Source" Means When the Founder Holds Every Vote Here is the tension worth sitting with. DeepSeek built its reputation on openness. Its R1 model, released earlier this year, rattled the AI industry precisely because it demonstrated that a lean, research-focused team could produce a model competitive with much larger Western labs and then publish the weights for anyone to use. That openness is a genuine product strategy, not a marketing posture. It drives adoption, it attracts researchers, and it builds a kind of ambient trust that closed-model companies have to spend marketing budgets to simulate. But the funding structure does something interesting in parallel. By routing all capital through a limited partnership that Liang controls, with no board seats granted to investors and no liquidity for five years, DeepSeek has raised institutional money without surrendering any of the governance levers that typically accompany it. The model weights stay open. The company strategy stays closed. That is a combination most founders only dream about, and Liang has apparently pulled it off at a scale that makes it real. According to The Decoder, Liang told investors ahead of the round that he prioritizes foundational AI research and AGI development over short-term profits. Investors agreed to those terms anyway, which tells you something about the perceived scarcity of the asset. ## What Product Builders Should Take From This There is a lesson here that extends well beyond AI fundraising. The open-source model has always carried a monetization question mark: if you give the core product away, where does the business come from? The standard answers are support contracts, hosted infrastructure, and enterprise add-ons. DeepSeek's round suggests a fourth answer that is less discussed: strategic capital positioning. By raising at a $50 billion valuation while keeping the model open, DeepSeek is not monetizing openness directly. It is using openness to build a moat of researcher loyalty and global adoption, and then raising capital against that moat's future value. The capital, in turn, funds the foundational research that keeps the moat deep. That is a flywheel, and it is a clean one. For anyone building on top of open-source foundations, including developers, edtech teams, and enterprise product managers, the implication is worth internalizing. Open-source AI infrastructure is increasingly being funded at a scale that closed-model companies cannot ignore. The research output from a well-capitalized DeepSeek, operating without quarterly earnings pressure and with a founder who explicitly told investors that short-term profits are not the priority, is likely to keep arriving faster than markets expect. The April whisper valuation of $10 billion became a $50 billion close in roughly six weeks. Watch what the research output looks like six months from now. ## Sources - DeepSeek's new round, CVC's prosthetics bet, and Databricks' buy - Axios

Sources