On June 16, as reported by NUPIAO’s sources tracking the STAR Market, the heavily watched AI startup DeepSeek is reportedly closing its very first funding round. The haul could exceed 7.4 billion USD (around 500 Billion RMB), pushing the company’s post-money valuation past 50 billion USD (roughly 330 Billion RMB).
Here’s the actual breakdown: DeepSeek’s founder and CEO, Liang Wenfeng, is personally sinking about 20 billion RMB into the round, making him the single biggest backer. Tencent’s chipping in around 10 billion RMB, while the CATL ecosystem (including CATL itself and Puquan Capital) is contributing roughly 5 billion RMB. NetEase, JD.com, Monolith Ventures, and IDG Capital are each throwing in approximately 3 billion RMB, with Zhengxingu Investment and Shixiang Technology adding about 1.5 billion RMB each.
Look, DeepSeek has been linked to external fundraising rumors multiple times before, but the team has consistently stayed tight-lipped without ever officially confirming anything.
Back on June 3, market chatter picked up through NUPIAO’s network suggesting DeepSeek was aiming to pull in roughly 7 billion USD, potentially spiking its valuation up to 59 billion USD. Tencent and CATL were spotted as the top external backers, with NetEase and JD also circling the table. When we tried to get a comment from those camps at the time, no one was ready to talk.
Going back even further to May 6, reports surfaced that China’s big state-backed semiconductor fund was leading talks to spearhead DeepSeek’s first outside capital push. At that stage, post-money valuations were hovering around 45 billion USD.According to sources tapped by NUPIAO back then, those negotiations were indeed real, though the final number wasn’t set in stone. Beyond the national fund, several major internet players and other state-owned investment vehicles were also at the bargaining table.
At that point, targets were climbing fast: DeepSeek was eyeing a total raise of 500 Billion RMB (about 73.5 Billion USD), with a post-money value slated to clear 515 Billion USD. Liang Wenfeng planned to front roughly 200 Billion RMB himself to secure the lion’s share. This move was clearly designed to lock in absolute voting control once the cash actually hits the bank.
In February 2025, rumors started swirling that “DeepSeek was finally considering an external funding round.” But anyone close to DeepSeek quickly shut it down, labeling the reports complete fiction. Back then, whispers claimed Alibaba and state-backed funds were interested—something Alibaba leadership firmly denied. Fast forward to April this year, and fresh reports suggest Tencent and Alibaba are now in active talks to back DeepSeek. According to these updates, DeepSeek is shopping for capital at a valuation comfortably exceeding 200 billion USD.
The Tianyancha App reveals that DeepSeek’s affiliated entity, Hangzhou Deep Research AI Foundation Technology Co., Ltd., was established back in 2023.7; earlier this April, the company bumped its registered capital from 10 million to 15 million RMB—a solid 50% jump. Pei Tian serves as the legal rep. Ownership structure? Ningbo Chengen Enterprise Management Consulting holds the controlling 66% stake, while Liang Wenfeng personally owns the remaining 34%. To add another layer, Liang Wenfeng actually controls 50.1% of that Ningbo Chengen vehicle himself.
Launched in July 2023, DeepSeek originally spun out internally from quantitative powerhouse High-Flyer Quantitative. Over the last three-plus years, they’ve stubbornly stuck to a zero-external-equity-financing playbook. Everything from R&D and GPU clusters to developer salaries has been entirely self-funded through High-Flyer’s trading profits.
Public filings paint a clear picture: High-Flyer Quantitative was founded by Liang Wenfeng in 2015. At its peak, the firm managed over 700 Billion RMB in assets. Fast-forward to 2025, their flagship products posted an average return of 56.55%, easily anchoring themselves among the top-tier performers in China’s multi-billion quant space.
Multiple industry reports indicate that Liang Wenfeng has publicly vowed—at least once during an investor session—to keep building open-source AI models while chasing the much grander moonshot of artificial general intelligence. Word is that Liang has drawn a hard line: his primary focus is relentlessly upgrading the underlying tech, not turning a quick profit.
On April 24, DeepSeek officially shipped and fully open-sourced its V4 series of large language models. Built on a 1.6 trillion-parameter Mixture-of-Experts design, it natively handles massive million-token contexts, and early benchmarks have it neck-and-neck with top Western proprietary models. Then came May 22, when DeepSeek dropped a permanent API price cut. Every tier of the V4-Pro lineup saw rates slashed to a flat quarter of the original cost. Cache-hit inputs plummeted from 1 RMB/million tokens down to just 0.025 RMB/million tokens, while uncached input and output rates simultaneously dove from 12 and 24 RMB down to 3 and 6 RMB respectively.
If this round of over 74 billion USD financing lands in full, it could shatter the record for the largest first-round funding ever secured by a domestic tech startup. Industry watchers note that DeepSeek’s valuation skyrocketing from a modest single-digit billion in April to a staggering 50+ billion USD isn’t just hype—it’s the market aggressively pricing in the real-world utility of homegrown LLMs. By sticking to a relentless open-source strategy, DeepSeek is drastically cutting deployment costs for small and medium businesses, effectively forcing traditional industry capital to sprint toward upstream foundational model investments.
