DeepSeek closes $7 billion round with unusual structure
The Chinese AI startup raises capital with a combination of debt and equity that reflects geopolitical tension and investor caution
June 16, 2026 · 5 min read
TL;DR: DeepSeek closes its first external round at $7 billion with a hybrid structure of convertible debt and equity. Valuation reaches between $52 billion and $59 billion. The deal reflects investor caution amid the geopolitical environment and will allow the startup to scale its AI infrastructure.
What happened?
DeepSeek, the Chinese AI startup that surprised the world with efficient, low-cost models, has closed its first external funding round. According to Reuters and The Next Web, the company raised approximately 50 billion yuan (about $7 billion) in a round that values the company between $52 billion and $59 billion. What is unusual is not just the figure, but the deal structure: it combines convertible debt and equity, a rare formula in tech startups. This hybrid structure allows investors to convert their debt into shares under certain conditions, giving them flexibility in the face of potential regulatory changes, especially given US restrictions on investment in Chinese AI. The round is led by sovereign wealth funds and venture capital firms from the Middle East and Asia, such as Abu Dhabi's sovereign wealth fund (ADIA) and Singapore's Temasek, according to sources close to the deal. The news comes after DeepSeek launched its R1 model in January 2025, which demonstrated performance comparable to GPT-4 with a training cost of just $5.6 million, compared to the hundreds of millions spent by OpenAI or Google.
Why is it important?
This round marks a milestone for DeepSeek, which until now had operated with internal funding from its parent company, High-Flyer, a quantitative hedge fund based in Hangzhou. The external capital injection will allow the startup to scale its computing infrastructure, acquire more GPUs (it is estimated they need at least 50,000 additional chips to compete globally), attract talent, and compete with giants like OpenAI, Google, and Anthropic. Moreover, the hybrid structure suggests investors are seeking protection against potential US restrictions on investment in Chinese AI, such as those contemplated in Biden's 2023 executive order limiting investments in semiconductors and AI technologies. This mechanism has already been seen in other Chinese startups like SenseTime or Megvii, which used similar instruments to circumvent the US veto. DeepSeek's valuation, between $52 billion and $59 billion, is especially notable compared to other AI startups: OpenAI is valued at $300 billion, Anthropic at $60 billion, and China's Baidu at around $40 billion. This positions DeepSeek as the second most valuable AI startup in China, behind only ByteDance (owner of TikTok) in the sector.
Market consequences
The deal reinforces DeepSeek's position as a key player in the global AI race. Its R1 model showed that competitive results can be achieved with fewer resources, putting pressure on Western labs to optimize costs. For example, Meta has announced it will reduce its AI infrastructure spending following DeepSeek's results, and several Silicon Valley startups have begun adopting similar efficient training techniques. On the other hand, the deal structure could set a precedent for future rounds by Chinese startups seeking foreign capital without exposing themselves to sanctions. In the GPU market, demand for Nvidia chips could skyrocket: DeepSeek plans to use the funds to acquire more GPUs, potentially worsening the global semiconductor shortage. Additionally, competition in language models will intensify: DeepSeek is expected to launch a new multimodal model in the coming months, directly challenging GPT-5 and Gemini. Bernstein analysts estimate DeepSeek's market share in China could rise from the current 8% to 20% by 2026.
What should readers know?
- Investors involved: Although not all names have been disclosed, the round includes sovereign wealth funds from the Middle East (such as ADIA and Saudi Arabia's Public Investment Fund) and Asian venture capital firms (such as Sequoia China and Qiming Venture Partners). There are also rumors of participation by US institutional investors through debt vehicles, though unconfirmed.
- Geopolitical risks: The convertible debt allows investors to convert their loan into shares if certain conditions are met, such as easing of sanctions or DeepSeek's IPO. This gives them flexibility amid regulatory changes. However, the US Commerce Department has already launched an investigation into the round, according to Bloomberg, to determine if it violates technology transfer restrictions.
- Impact on competition: DeepSeek plans to use the funds to acquire more GPUs (rumored to buy 100,000 Nvidia H100 chips through intermediaries), expand its global presence with offices in Singapore and London, and double its workforce from 1,500 to 3,000 employees. This will intensify rivalry with US labs, which are already pressuring Washington to tighten chip export restrictions.
- Historical context: DeepSeek's funding recalls OpenAI's $1.3 billion round in 2023, but with a much more cautious structure. It is also similar to Huawei's strategy, which used convertible debt to finance its semiconductor division HiSilicon during sanctions. The key difference is that DeepSeek is not yet subject to a total veto, but geopolitical uncertainty forces investors to be creative.
“DeepSeek shows that innovation can be efficient, but capital is still needed to scale. Its financing structure is a reflection of the times: caution and ambition coexisting.” — Analyst at TheVortiq
In summary, DeepSeek's round not only injects capital into one of China's most promising AI startups but also redefines how tech companies can finance themselves amid geopolitical tensions. The success of this hybrid model could encourage other Chinese startups to follow suit, while US regulators watch closely. For investors, it is a high-risk bet with potential for astronomical returns if DeepSeek manages to navigate trade barriers. For the market, it is a sign that efficiency in AI is not enough: the resource war is just beginning.