Chinese artificial intelligence pioneer DeepSeek has verbally informed prospective investors that it is suspending its second fundraising round, according to people familiar with the matter cited by Bloomberg. The unexpected halt comes just days after leaked transcripts attributed to founder Liang Wenfeng sparked widespread online discussion about China's reliance on US hardware and its technological lag relative to American rivals.
The suspension marks a sudden shift for the Hangzhou-based startup, which had been seeking between CNY10-50 billion (approx. US$1.48-7.38 billion) in follow-on capital. According to reports from both Bloomberg and The Information, the fresh capital drive was set to value DeepSeek at between CNY480-500 billion, a steep increase from its roughly US$50 billion valuation following a US$7 billion first round that closed in June.
DeepSeek, founded in 2023 under hedge fund Zhejiang High-Flyer Asset Management, had secured backing from major Chinese entities, including Tencent Holdings, Contemporary Amperex Technology Co. Ltd. (CATL), and the state-backed National Artificial Intelligence Industry Investment Fund, per Bloomberg.
Sources cited by Bloomberg said Liang felt frustrated by public reports detailing his private remarks to investors. Chinese tech media, including Yicai and Tencent Technology, published transcripts of a nearly 4-hour investor Q&A in which Liang observed that Chinese AI models remain approximately 6 to 18 months behind top US counterparts.
Liang reportedly emphasized that the gap stems primarily from severe compute constraints rather than talent shortages, stating that "what stops us from scaling is not willingness, but insufficient compute." Liang also noted that while DeepSeek continues to rely on Nvidia graphics processors, the startup developed its own compiler, TileLang, to reduce its dependence on Nvidia's CUDA ecosystem. Furthermore, Liang highlighted that Huawei's Ascend platform shows strong competitiveness, though it continues to face supply and capacity limitations.
Despite the temporary fundraising freeze, DeepSeek has demonstrated substantial commercial traction and rapid institutional evolution. According to three sources cited by The Information, DeepSeek's annualized revenue recently reached between US$400 million and US$500 million, propelled by enterprise sales of its application programming interface (API) access. Operating at gross margins of 70% to 80%, the startup has maintained profitability while pricing its flagship V4 model at a fraction of Western competitors' prices, largely due to architectural innovations that minimize the required chip infrastructure.
Commercial growth and IPO preparations continue despite fundraising pause
The developments surrounding DeepSeek carry broader implications for both global venture capital and the Sino-American technological rivalry.
Firstly, Liang's candid assessment underscores that while Chinese labs can achieve high algorithmic efficiency under hardware constraints, access to advanced semiconductor fabrication remains the primary bottleneck for China's AI sector.
Secondly, DeepSeek's shift toward accepting foreign capital in dollars and pursuing a domestic public listing underscores the intense pressure on generative AI firms to secure sustainable, long-term financing for expanding data center infrastructure.
Finally, the episode demonstrates the high degree of political and public sensitivity surrounding executive commentary in China's AI sector, where public evaluations of national competitiveness can directly impact corporate dealmaking and broader market dynamics.
Article edited by Jack Wu
