DeepSeek Halts $74B Funding Round After CEO Transcript Leak

The U.S. Bureau of Industry and Security is formally investigating whether Chinese AI companies like Moonshot are improperly accessing advanced American chips and training models on intellectual property from U.S. labs such as Anthropic. Trump administration officials have publicly accused Moonshot and other Chinese open source AI firms of stealing IP from American AI developers. If the investigation concludes misconduct occurred, the Commerce Department could add Moonshot to its entity list, restricting access to U.S. advanced chip technology.

DeepSeek, a Chinese AI lab, is planning another funding round weeks after closing a $7.4 billion raise at a post-money valuation exceeding $50 billion. The rapid succession of fundraising rounds signals aggressive capital deployment and expansion plans. The move comes as competition in AI development intensifies globally.

NVIDIA claims its inference software stack has reduced token costs by up to 5x on the DeepSeek V4 model within one month on its Blackwell platform. The company argues that as AI moves from pilots to production, software optimization across serving, acceleration, and infrastructure layers becomes critical to cost efficiency. Leading inference providers including Baseten, Cognition, Deep Infra, and Together AI is already deploying these tools to improve throughput and reduce latency on Blackwell GPUs.

DeepSeek has open-sourced DSpark, an MIT-licensed framework that accelerates large language model inference by up to 85% without altering model outputs. The system uses speculative decoding, where a smaller draft model predicts likely token sequences that a larger model then validates, reducing computational overhead. DeepSeek has released technical papers, model checkpoints, and training code via GitHub and Hugging Face, making the technique available to researchers and enterprises running open-weight models.