Participated · Equity · Jun 2026

DeepSeek is known for developing high-performance large AI models and maintaining popular open-source projects that have attracted over 170,000 GitHub stars and about 5.7 billion monthly API calls. Historically self-funded by its parent, High-Flyer Capital, the firm focused on foundational model research and resisted external investment until this recent financing move. Its next-generation flagship, V4, reportedly uses a trillion-parameter MoE architecture and is being deeply adapted to Huawei's Ascend chips to reduce dependence on NVIDIA's CUDA ecosystem. The company is raising external capital to cover the enormous compute and engineering costs of V4, to establish a market valuation for employee equity incentives, and to accelerate productization and ecosystem development. DeepSeek faces engineering and tooling challenges in migrating to a non-NVIDIA stack and must translate research-led capabilities into sustainable business products post-funding.

Participated · Series B · Feb 2024

Moonshot AI was founded in 2023 by Yang Zhilin, a former Meta AI and Google Brain researcher. The lab gained prominence with its open-weight Kimi models—Kimi K2.5 notably drove rapid adoption and Kimi K2.6 became one of the most-used LLMs on the OpenRouter distribution platform. Moonshot competes with models from OpenAI, Google, Anthropic, ByteDance, Alibaba, Zhipu and other Chinese AI labs. The company reported annual recurring revenue exceeding $200 million in April, fueled by paid subscriptions and API usage. Moonshot attracted significant investor support prior to this round, having raised about $3.9 billion over the prior six months and counting backers such as Alibaba, Tencent, HongShan, ZhenFund, IDG Capital and 5Y Capital. Its valuation progressed from about $4.3 billion at the end of 2025 to roughly $10 billion in early 2026 before the current $20 billion valuation.