DeePhi Tech / 深鉴科技

DeePhi Tech, or 深鉴科技, is Yao Song / 姚颂’s first startup in 173: 对话姚颂：深鉴、东方空间、再出发，「天才少年」十年后. The source describes its origin as a Tsinghua University / 清华大学 AI-chip demo built from complementary work in algorithms, architecture, software systems, and compiler-like tooling, with Wang Yu / 汪玉 and Han Song / 韩松 as important technical anchors.

The company is a Hard Tech Fundraising case because early investors did not know how to underwrite AI-chip risk. Yao says the team met roughly 50 institutions before learning to explain the company through market size, product need, and technical landing rather than only through technical possibility. Its later sale to Xilinx / 赛灵思 becomes a Strategic Acquirer Fit case: the acquirer needed software, algorithm, and solution capability to move from devices toward platform strategy.

Key Claims

• DeePhi’s startup timing depended on deep learning moving toward commercial application and CPU/GPU alternatives becoming more interesting.
• The source connects AI Chip Specialization to productization: a hardware acceleration demo had to become a platform, solution, and customer story.
• The company cut its smart-drone direction after Yao judged that DJI / 大疆 controlled too much of the market, then moved attention toward smart security and assisted driving.
• The acquisition decision combined valuation, strategic fit, IPO uncertainty, future risk, and team endurance rather than a single founder-liquidity motive.

Connections

• Yao Song / 姚颂, Wang Yu / 汪玉, and Han Song / 韩松 — founder and technical-origin figures.
• Xilinx / 赛灵思 — acquirer.
• AI Chip Specialization, Inference Chip Startup Narrowing, Hard Tech Fundraising, and Investor Risk Narrative — core concept links.
• Strategic Acquirer Fit and Post-Acquisition Founder Identity — acquisition interpretation.