Physix Frontier · News Briefing Card (QbitAI · Nov 16, 2026)

Tongxin Intelligent raises hundreds of millions in six months for AGI inference chips

KEY FACTS

  • Tongxin Intelligent was initiated by Zhu Songchun and incubated by the Beijing Institute for General Artificial Intelligence, and was founded in April this year.
  • The company closed seed and angel rounds within six months, raising hundreds of millions of yuan in total, with the funds going into pre-tapeout architecture design.
  • Its first-generation product is positioned as a cost-effective inference chip, covering cloud large models, world models, and embodied-intelligence edge devices.
  • The company completed architecture design and project approval three months after founding, and has now entered the chip design phase.
  • FPGA prototype validation will begin in early December, and the chip is planned with three types of compute units: Vector, Tensor, and AGI Core.

KEY DATA

hundreds of millions of yuanCombined seed and angel round funding
3 monthsTime from founding to completion of architecture design and project approval
about 1.17 trillion yuanForecast China AI inference chip market size in 2030
about 8.2 billion USDAmount AMD plans to acquire World Labs for

PHYSIX OBSERVATION

Tongxin presses Zhu Songchun's CUV theory directly into chip architecture, attempting to open dedicated units for graph reasoning, causality, and value computation beyond the GPU; this is a fight over the right to define, not a fight over compute. But the cycle from theory to mass production is long, and the first-generation product still has to win customers by supporting mainstream models; the real test is whether AGI Core can deliver its efficiency advantage on real tasks.

Source: QbitAI report