Physix Frontier · News Briefing Card (ifanr · Nov 20, 2026)

Li Auto's in-house Mach M100 chip debuts at TSMC forum, now in mass production

KEY FACTS

  • On November 17, Li Auto presented its in-house Mach M100 chip at the TSMC OIP Forum.
  • The Mach M100 uses a 5nm automotive-grade process, delivers 1280 TOPS per chip, and achieves compute utilization above 82%.
  • The chip entered mass production and shipped in vehicles including the new Li Auto L9 in May this year.
  • OTA 9.2, released on November 6, added Mach VLA 2.5 and deployed an in-house 3D ViT vision encoder.
  • In October, Li Auto's chip subsidiary Mach Core closed a 1.625 billion yuan all-equity funding round at a post-money valuation of 17.1 billion yuan.

KEY DATA

1280 TOPSPer-chip compute
Above 82%Compute utilization
300 metersForward perception range
17.1 billion yuanMach Core post-money valuation

PHYSIX OBSERVATION

Li Auto pulls in-house chip development away from the spec race and back to the experience layer: its dynamic dataflow architecture raises compute efficiency so the VLA large model runs faster on-device, and what users perceive is more decisive driving decisions. More noteworthy is that the Mach M100 is positioned as an on-device general-purpose AI chip, and combined with the subsidiary's funding, Li Auto is extending from an automaker toward a physical-AI compute supplier. If this path works, it will rewrite the value ceiling of automakers' in-house chips.

Source: ifanr report