C600 GPU by MetaX: status and timeline

C600 GPU

Provide Chinese AI labs and cloud providers with a high-performance domestic GPU alternative unconstrained by US export controls on Nvidia accelerators.

DeployStage 4 of 5

Pilot production completed late 2025; full mass production rolling out through 2026.

Updated Q4 2025Checked 11 Oct0 updates this week

Milestones

Next · C600 mass production

C600 announced2025Complete.
C600 pilot productionQ4 2025Complete.
C600 mass productionTarget Q1 2026Current milestone.
Hong Kong IPO expectedTarget Q4 2026Not yet reached.

Most important updates

MetaX begins C600 pilot production; mass production planned for 2026Q4 2025

Upcoming

Q1 2026C600 mass production (next)
Q4 2026Hong Kong IPO expected
2027C700 introduction

Current obstacles

Software ecosystem fragmentationDeep learning frameworks must add explicit C600 support; many companies maintain CUDA codebases, requiring parallel development for C600.
Manufacturing capacity rampTSMC N5 process access is limited; scaling from pilot to mass production must compete with demand from other fabless semiconductor companies.

Physics limits

Manufacturing node leadershipTSMC N5 is no longer cutting-edge; advanced processes require investment in next-gen photolithography and infrastructure, slowing iteration speed.
HBM supply constraintsHBM3e production is limited by a small number of suppliers (SK Hynix, Samsung, Micron); any supply shortage directly limits GPU production volumes.

How it works

2 parts

Memory Hierarchy

HBM3e bandwidth optimization

HBM3e stacks memory directly on the chip, providing 4-5x higher bandwidth than GDDR memory, reducing compute bottlenecks during training.

Precision Modes

Multi-precision arithmetic

FP8 precision reduces memory and bandwidth by 75% versus FP32, accelerating training while maintaining model accuracy.

Spec sheet

Spec	C600 GPU
Memory Interface	HBM3e^R (reported)
Supported Precision	Float32, Float16, BFloat16, TFloat32, FP8^R (reported)
Thermal Design Power	Lower than H100^E (estimated)

^R reported by the company · ^E our estimate

Update log

1 update

Q4 2025

Major: MetaX begins C600 pilot production; mass production planned for 2026FilingManufacturing

About MetaX

The team behind C600 GPU

MetaX

Chinese GPU, C600, C700, STAR listing

MetaX is a Chinese GPU chip designer founded in 2020 by former AMD executives. The company develops AI accelerator GPUs, primarily the C-series line, positioned as alternatives to restricted foreign chips. It listed on Shanghai's STAR Market in December 2025 and is pursuing a Hong Kong IPO.

MetaX develops C-series AI GPUs free from US export restrictions, a key advantage in the Chinese AI market.
Listed on Shanghai STAR in Dec 2025 at 700% debut jump; 2025 revenue grew 121% to ¥1.6B (US$230M).
C600 entered pilot production in Q4 2025 with mass production ramping through 2026.

metax.com.cn (opens metax.com.cn)Company profile →

Founded: 20206 yrs
Headquarters: China
Status: Public
Valuation: $3.4Best. market capMar 2026estimate
Staff: ~250Mar 2026
Last round: STAR Market IPO · $600MDec 2025
Works in: ComputingAI accelerators & CPUs
Coverage: 1 program · 3 updateslatest 1 Mar 2026checked 11 Oct

Spec

C600 GPU

Memory Interface

HBM3e^R (reported)

Supported Precision

Float32, Float16, BFloat16, TFloat32, FP8^R (reported)

Thermal Design Power

Lower than H100^E (estimated)