Biren Technology, MoKe Shensi and Step Build the First Optoelectronic AI Product

MoKe Shensi, an optical chip startup, Biren Technology, an AI accelerator maker, and the AI company Step signed a strategic agreement in Beijing. The partners jointly unveiled the world's first AI product that combines optical and electrical chips in a single system, along with a platform for optoelectronic AI computing.

AI-processed from 36Kr (36氪); edited by Hamidun News
Biren Technology, MoKe Shensi and Step Build the First Optoelectronic AI Product
Source: 36Kr (36氪). Collage: Hamidun News.
Listen to article
Article Fact Sheet & Key Figures
ParameterDetails
Primary source36Kr (36氪) ↗
Editor-in-Chief
Key metric #1The world's first AI product in the intelligent computing category: a commercial system in which optical and general-purpose electrical components work together

In July 2026 in Beijing, three Chinese technology companies signed a strategic cooperation agreement: MoKe Shensi (每刻深思), a startup working on spatial optical chips; Biren Technology (壁仞科技), an AI accelerator maker; and Step (阶跃), an AI company. The partners jointly unveiled the world's first AI product that combines optical chips and general-purpose electrical accelerators in a single computing system, along with a platform for optoelectronic AI computing.

What is an optoelectronic AI architecture

Today the vast majority of AI computation runs on GPUs, graphics processors built on electrical circuits. NVIDIA's dominance in this segment is so strong that access to its chips largely determines the pace of AI development in different countries.

Optical computing takes a fundamentally different approach: data is processed with photons, particles of light, instead of electrons. This lets a system perform matrix multiplications, the core workload of neural networks, at the speed of light and with significantly lower power consumption than conventional GPUs.

The historical barrier has been the limited programmability of optical circuits: they handle fixed matrix operations well but lag behind electronics in flexibility. A hybrid approach, in which the optical subsystem takes on the heavy computation while the electrical one handles control and adaptation, gets around this limitation. MoKe Shensi's architecture is built on exactly this idea.

What the three partners announced

At the signing ceremony in Beijing, the alliance presented two joint products:

  • An optoelectronic AI computing platform: a single solution that combines MoKe Shensi's optical chips and Biren Technology's electrical accelerators in a shared hardware and software environment
  • The world's first AI product in the intelligent computing category: a commercial system in which optical and general-purpose electrical components work together

The three participants cover different links in the chain. MoKe Shensi develops photonic processors. Biren Technology makes electrical AI accelerators and has experience building closed computing ecosystems. Step trains and runs large language models, and will act as the practical tester of the new platform.

Why this matters for China

US export restrictions in recent years have effectively cut China off from advanced silicon GPUs, above all NVIDIA's A100 and H100 classes. That created an acute shortage of computing capacity and pushed local players to look for alternative paths.

Optical components are made under different manufacturing standards and fall under US sanctions to a much lesser degree. A successful hybrid optoelectronic platform would open a new technological corridor for China, one independent of the restrictions that govern shipments of silicon GPUs.

What this means

Optical computing has been discussed in academic circles since the 1980s, but the technology has not previously reached the level of commercial AI products. The agreement between MoKe Shensi, Biren Technology and Step is the first publicly announced case of a commercial AI product built on a hybrid optoelectronic architecture.

If the platform lives up to expectations, it could open a new class of AI infrastructure. That would matter both for China, which is pursuing technological independence, and for other markets interested in AI computing that does not depend on the dominant Western platforms.

Key Questions Answered

What is an optoelectronic AI architecture?

Today the vast majority of AI computation runs on GPUs, graphics processors built on electrical circuits. NVIDIA's dominance in this segment is so strong that access to its chips largely determines the pace of AI development in different countries. Optical computing takes a fundamentally different approach: data is processed with photons, particles of light, instead of electrons. This lets a system perform matrix multiplications, the core workload of neural networks, at the speed of light and with significantly lower power consumption than conventional GPUs. The historical

What the three partners announced?

At the signing ceremony in Beijing, the alliance presented two joint products: An optoelectronic AI computing platform: a single solution that combines MoKe Shensi's optical chips and Biren Technology's electrical accelerators in a shared hardware and software environment The world's first AI product in the intelligent computing category: a commercial system in which optical and general-purpose electrical components work together The three participants cover different links in the chain. MoKe Shensi develops photonic processors. Biren Technology makes electrical AI accelera

Why this matters for China?

US export restrictions in recent years have effectively cut China off from advanced silicon GPUs, above all NVIDIA's A100 and H100 classes. That created an acute shortage of computing capacity and pushed local players to look for alternative paths. Optical components are made under different manufacturing standards and fall under US sanctions to a much lesser degree. A successful hybrid optoelectronic platform would open a new technological corridor for China, one independent of the restrictions that govern shipments of silicon GPUs.

ZK
Zhemal Khamidun — CPO at Alpina Digital, product manager of AlpinaGPT. Building AI products since 2019. Hamidun News is a pet project that grew into a production aggregator with 6,000+ readers.
✦ Private AI Builders Community · Vibe Club

Ship AI Faster with Founders, Engineers & Weekly Architecture Reviews

Inside Vibe Club (hamidun.com/club) — weekly live masterminds with Zhemal Khamidun, Tech Radar, full LMS access, and 370+ production Claude Code skills.

Join Vibe Club →Explore Hamidun Academy
What do you think?
Loading comments…