Russia registers an AI chiplet topology built on AMD and Nvidia principles
Rospatent has registered the topology of a Russian tensor chiplet, a counterpart to Nvidia's designs. The rights holder is the Foundation for Advanced Research Projects. The architecture follows the principles used by AMD and Nvidia and is built on a 28 nm process. The state can now hand out production licenses to domestic design centers and fabs.
AI-processed from CNews AI; edited by Hamidun News
Rospatent has registered the topology of a Russian tensor chiplet, a specialized accelerator for artificial intelligence workloads. The rights holder is the Russian Federation, represented by the Foundation for Advanced Research Projects (FPI). This means the state has obtained the exclusive right to distribute the architecture among Russian manufacturers.
What a chiplet is and why this is an
Nvidia-style architecture
A chiplet is a modular architecture for building powerful processors. Instead of one huge monolithic die, the system consists of several ready-made blocks (cores, cache, controllers) joined into a single whole. AMD and Nvidia use this approach to cut development costs. The Russian AI chiplet is built on the same principle.
The process node is 28 nm, which is outdated by current standards: Nvidia H100 chips run on 5 nm and deliver several times the performance. But for a first domestic design, 28 nm is a realistic choice given the technological base available.
Why the state wants its own architecture
Until now, Russian companies that needed AI accelerators either bought Nvidia hardware (when available) or copied the architecture illegally (with legal risks). Now the state can license chiplet production under its own standard. This opens up several paths:
- Domestic design centers will develop AI systems that support the Russian accelerator
- Russian fabs will gain a legal right to manufacture it
- Companies working on government contracts will be able to use domestic chips
- Exports to countries under sanctions are possible
From registration to production
Registering the architecture with Rospatent is an important but intermediate step. Going from design to volume production takes 2-3 years and requires solving numerous technological problems.
"This is a 28 nm architecture created with the available Russian and friendly-country technological base in mind,"
FPI explained.
An older process node means lower performance and higher power consumption. But it is a trade-off: access to modern process nodes requires Western equipment, which is unavailable under sanctions.
What this means
Russia is moving from being a consumer to a developer of AI chips. State backing shows the seriousness of the intent. The main challenge is actually getting production running under technological restrictions.
- Nov 1, 2026Local models lose to Opus and Codex: 0.21 on a MacBook versus 1.00 for Codex
- Nov 1, 2026Habr author fine-tunes YOLOv8l on dental X-rays and pushes mAP50 to 0.642
- Nov 1, 2026Voice bots built on LLMs turn to Jev, a fast classifier, for decisions in 200–500 ms
- Nov 1, 2026MoneyGate safety layer blocked all 42 fraudulent requests to local AI models
Key Questions Answered
What a chiplet is and why this is an Nvidia-style architecture?
A chiplet is a modular architecture for building powerful processors. Instead of one huge monolithic die, the system consists of several ready-made blocks (cores, cache, controllers) joined into a single whole. AMD and Nvidia use this approach to cut development costs. The Russian AI chiplet is built on the same principle. The process node is 28 nm, which is outdated by current standards: Nvidia H100 chips run on 5 nm and deliver several times the performance. But for a first domestic design, 28 nm is a realistic choice given the technological base available.
Why the state wants its own architecture?
Until now, Russian companies that needed AI accelerators either bought Nvidia hardware (when available) or copied the architecture illegally (with legal risks). Now the state can license chiplet production under its own standard. This opens up several paths: Domestic design centers will develop AI systems that support the Russian accelerator Russian fabs will gain a legal right to manufacture it Companies working on government contracts will be able to use domestic chips Exports to countries under sanctions are possible
What are the key details on From registration to production?
Registering the architecture with Rospatent is an important but intermediate step. Going from design to volume production takes 2-3 years and requires solving numerous technological problems. "This is a 28 nm architecture created with the available Russian and friendly-country technological base in mind," FPI explained. An older process node means lower performance and higher power consumption. But it is a trade-off: access to modern process nodes requires Western equipment, which is unavailable under sanctions.
Test Frontier LLMs & Autonomous Agents in One Workspace
Run Claude Opus 5.5, GPT-6, Gemini 3.1, and DeepSeek side-by-side in Nomad AI (nomadnet.ai) — unified model routing, deep research, and multimodal workflows under one subscription.
The AI world, distilled — once a week
Seven stories that actually mattered, hand-picked. No noise, no reposts, no press releases.
Done! Check your inbox for a confirmation.