OpenAI signs deal, worth $10B, for compute from Cerebras

Good news! Chip competition is good, more chip competition is better!

"... OpenAI announced Wednesday [14/1/2026] that it had reached a multi-year agreement with AI chipmaker Cerebras. The chipmaker will deliver 750 megawatts of compute to the AI giant starting this year and continuing through the year 2028, Cerebras said. ..."

"OpenAI and Cerebras have signed a multi-year agreement to deploy 750 megawatts of Cerebras wafer-scale systems to serve OpenAI customers. This deployment will roll out in multiple stages beginning in 2026, making it the largest high-speed AI inference deployment in the world. ..."

"Cerebras builds purpose-built AI systems to accelerate long outputs from AI models. Its unique speed comes from putting massive compute, memory, and bandwidth together on a single giant chip and eliminating the bottlenecks that slow inference on conventional hardware. ..."

OpenAI signs deal, worth $10B, for compute from Cerebras | TechCrunch

OpenAI Partners with Cerebras to Bring High-Speed Inference to the Mainstream (original news release)

OpenAI partners with Cerebras (original news release) "OpenAI is partnering with Cerebras to add 750MW of ultra low-latency AI compute to our platform."

Credits: Last Week in AI

A gigantic chip indeed!

Giant Chips for Supercomputers

Should we not call these chips bricks? How to make this "chip" fit in my smartphone? 😊

"... the company has been squeezing as many processors as it can onto one giant wafer. The main advantage is in the interconnects—by wiring processors together on-chip, the wafer-scale chip bypasses many of the computational speed lossesthat come from many GPUs talking to each other, as well as losses from loading data to and from memory. ..."

Giant Chips Give Supercomputers a Run for Their Money - IEEE Spectrum Cerebras’s wafer-scale chips excel at molecular dynamics and AI inference

Cerebras' second-generation Wafer-Scale Engine (WSE-2) is a massive chip tailored for AI applications

Cerebras WSE-3: Third Generation Superchip for AI with 4 trillion transistors

Impressive! Mind boggling!

"AI supercomputer firm Cerebras says its next generation of waferscale AI chips can do double the performance of the previous generation while consuming the same amount of power. The Wafer Scale Engine 3 (WSE-3) contains 4 trillion transistors, a more than 50 percent increase over the previous generation thanks to the use of newer chipmaking technology. The company says it will use the WSE-3 in a new generation of AI computers, which are now being installed in a datacenter in Dallas to form a supercomputer capable of 8 exaflops (8 billion billion floating point operations per second). Separately, Cerebras has entered into a joint development agreement with Qualcomm that aims to boost a metric of price and performance for AI inference 10-fold. ..."

"Key Specs:

• 4 trillion transistors
• 900,000 AI cores
• 125 petaflops of peak AI performance
• 44GB on-chip SRAM
• 5nm TSMC process
• External memory: 1.5TB, 12TB, or 1.2PB
• Trains AI models up to 24 trillion parameters
• Cluster size of up to 2048 CS-3 systems

"

Cerebras WSE-3: Third Generation Superchip for AI - IEEE Spectrum Which will power an 8-exaflop AI supercomputer

Cerebras Systems Unveils World’s Fastest AI Chip with Whopping 4 Trillion Transistors Third Generation 5nm Wafer Scale Engine (WSE-3) Powers Industry’s Most Scalable AI Supercomputers, Up To 256 exaFLOPs via 2048 Nodes

Cerebras unveils world's largest AI training supercomputer with 54M cores

Sheer mind boggling! The potential for progress and advances is enormous!

"... Rather than make individual chips for its centralized processing units (CPUs), Cerebras takes entire silicon wafers and prints its cores on the wafers, which are the size of pizza. These wafers have the equivalent of hundreds of chips on a single wafer, with many cores on each wafer. And that’s how they get to 54 million cores in a single supercomputer. ..."

"Sunnyvale, CA— July 20, 2023 – Cerebras Systems, the pioneer in accelerating generative AI, and G42, the UAE-based technology holding group, today announced Condor Galaxy, a network of nine interconnected supercomputers, offering a new approach to AI compute that promises to significantly reduce AI model training time. The first AI supercomputer on this network, Condor Galaxy 1 (CG-1), has 4 exaFLOPs and 54 million cores. Cerebras and G42 are planning to deploy two more such supercomputers, CG-2 and CG-3, in the U.S. in early 2024. With a planned capacity of 36 exaFLOPs in total, this unprecedented supercomputing network will revolutionize the advancement of AI globally. ..."

Cerebras unveils world's largest AI training supercomputer with 54M cores | VentureBeat

Cerebras and G42 Unveil World’s Largest Supercomputer for AI Training with 4 exaFLOPs to Fuel a New Era of Innovation (company press release)

Cerebras Systems sets record for largest AI models ever trained on one device

We are coming closer to a brain on chip! What a gigantic chip! Probably it does not yet fit in a typical smartphone! 😄

This could be a game changer!

"... for the first time ever, the ability to train models with up to 20 billion parameters on a single CS-2 system – a feat not possible on any other single device. By enabling a single CS-2 to train these models, Cerebras reduces the system engineering time necessary to run [train] large natural language processing (NLP) models from months to minutes. It also eliminates one of the most painful aspects of NLP — namely the partitioning of the model across hundreds or thousands of small graphics processing units (GPU)."

Cerebras Systems sets record for largest AI models ever trained on one device | VentureBeat

Cerebras Systems Sets Record for Largest AI Models Ever Trained on A Single Device Single CS-2 System trains multi-billion parameter NLP models including GPT-3L 1.3 Billion, GPT-J 6 Billion , GPT-3 13 Billion and GPT-NeoX 20 Billion; provides simple setup, faster training and enables switching between models with a few keystrokes

Cerebras' wafer-size chip is 10,000 times faster than a GPU

Wow! Could this be a major breakthrough in chip technology? Wafer scale technology seems to be a very different paradigm in microprocessor technology!

"On a practical level, this means AI neural networks that previously took months to train can now train in minutes on the Cerebras system. ... But Cerebras ... takes that wafer and makes a single, massive chip out of it. Each piece of the chip, dubbed a core, is interconnected in a sophisticated way to other cores. The interconnections are designed to keep all the cores functioning at high speeds so the transistors can work together as one. ... 
The CS-1 beat the Joule Supercomputer [which is No. 82 on a list of the top 500 supercomputers in the world] at a workload for computational fluid dynamics, which simulates the movement of fluids in places such as a carburetor. The Joule Supercomputer costs tens of millions of dollars to build, with 84,000 CPU cores spread over dozens of racks, and it consumes 450 kilowatts of power. In this demo, the Joule Supercomputer used 16,384 cores, and the Cerebras computer was 200 times faster, ... Cerebras costs several million dollars and uses 20 kilowatts of power. ..."

Cerebras' wafer-size chip is 10,000 times faster than a GPU | VentureBeat

Most Powerful AI Chip Yet Released

Posted: 8/24/2019  Updated: 8/30/2019

Update Of 8/30/2019

The large size of this chip (215 mm on each side) could indicate a major paradigm shift in chip production! In the past, the major trend was to make chips more compact and dense on ever tinier scales.

However, “... Cerebras argues that the world’s semiconductor companies have spent decades developing ever tinier chips that can then be bundled together to create super-powerful processors. But hooking lots of small chips together creates latencies that slow down training of AI models—a huge industry bottleneck. In contrast, an extra giant chip could shift data between processing and memory incredibly fast and turbocharge AI applications” (source; emphasis added)

Original Post

Could this be a game changer we have been waiting on? AI & machine learning have a serious bottleneck (limiting factor) and that is the computing hardware itself. Better and more powerful chips are urgently needed. 

This is a giant of a WSE (Wafer Scale Engine) chip measuring 215 mm each side or 46,225 mm2 and featuring 1.2 Trillion transistors and 400,000 AI-optimized cores! “By comparison, the largest Graphics Processing Unit is 815 mm2 and has 21.1 Billion transistors.” (S2) This chip was just released by Cerebras Systems (not exactly one of the familiar names in the chip making industry, but this may change).

Sources (S):

1. Cerebras Systems unveils a record 1.2 trillion transistor chip for AI (8/19/19)
2. Cerebras Systems