Cerebras

Cerebras Systems is an AI hardware company founded in 2016, known for building the world's largest computer chips. The company's Wafer-Scale Engine (WSE-3) is a single chip that occupies an entire silicon wafer — containing 4 trillion transistors and 900,000 AI-optimized cores, making it roughly 50 times larger than the largest GPU.

Cerebras's CS-3 systems using the WSE-3 are designed for AI training and inference workloads that benefit from massive on-chip memory and bandwidth, eliminating the communication bottlenecks that occur in multi-GPU clusters. The company also operates Cerebras Inference, a cloud service offering some of the fastest inference speeds available — delivering over 1,800 tokens per second for Llama models, significantly faster than GPU-based alternatives.

Cerebras raised approximately $2.9 billion across 8 rounds privately before going public in May 2026, completing the largest US tech IPO of the year: 28 million shares priced at $185 (above the $115 to $160 range) for roughly $5.5 billion in proceeds, with the stock more than doubling on its first day of trading to close at $311, which valued the company at roughly $66 billion at that closing price. The S-1 named OpenAI, Group 42, Saudi Arabia's MBZUAI (Mohamed bin Zayed University of Artificial Intelligence), and Amazon Web Services as top customers, and Cerebras swung to profitability on $510 million of 2025 revenue. OpenAI is one of Cerebras's largest customers under a multi-year contract worth more than $10 billion signed in January 2026, and holds a $1 billion secured loan plus warrants for over 33 million shares — making OpenAI a meaningful post-listing shareholder. That relationship became publicly visible in the product in August 2026, when OpenAI launched an Ultrafast service tier that serves its GPT-5.6 Sol model from Cerebras wafer-scale hardware rather than GPUs, reaching up to 750 output tokens per second. Cerebras reports a 5.6-times end-to-end speedup on the GDP-Val benchmark with no quality loss. The tier launched in limited preview in the OpenAI API, and it is the clearest demonstration so far of Cerebras serving a frontier lab's flagship model rather than open-weight alternatives. The unconventional approach to AI computing challenges the dominant NVIDIA GPU paradigm, offering an alternative architecture that excels at high-bandwidth answer-inference workloads where on-chip SRAM gives Cerebras a practical latency advantage over HBM-based GPUs. Other major customers include AWS, Meta, Mistral, Perplexity, Mayo Clinic, US defense agencies, and pharmaceutical and research labs that need extreme computational throughput.

Cerebras went public on Nasdaq on May 14, 2026 in the largest US technology listing since Snowflake, pricing 28 million shares at $185 and raising roughly $5.5 billion. On August 18, 2026 it launched the CS-4 rack system, which reaches its speed gains by overclocking the existing wafer-scale engine from 1.4 to 2.8 gigahertz and packing three wafers per rack rather than by fabricating new silicon — a genuine improvement that does not raise the on-wafer memory ceiling that forces frontier models across multiple systems. Alongside it, AMD joined OpenAI as a named partner in a disaggregated inference design that splits serving between vendors, with AMD graphics processors handling the prefill phase and Cerebras handling token generation.