Cerebras Systems

Hagay Lupesko foreshadowed an important reality of operating at scale: “If something can fail, it will fail. And it will fail at the worst possible time. Your cloud region will go down, your CDN will go down with it, your own software will fail in ways you never imagined, and every hardware device you rely on will eventually fail.” In his keynote at AI_dev Europe 2025 (Open Source GenAI & ML Summit by The Linux Foundation), Hagay, SVP of Engineering, shared how Cerebras scaled from zero to 50 exaFLOPS of inference compute in under a year and the reliability, observability, and operational lessons learned along the way. 🔮 Full video here: https://lnkd.in/grjcM_jt

The Cerebras Mindset: “You can’t be 1–2× better. You have to be 10 or 20× better.”

Thank you to Ivo Bolsens and the Stanford SystemX Alliance for inviting me to speak at the seminar today. To learn more about how the Cerebras Systems wafer scale architecture uniquely enables a new generation of instant and realtime AI, and sign up to try it for yourself, go to https://www.cerebras.ai/.

👻 Meet REAP: the Reaper for redundant experts.

Cerebras is now powering Cognition's latest code retrieval models directly in Windsurf Context retrieval has been one of the biggest bottlenecks in agentic coding. When you ask an agent to work on a large codebase, it can spend 60% of its time just searching for relevant files. This retrieval process not only keeps you waiting, but pollutes the context window with irrelevant snippets and racks up your inference bill. Cognition trained two specialized, compact models using RL and deployed them on Cerebras for maximum speed. Querying huge ~1M line codebases like React, Vercel, and PyTorch, swe-grep finds relevant code snippets in a few seconds, compared to minutes on Claude and Cursor. End-to-end retrieval, reasoning, and summarization runs ~5x faster than today’s leading models. Cognition’s Fast Context is available now in Windsurf Cascade. Cerebras is proud to be putting our wafer to work for every Cognition user!

🧮 What does 8×7B actually mean? It is NOT 8 experts with 7B active parameters per token. Turns out it’s actually 13B active parameters. But wait — where does 13B come from? In the world of Mixture of Experts (#MoE), even the simplest questions get surprisingly complex: ❓ How much storage do you actually need? ❓ How much compute does that translate to? ❓ What are the real bottlenecks — memory, compute, or communication? ⁉️ And how does Cerebras solve GPU bottlenecks? If you’ve ever tried to make sense of MoE math, our next post in the MoE 101 guide by Daria Soboleva (and interactive calculator) breaks it all down. https://lnkd.in/g7N6dh69

The energy this week in Dubai has been electric here at GITEX GLOBAL Largest Tech & Startup Show in the World. We’ve been connecting with innovators, founders, and leaders who are shaping the next wave of AI! From global giants to fast-growing startups, one thing is clear: everyone wants to go faster in AI. 𝗖𝗲𝗿𝗲𝗯𝗿𝗮𝘀 is here to deliver with inference that is 𝟮𝟬𝗫 𝗳𝗮𝘀𝘁𝗲𝗿 𝘁𝗵𝗮𝗻 𝗚𝗣𝗨𝘀. Thank you to everyone who stopped by our stand to meet with Andrew Feldman and our team. Your energy inspires us to keep pushing what’s possible.

Our Oklahoma City datacenter was designed around water, not air. To keep our wafer-scale systems running at peak performance, we use a Tier III, 6,000-ton chilled-water plant inside a 100,000 sq ft F5-rated facility. Unlike traditional air-cooled systems, our chilled-water design moves heat quietly and efficiently, even under the heaviest AI workloads. It keeps hundreds of millions of cores perfectly balanced as they drive real-time inference at massive scale. A closed-loop system recycles and stores water, maintaining stable cooling without drawing on external sources...even when winds might hit 318 mph. 🌪️ For AI infrastructure at this level, water-based thermal management isn’t just smart, it’s essential.  It’s how we push the boundaries of speed, efficiency, and reliability every single day.

This is what the fastest AI inference on Earth looks like. Live from Dubai at GITEX GLOBAL Largest Tech & Startup Show in the World and Expand North Star 2025 🚀 Blazing Fast Inference takes center stage Andrew Feldman took the stage to discuss Cerebras Inference delivers breakthrough speed and scale...powering next-gen applications from code generation to agentic AI. 🎉 Packed House At the Booth Developers, enterprises, and world leaders alike are seeing firsthand what’s possible when inference runs 20x faster than GPUs. 🔥 Just Getting Started It’s only the first few days, and the energy is incredible. GITEX is where the world gathers to shape the future of AI, and Cerebras is here to help build it. Stop by Hall 3, Booth H3-B12

Cerebras is now available via Pay Per Token It's here - the world's fastest AI inference available via self-serve. For $10, you get instant access to leading open-weight models like Qwen3 235B, GPT OSS 120B, and Qwen3 Coder 480B—running at 20x the speed of GPU-based model providers. Get started at cloud.cerebras.ai