Skip to content
No PriorsNo Priors

The Story Behind Cerebras’ $63 Billion IPO with Founder and CEO Andrew Feldman

Companies in Silicon Valley from Nvidia to AMD are racing to fuel the AI revolution with postage stamp-sized AI chips. Meanwhile, a chip the size of a dinner plate just fueled a $63 billion IPO for Cerebras. Elad Gil and Sarah Guo sit down with Cerebras founder and CEO Andrew Feldman to discuss the company’s journey to making one of the largest tech go-publics in history. Andrew details the multi-year journey of pioneering wafer-scale AI computing, including surviving a brutal period of being ahead of market demand. He also explains the engineering breakthroughs that led to delivering inference speeds at 20x that of standard GPUs. Andrew then shares how a remarkable $20 billion deal with OpenAI came together in only four weeks. Plus, Andrew’s thoughts on why architecting the future of AI requires the fortitude to be a “professional David” against the Goliaths of tech. Sign up for new podcasts every week. Email feedback to show@no-priors.com Follow us on Twitter: @NoPriorsPod | @Saranormous | @EladGil | @andrewdfeldman | @Cerebras Chapters: 00:00 – Cold Open 00:41 – Andrew Feldman Introduction 00:48 – Cerebras’ Evolution 02:17 – Wafer-Scale Bet Pays Off 06:38 – Challenges and Breakthroughs 08:37 – Crossing the Market Chasm 10:38 – Scaling Software and Hardware 12:03 – Relevance of AI-Generated Coding 13:31 – Leadership and Hiring Culture 17:16 – When to Quit vs. Persist 19:40 – Why Cerebras Went Public 22:57 – The OpenAI Deal 25:54 – Open Source and Post-Trained Workloads 27:37 – How Speed Opens Up New Business 30:07 – Conclusion

Andrew FeldmanguestElad GilhostSarah Guohost
May 21, 202630mWatch on YouTube ↗

At a glance

WHAT IT’S REALLY ABOUT

Cerebras’ wafer-scale gamble pays off with fast inference demand surge

  1. Cerebras built a wafer-scale “dinner plate” chip architecture to achieve ~15–20x faster inference than GPUs, betting that AI would eventually demand extreme speed at scale.
  2. The company endured a high-burn, high-uncertainty engineering stretch (2017–2019) before successfully yielding wafer-scale hardware, followed by years of being “ahead of the market.”
  3. Demand inflected when AI became daily-useful in 2025, making latency and throughput decisive and turning Cerebras’ speed into a mainstream product advantage.
  4. A sovereign partner (G42) provided a crucial bridge across the hardware commercialization chasm with large orders and real-world scale testing, positioning Cerebras for later hyperscaler and model-provider deals.
  5. Feldman argues going public reduced cost of capital, increased credibility, and offered a rare “AI pure play” equity story, while emphasizing culture, hiring bar, and fearless execution as scaling risks.

IDEAS WORTH REMEMBERING

5 ideas

Radical performance gains usually require radical architecture changes.

Feldman argues you don’t get 15–20x improvements via “minor modifications” to incumbents; Cerebras’ wafer-scale approach was intentionally non-derivative and initially viewed as “impossible.”

Being early can look like being wrong—until the workload becomes daily-critical.

Cerebras had working, extremely fast systems before the market cared; once AI became embedded in everyday workflows, “slow inference” became economically unacceptable and demand surged.

Hardware companies often must start in HPC to prove value before mainstream volume arrives.

Early wins at national labs and similar environments validated speed despite immature software, but those markets alone don’t provide the unit volume needed for broad adoption.

A “bridge customer” can be existential for crossing the hardware scaling gap.

G42’s billion-dollar order enabled supply-chain transformation and large-scale battle testing—capabilities Cerebras needed to be ready when OpenAI and AWS came calling.

Scaling hardware is constrained by real-world lead times, not just ambition.

Doubling output requires manufacturing lines, buildings, power, test fixtures, and coordination; Cerebras targets ~10x manufacturing growth in a year, which Feldman frames as near-historic pace.

WORDS WORTH SAVING

5 quotes

We'd built a really, really fast machine, and for a long time nobody cared.

— Andrew Feldman

How big is the market for slow search? It's zero. How big is the market for dial-up internet? It's zero. That's how big the market for slow inference will be.

— Andrew Feldman

You're having board meetings every six weeks saying, "I, I can't build it. No, still not working."

— Andrew Feldman

We would much rather fail in pursuit of the extraordinary than succeed in the ordinary.

— Andrew Feldman

The slippery slope is a beast.

— Andrew Feldman

Wafer-scale chip design vs. GPUsInference speed as the key product wedgeLong R&D valley: yield, reliability, and compiler maturityCrossing the commercialization chasm (HPC → sovereign → hyperscalers)OpenAI and AWS partnerships and deal velocityScaling hardware manufacturing and supply chain constraintsLeadership psychology: persistence vs. quitting; hiring and culture

High quality AI-generated summary created from speaker-labeled transcript.

Get more out of YouTube videos.

High quality summaries for YouTube videos. Accurate transcripts to search & find moments. Powered by ChatGPT & Claude AI.

Add to Chrome