AI Mornings with Andreas Vig
Your daily AI news briefing in under 10 minutes. New models, product launches, research breakthroughs, and industry shifts, explained clearly, no hype.
AI Mornings with Andreas Vig
Cerebras' $26B IPO & Sierra's $950M Enterprise AI Play
Use Left/Right to seek, Home/End to jump to start or end. Hold shift to jump forward or backward.
Hey, welcome to AI Mornings with Andreas Vig. It's Monday, May 5th, and we've got a big funding and IPO story kind of day. Let's start with the biggest money story of the day. AI chipmaker Cerebro Systems just filed to go public in what's shaping up to be the largest tech IPO of 2026 so far. They're planning to sell 28 million shares at $115 to $125 each, which would raise $3.5 billion and value the company at up to $26.6 billion at the high end. And here's the thing banks are already fielding about $10 billion in orders for those shares, suggesting they might price even higher than the announced range. Cerebrus makes the Waferscale Engine 3, an AI-specific chip that goes after the GPU market by claiming faster inference with lower power consumption. But what makes this IPO really interesting is the OpenAI connection. Sam Altman, Greg Brockman, Ilya Sutzkeva, and Adam D'Angelo all invested in Cerebrass as angel investors. More significantly, OpenAI loaned Cerebras $1 billion in December, secured by warrants that let OpenAI buy over 33 million shares. OpenAI also signed a multi-year agreement worth more than 10 billion with them. So this IPO is going to be a win for OpenAI in multiple ways. Speaking of funding rounds, Brett Taylor's enterprise AI startup Sierra just announced a $950 million round led by Tiger Global and GV, pushing their valuation above $15 billion. Sierra is one of those enterprise AI plays that's actually showing real traction. They claim over 40% of the Fortune 50 as customers. And they hit 150 million in annual recurring revenue by February after being at 100 million just a few months earlier. They recently launched something called Ghost Rider, which is an agent as a service tool that builds other agents from natural language descriptions. Taylor, who's also OpenAI's chairman and was previously co-CEO of Salesforce, is betting that the future means employees never have to navigate complex enterprise systems directly. That funding story actually came with a pretty compelling enterprise AI data point from Uber. Their CTO said that 10% of all code at the company is now generated autonomously by AI agents across 8,000 engineers and technical workers. They also ran a proof of concept where one team built a hotel booking integration using only agentic workflows. Work that would normally take a year got done in six months. The flip side is that Uber, quote, blew through their AI budget after opening the door to Agentic Tools, which tells you something about the real cost of these deployments before the returns kick in. Now, here's a governance story that caught my eye. YCombinator owns about 0.6% of OpenAI, which at OpenAI's current $852 billion valuation is worth over $100. The stake goes back to 2016, when OpenAI was ceded by YC Research, while Sam Altman was running Y Combinator. This matters because Paul Graham, who is one of YC's four founding partners, has been commenting on Sam Altman's character in the press. And it turns out he has billions of dollars personally at stake with OpenAI's success. That doesn't make his opinions invalid, but it's exactly the kind of thing that probably should have been disclosed when he was being quoted as a character reference. Finally, OpenAI published a technical deep dive today on how they deliver low-latency voice AI at scale to over 900 million weekly active users. They re-architected their WebRTC stack using what they call a relay plus transceiver architecture that separates packet routing from protocol termination. The interesting part is their global relay layer, which puts geographically distributed entry points close to users to reduce that first hop latency. It's a reminder that making voice AI feel conversational rather than push to talk requires serious infrastructure work under the hood. Alright, that's it for today. See you tomorrow.