After Cerebras's "HOT" IPO, 8 Takeaways for AI IPO Wave. ARD #76
Today’s frame: What to take away from Cerebras’s “HOT” IPO (CBRS). It’s a big topic touching on a wide range of AI issues I’ve discussed in recent months — both on the Substack and on this podcast — and there are a range of implications for you AI Curious Folk that I’d like to go over in more detail.
Thus today we’re going to focus on just this one AI news event for some deeper thoughts and takes — not the usual three sets of events and takes you’ve come to expect from ARD. Cerebras’s day-one pricing + after-market trading echoes the Internet Days, and there’s a lot in that echo worth unpacking carefully.
Cerebras Systems (ticker: CBRS) priced today on Nasdaq. The stock was priced at $185, opened at $330, and is currently around $350 — up 80-90%. They raised approximately $6 billion. The valuation went from $50-60 billion initially to over $100 billion. For someone like me — ex-Goldman partner heading the internet research effort starting in the 93-94 timeframe — this has big echoes. And to be sure, there have been a lot of AI IPOs already (CoreWeave was a big successful one with lots of volatility), but this one is a marker event for the AI IPO Mega-Wave ahead.
Cerebras is a hot AI inference chip company. I wrote about it a couple of days ago on the Substack. They compete in the same space as Nvidia (which has the lion’s share of both training and inference markets) — these are technical terms basically about training the big models from OpenAI, Anthropic, Meta AI, xAI’s Grok, Google Gemini, etc. The core issue is that there are many ways to do these techniques as model architectures evolve. Cerebras has a unique mousetrap — a very big chip that does inference with memory inserted much closer to the inference task. It’s been a long road for this company — I’ve been tracking them for years and years, way before this AI thing got hot.
MY TAKES: Now to the eight implications you should track from today’s CBRS print — every one of them tied back to threads we’ve discussed on AI-RTZ and ARD across the spring:
(1) There is a secular growth story here, not just for Cerebras, but for chips going into this stuff broadly. Lots of innovation ahead. I wrote about this in detail in AI-RTZ #1086 yesterday on why chips are really helping Software Eat the World — Mark Andreessen’s famous essay 15 years ago has been true for the most part. The core driver here: a lot of innovation in both hardware and software. Nvidia is at the forefront — Jensen Huang really orchestrates the roadmap for all the opportunities + sub-opportunities opening around AI processing. The probabilistic AI models from OpenAI, Anthropic etc. are amazing, but none of it is actionable until you have the chips — not just GPUs but everything around them: memory, networking, etc. Huge array of private and public companies will have ample opportunity. And we’re in the early days — even though AI has been around for decades, this version (transformer-based models around ChatGPT) is in its fourth year. Secular growth story is intact — this is the foundation level I keep tracking.
(2) The number of IPOs in general has really shrunk in recent years. Last year there were about 30 IPOs. Four years ago there were about 121 IPOs. A lot of tech companies — not just AI — are staying private longer. Companies like Stripe have been around for a decade or more. Today’s capital markets are very different from the Internet days. Huge amount of institutional capital, sovereign funds, family wealth funds, crossover funds — all participating early in the private rounds. By the time these companies come public, they’re at the other side of the curve in their rapid growth. Different from the dot-com era where companies were barely a company with an idea. Today’s companies have billions in revenue (maybe not billions in profit yet, since these are capital-intensive AI businesses).
(3) IPOs are generally controlled events in terms of the amount of shares available — the float — relative to total outstanding. Even more so today when valuations are eye-catching in the hundreds of billions. For Cerebras: about 30 million shares floated vs about 171 million more to be released after a 6-month lockup. Roughly one-sixth of the cap table at the open — the bulk of supply comes in waves. The float is designed to go up and to the right and grab headlines for the company. Then months later, secondaries. Cerebras itself is aggressively trying to widen its customer base. A year ago over 67% of revenues was one customer — government entities in Saudi Arabia, one of their big investors (a Sovereign Fund). Today it’s down to maybe under 30%. They’ve got to really diversify. The IPO publicity helps build business and credibility for that.
(4) Institutions are a much more active participant in this wave of AI private and public companies than they were in the Internet era. A lot more strategic investors, sovereign wealth funds, huge mega family offices. It basically covers a much bigger chunk of the alphabet. When a private company initially does a seed round, then A round, B round, C round — in the old days they would typically go public after the C or D round. Today these companies have G, H, I rounds — half a dozen or more private rounds at ever-larger valuations. Anthropic yesterday just closed another $40-50 billion at a $950 billion valuation — even though they’re absolutely prepping to go public by end of year. These companies don’t really care that much whether they fill the gas tank from the private or the public bucket. But if you’re investing through your Robinhood account, you want to be knowledgeable about these dynamics.
(5) Recognize a much deeper and wider established MEME stock environment than we had in the dot-com era. Mechanisms like Robinhood let millions and tens of millions of individual investors not just participate but talk about it through social media. So there’s waves and counter-waves. And a much larger universe of investment structures like ETFs and mutual funds — plus derivative securities that allow institutions to hedge risks. Thoughtful, mindful element to factor in.
(6) As a former analyst — and I still consider myself an analyst at the core — you want to see these companies perform publicly. It’s a very different game. The next two quarters at least — typically I like to watch at least three or four quarters — of public performance in terms of meeting and beating expectations is critical. That’s the seasoning aspect. Don’t just be enamored with the one-day, one-week pops and drops — watch the fundamental performance of these companies. Often that talks to governance, product maturity, customer demand organicity, systems to maintain rising-scale selling. Lots of details in the weeds. Just wait and watch the quarterly performance — a critical reminder when we’ve got situations like $185 priced talk opening at $330.
(7) ‘Frenemies’ dynamics. This has always been the case in tech but it’s particularly concentrated here. There are a lot of participants — especially corporate ones — that have lots of shares in this stuff. OpenAI shares Cerebras. These companies are building their separate chips as well. So there’s a lot of co-opetition dynamics in these companies. Something to track on a fundamental basis through the quarters as a public company. A much more pronounced set of realities than in prior tech waves. (See ARD #73 on corporate AI investments for the deeper read.)
(8) PASSIVE INDEX FUND DYNAMICS for mega AI IPOs ahead. Especially SpaceX (expected in June at a $1.75 trillion number), Anthropic and OpenAI (by end of year). A lot of developments here — especially SpaceX, where Elon very cleverly has been really pressuring and looks like he’s getting the major index funds (NASDAQ, S&P, etc.) to basically include his private company into the indices far earlier. Normally it takes a year of seasoning before these things are entered into the index. In SpaceX’s case, the rumor is they’ll be there in 15 days after they’re public. The reason this matters: today there are hundreds of billions — not trillions — of dollars in passive investment funds, ETFs, mutual funds that track these indices. Whatever is public + part of the index, the fund is obligated to buy. So this has been orchestrated to get billions of capital to buy in before the 6-month lockup is out — because they have to, to participate in the index. They’re non-fundamental buyers. This is a dynamic we did not have in prior markets. I wrote about this phenomenon as “We Are the Geese” — if you’ve ever thought about how foie gras is made, it’s force-fed into geese with tubes (called gavage). That’s what these big IPOs are potentially going to be doing — force-fed into passive index funds. And if you’re a long-term investor, watch this carefully. It’s on my tickler — depending on how this stuff is ultimately done, I’m actually going to take funds out of some of the indices that are including some of these companies and diversify or build my own artificial index by buying individual stocks. Use of the word “corruption” — not criminal corruption, but a corruption of the process in terms of what these structures are supposed to be. Pools of capital with certain seasonings + maturity — and that is what is being bypassed here.
MP Take: AI IPOs will make headlines all this year and next. Today is the beginning of a prominent one. Recognize these fundamental vs financial trading tail-winds and disaggregate them. The race is anyone to take. Anthropic is running hard with “Mythos.” OpenAI has their 5.5 Jagged PD model and various Codex iterations. The race continues even way past the IPOs. The IPOs are just a beginning point, not an end point. Keep that in mind.
Plus: Gadget AI — Trouble in Paradise: OpenAI vs Apple. Bloomberg reports the OpenAI/Apple partnership is fraying, setting up a possible legal fight. OpenAI is reportedly suing or considering suing Apple for not doing enough to integrate the two products — especially as Apple is racing to publicly integrate Google Gemini with Siri by WWDC next month in June. A bit of drama you’ll see a lot of details on.
MP Take: This is a ‘buyer’s remorse’ deal on the OpenAI side. Yes, they did get a coup in 2024 securing the Siri deal, but it was always non-exclusive. Apple is playing its cards well, especially at a time when OpenAI did two things in the last two years. First: $6.5 billion for Jony Ive’s vehicle, basically saying “we’re going to create our own devices, compete with Apple.” Frenemies. And all these humans, even if they’re billionaire CEOs, are emotional — you tweak someone in the nose, they tweak back, look at other alternatives. Second: OpenAI is now planning, with Jony Ive’s help, to build its own AI smartphone. No different from when Eric Schmidt from Google (CEO in the 2000s) was on Apple’s board, decided to get off, then literally announced its own Android phone with touch months before Apple announced the iPhone. These are corporate political negotiations + dynamics. Apple is playing its cards well. OpenAI seems to be using the legal card tactically. The game on AI smartphones has barely begun.
Bonus — today’s AI-RTZ companion #1086 — Why Chips Are REALLY Helping Software Eat the World — the deeper written read on why the chip side of the AI Tech Wave is the durable fundamental story underneath the financial trading tail-winds.
Closing Questions —
-
What AI wearable that has been leaked is most intriguing to me? Apple’s rumored AirPods with Cameras. It’s been out there for a while, but apparently it’s accelerated. AirPods have been a hugely successful win — Tim Cook’s legacy. This business by itself, as a standalone business, would be a Fortune 50-level company. Apple Watch is not far behind. The notion of putting cameras into AirPods is very innovative — most other companies are not doing this. It’s literally putting two human senses into one device that goes into one of our core senses (hearing), which I’d argue is one of our most important senses. Lots of technical innovations and functionality that can be created for mainstream users. Apple has a track record — music players in our ears, phones without keyboards. Could come as early as later this year. Source: Bloomberg — Apple’s Camera-Equipped AirPods.
MP Take: Cameras in AirPods are not necessarily about taking pictures. We don’t know how it’ll be deployed yet. It’s about ambient awareness data that goes into the systems for AI. And Apple being Apple — probably more privacy and trust sensitive than most competitors, especially Meta and Google. That’s a big important differentiation.
-
What AI wearable am I less excited or least enthused about today? AI Smart Glasses. Especially what’s been tried with Meta, Samsung, and others. Apple does have its own level of smart glasses. Over half the people actually wear glasses, but they’re prescription. Glasses also have a fashion element — we typically want three or four pairs for different things. That’s very different from a gadget that costs $400-500 to start (before transparency lenses, prescription costs). And the technology changes every year — I want next year’s glasses with better processors and memory. Tough segment. Standing backcat: AI-RTZ #872 — Glass Half Empty on AI Devices. And on the Meta vs Apple smart-glasses framing: AI-RTZ #1055.
MP Take: Technically cool. But too many social/privacy and tech issues to make this a global, mainstream cultural phenomenon. Too fragmented in terms of major and minor players. I won’t belabor it — but that’s why I’m less excited about glasses and more intrigued by the cameras in AirPods.
Hope you find these items thought provoking. As usual, questions and feedback welcome. Stay tuned.
More source links and clips below.
(NOTE: The discussions here are for information purposes only, and not meant as investment advice at any time. Thanks for joining us here)
Clips from today’s episode
Short — Cerebras IPO — $185 to $350 Day-One Pop Cerebras Systems (CBRS) priced today on Nasdaq. The stock was priced at $185 and opened at $330, currently $350. Up 80-90%. Raised about $6 billion. Valuation went from $50-60B initially to over $100B. For someone like me — ex-Goldman partner heading the internet research effort in the 93-94 timeframe — this has big echoes of the late-1990s Internet Days. Cerebras competes with Nvidia in the AI inference chip market.
MP Take: AI IPOs will make headlines all this year and next. Recognize the fundamental vs financial trading tail-winds. Secular growth story in chips is real. But market structure underneath is materially different from Internet days — institutions are more active across private G/H/I rounds, float is tightly controlled, and meme-stock + passive-index dynamics are much deeper. Don’t get enamored by week-one pops; watch the next 3-4 quarters of public performance for true quality.
Short — Modern AI Cos Stay Private Longer — G, H, I Rounds Now Modern AI companies are staying private much longer than dot-com era startups. In the old days, companies went public after the C or D round. Today these companies have G, H, I rounds — half a dozen or more private rounds at ever-larger valuations. Anthropic yesterday closed another $40-50B at a $950B valuation while absolutely prepping to go public by end of year. So these companies don’t really care whether they fill the gas tank from the private or public bucket.
MP Take: Different physics this cycle. There’s huge institutional capital — sovereign funds, family wealth funds, institutions, crossover funds — all participating early in the private rounds. By the time these companies come public, they’re past their fastest growth phase. Today’s mega-IPOs have billions in revenue (not necessarily profit yet — these are very capital-intensive businesses).
Short — AI IPO Float Math — 30M Shares Out, 171M Locked Up IPOs are generally controlled events in terms of the amount of shares available — the float — relative to total outstanding. It’s been the case always, and even more so today when valuations are eye-catching in the hundreds of billions. For Cerebras (CBRS): about 30 million shares floated vs about 171 million more to be released after a 6-month lockup. Roughly one-sixth of the cap table at the open. The float is designed to go up and to the right and grab headlines.
MP Take: Watch the lockup expiry calendars carefully — that’s when true long-term price discovery happens. Cerebras is aggressively widening their customer base. A year ago over 67% of revenues was one customer — government entities in Saudi Arabia (Sovereign Fund investor). Today maybe under 30%. They’ve got to diversify. The IPO publicity helps build business and credibility for that.
Short — Internet-Days Echoes — The AI IPO Wave Has Begun Cerebras CBRS pricing today is a marker event for the AI IPO Mega-Wave ahead. We’ve already had a bunch of AI IPOs — CoreWeave was a big successful one with lots of volatility up and down. This is on the lower end of the tech stack. There have also been AI IPOs further up in the application layer. We’re going to see a lot more of these — SpaceX/xAI expected June at $1.75 trillion, Anthropic and OpenAI by end of year.
MP Take: For someone like me — ex-Goldman partner heading the internet research effort starting in the 93-94 timeframe — this has big echoes. But with materially different market structure underneath. Institutions are more active. Float is tightly controlled. MEME-stock environment is much deeper. Passive-index fund dynamics are gameable in ways that didn’t exist in the 1990s. The race continues way past the IPOs.
Short — AirPods with Cameras — Two Senses in One Device Most intriguing leaked AI wearable from my perspective: Apple’s rumored AirPods with Cameras. AirPods are already Fortune 50-scale as a standalone business. Apple Watch is not far behind. The notion of putting cameras into AirPods is innovative — most other companies aren’t doing this. It’s literally putting two human senses into one device that goes into one of our most important core senses: hearing. Apple has a track record of making unusual tech habits cool at scale.
MP Take: It’s not necessarily about taking pictures. It’s about ambient awareness data that goes into the systems for AI. And Apple being Apple — likely more privacy and trust sensitive than most competitors, especially Meta and Google. That’s a big differentiation. Could ship as early as later this year.
Ep 76 scope: 1 Main + 5 Shorts (vs the usual 4) — MP’s variant pre-recording scope today. No Segments or Hooks.
About AI Ramblings Daily (ARD), and AI-RTZ
Both are daily. Both are free. Both are about AI. But they’re different mediums carrying different messages.
AI-RTZ is the morning text — a deeper written take on one idea, published by at least 5 AM EST. Today: post #1086 — Why Chips Are REALLY Helping Software Eat the World.
AI Ramblings Daily is the afternoon video + podcast — my ad hoc takes and perspective on the day’s AI issues & news flow, around 23 minutes today (single-event deep-dive variant), with short 1-minute clips for quick topic views. Today: episode #76.
Subscribe to either or both on michaelparekh.substack.com. They run as separate Sections you can opt into or out of.
Links used in today’s show (already embedded inline above; listed here for reference)
Take 1 — Cerebras IPO Pricing + Performance Echoes Internet Days (8 sub-bullets):
-
The Information — Cerebras IPO Winners Include Foundation, Benchmark, OpenAI
-
AI-RTZ #1084 — AI Chip Boom + AI IPO This Week with Cerebras
-
AI-RTZ #1066 — SpaceX/xAI IPO “We Are the Geese” Trillion+ Dollar Passive Index Games
-
AI-RTZ #1086 — Why Chips Are REALLY Helping Software Eat the World
-
AI-RTZ #1003 — Mega AI IPOs in 2026: SpaceX/xAI, OpenAI, Anthropic
-
“We Are the Geese” — MP on X (foie gras gavage applied to mega AI IPO passive-index dynamics)
Gadget AI — Trouble in Paradise: OpenAI vs Apple:
-
Bloomberg — OpenAI/Apple Partnership Frays, Setting Up Possible Legal Fight
-
AI-RTZ #728 — OpenAI’s ‘Sam/Jony’ AI Device $6.5B Acquisition
-
AI-RTZ #379 — Examining the Potential Apple/OpenAI Tie-Up (the 2024 deal)
Q1 + Q2 — AI Wearables (AirPods with Cameras / AI Smart Glasses):
Companion text:
