May 2026 · CBRS · Gojo · AI

Cerebras (CBRS) — IPO Follow-Up

Priced at $185. Opened at $350. Closed day 1 at $311 — a $95B market cap on $510M of trailing revenue. The AI hardware narrative won. Now Cerebras has to earn it.

IPO Result

IPO price: $185/share — priced May 13, 2026
Shares sold: 30 million shares, raising $5.55B
Opening print: $350 (+89% above offer)
Intraday high: ~$386 (+109%)
Day 1 close: $311.07 (+68%)
Day 1 market cap: ~$95B (briefly touched $100B fully diluted at the open)

This was the largest US tech IPO of 2026. The book was massively oversubscribed, and underwriters priced at $185 — already above the original S-1 range — but the grey market had been pricing CBRS well above that for weeks. The 68% first-day pop reflects just how much demand was left on the table.

First Week

Day 2 (May 15) gave back roughly 10%, with CBRS trading near $294. That’s not surprising — IPO pops of this magnitude almost always see a wave of flippers locking in gains the next session. The stock remained approximately 59% above the IPO price entering the weekend, which for a company at 186x trailing revenue is still a very stretched starting point.

No major analyst initiations had crossed as of May 15; the standard 25-day quiet period applies to underwriters. That will be the next meaningful catalyst for price discovery — watch for Goldman, Citi, and Morgan Stanley coverage in mid-June.

Pre-IPO Scorecard

We published a pre-IPO research note on April 27. Here’s how the predictions held up:

❌ Valuation range ($23B–$35B): Wrong, and badly so. The market priced CBRS at ~$48B at the IPO and $95B by day’s end. Our ceiling was $35B — the actual open blew past $100B on a fully diluted basis. The AI hardware narrative overwhelmed any fundamental anchor. We were too conservative on price, but we correctly identified the valuation as the key swing factor.
✓ UAE concentration as the key risk: Still real. G42 + MBZUAI accounted for 86% of 2025 revenue (G42: 24%, MBZUAI: 62%). The IPO prospectus confirmed these figures. The market is pricing this risk away — that’s either the market being smart about forward revenue mix, or setting up a nasty surprise when quarterly numbers drop.
✓ WSE-3 as a genuine differentiator: The 68% pop validated the hardware moat thesis. No one is building a 4-trillion-transistor wafer-scale chip at this scale except Cerebras. The market agreed.
✓ “Wait for lock-up expiry at the top end of range”: The spirit of this call was right — don’t chase euphoria — even if our “top end” was $35B and the stock went to $95B. Lock-up expires approximately November 10, 2026. That advice still stands, just at a very different price level.
↗ OpenAI deal bigger than reported: The S-1 described a $10B+ commitment; the updated prospectus filed ahead of the IPO disclosed the deal value exceeds $20B, with 750MW of compute capacity locked through 2028. A meaningful upgrade to the bull case.

New Developments Since April 27

AWS Deal

The biggest news that landed between our pre-IPO note and the listing: Amazon Web Services will deploy Cerebras CS-3 systems in AWS data centers and expose the solution through Amazon Bedrock. This is material. It gives Cerebras distribution reach it could never build organically, and it establishes a hyperscaler relationship that is not UAE-dependent. The AWS deal does not change 2025 revenue (it’s forward-looking), but it meaningfully de-risks the “what if G42 walks?” scenario.

OpenAI Deal Upgrade

The OpenAI commitment grew from $10B+ to $20B+ in the updated filing — 750MW of Cerebras-backed low-latency inference capacity through 2028. If that revenue flows at the implied rates, Cerebras’s 2026 and 2027 revenue profile could look very different from the 2025 baseline. The catch: OpenAI is itself a concentration risk. Replacing 86% UAE dependency with heavy OpenAI dependency is trading one single-counterparty exposure for another.

Updated Thesis

At $311 and a $95B market cap, Cerebras is trading at approximately:

186x trailing revenue ($510M, 2025)
~95–100x estimated 2026 revenue — assuming the 76% growth rate holds and they approach $900M–$1B
For context: Nvidia at its peak AI euphoria traded at roughly 35–40x trailing revenue. CBRS is pricing in Nvidia-level dominance before it has Nvidia-level revenue diversity, margins, or customer breadth.

The bull case is real: WSE-3 is a technically superior inference chip, the OpenAI deal provides massive near-term revenue, AWS Bedrock gives enterprise distribution, and the 76% growth rate shows no sign of slowing. If Cerebras can hit $2B+ in revenue by 2027 and expand beyond UAE and OpenAI, the $95B market cap becomes defensible in retrospect.

The bear case is equally real: 86% of 2025 revenue came from two UAE government-adjacent customers. The OpenAI deal is a single counterparty with its own volatile competitive positioning. There is no GAAP profitability. Nvidia will not sit still on inference. And the stock is priced for perfection — any crack in execution will be punished.

Current call: Don’t chase. The IPO pop was real and deserved — the company is more interesting than our April note priced. But $95B for $510M in trailing revenue, with 86% customer concentration still unresolved, is not the setup we want. The lock-up expiry in November 2026 will be the first real test: if the stock holds above $250 through lock-up and quarterly revenue confirms the OpenAI ramp, the thesis strengthens materially. That is the entry point to revisit seriously.

Watch For

Q1 2026 revenue (first public earnings): Does the 76% growth rate hold? More critically — what percentage came from UAE vs. OpenAI vs. AWS? That revenue mix shift is the single most important data point going forward.
Analyst initiations (~mid-June): Quiet period lifts around June 8. Price target distribution from IPO underwriters will set the near-term trading range and often drives a secondary momentum leg (or reset).
AWS Bedrock deployment timeline: When do CS-3 systems actually go live in AWS data centers? A specific public date would be a meaningful positive catalyst.
Nvidia’s response: Jensen Huang does not ignore competitive threats. Watch for any NIM or Blackwell-class inference optimization aimed at Cerebras’s latency advantage.
Lock-up expiry (~November 10, 2026): 180 days of insider and early investor supply overhang. Sequoia, Foundation Capital, and other early backers hold large positions. Supply pressure typically hits 2–4 weeks before the official date as informed money front-runs the expiry.
UAE geopolitical risk: G42 and MBZUAI are government-adjacent entities in the UAE. Any shift in US–UAE relations, AI export controls, or G42’s own strategic direction is a direct revenue risk that the current valuation does not price seriously.