Cerebras closed Friday at roughly $280, down about ten percent on its second-ever session of public trading after opening at $350 and printing all-time highs near $400 on Thursday. [1] The $185 IPO price set the bar; Thursday's +68 percent close pushed it above $310; Friday's reversal snapped the surge. [2] The semiconductor index lost four percent on the same session: Nvidia -4.4 percent, AMD -5.7 percent, Intel -6.2 percent. [3] The slide is now a session, not a guess.

The May 15 paper held that Day 2 would test whether AI-IPO demand could absorb the UAE concentration risk printed in the prospectus. Saturday answers: the test arrived simultaneously with a sector-wide concentration unwind that pulled the entire SOX down four percent. Disentangling the Cerebras-specific risk from the semis-wide derate is the analytical question of the next session. The market did not give the IPO-specific story the room it needed to publish cleanly.

CNBC and Yahoo Finance framed the slide as healthy froth correction. [1][2] That framing is consistent with the precedent — first-day pops of 50 to 70 percent on AI-adjacent IPOs typically give back half their gain within the first five sessions as flippers rotate out and longer-duration holders calibrate. The framing is also incomplete. The Day 2 reversal coincided with both a sector-wide concentration unwind and the UAE concentration risk that the prospectus disclosed in plain language. A single-name slide that lines up with a sector-wide slide is the chart pattern the bear case wants. A single-name slide that decouples from the sector slide would have been the bull case.

The UAE concentration is the part the prospectus prints and the post-IPO coverage tends to skip. Cerebras's largest customer — by a long margin — is the Group 42 / UAE-anchored compute footprint. The dependency on a single sovereign-adjacent customer is the risk a public-market shareholder takes on at $280 that they did not take on at $185. The Thursday surge to $311 was the market pricing the demand side without pricing the customer concentration. The Friday slide was the market pricing the concentration side without yet incorporating any new demand-side news.

The OpenAI counterparty map the paper has been building this week is the relevant adjacency. Cerebras is one of the labs OpenAI buys compute from. The Cerebras share-price discovery is therefore a backdoor read on OpenAI's training and inference economics — a public-market window into a private-company supplier relationship. The Daybreak launch (covered in today's tech feature), the Huang/H200 non-shipment to Chinese firms (today's tech feature), the Musk-Altman litigation, and the Vercel OAuth incident silence all sit on the same map. The Cerebras slide on Day 2 is the market's first audit of one node on that map.

The 10-year Treasury hit its highest yield since May 2025 on the same Friday. The Dow lost 1.07 percent, the S&P 500 lost 1.24 percent, the Nasdaq lost 1.54 percent. [3] The CME FedWatch tool is now pricing a December rate hike at approximately 40 percent, up from 13.6 percent a week ago. Top economic forecasters now project Q2 CPI at six percent. The macro backdrop is tightening into the same window Warsh inherits the Fed.

The sequencing matters for what the Cerebras slide means. If the SOX -4 percent is the dominant signal, Cerebras's -10 percent is a beta amplification of a sector move that will mean-revert. If the UAE concentration is the dominant signal, the Day 2 slide will continue into Day 3 and Day 4 as the prospectus risk gets re-priced session by session. The way to tell the two stories apart is to watch whether Nvidia and AMD bounce in early Monday trading. If they do and Cerebras does not, the Cerebras-specific story is the price discovery.

What the Day 2 print does not yet contain: any new customer disclosure, any prospectus-supplemental filing, any insider lockup news. The next dataable point is the first quarterly print after the IPO. The UAE customer concentration ratio that appears in the 10-Q will be the document the price either responds to or ignores. Until then, Day 3 trading is the next instrument.

-- THEO KAPLAN, San Francisco