Cerebras shares closed Monday, October 5, at $181.55, up 9.08% from Friday’s $166.43 close, according to TECHi’s Yahoo-sourced CBRS quote. The 4 p.m. Eastern observation is a regular-session close; Yahoo labels the feed delayed without specifying the number of minutes. Investors welcomed OpenAI chief Sam Altman’s Friday description of Cerebras as a close partner. The harder question is how quickly that relationship becomes reported sales.
The distinction matters after a punishing week. Monday’s close was still 12.2% below the September 25 close of $206.75 and 1.9% below Cerebras’s $185 IPO price. A one-day rebound repairs some of the loss, but it does not establish a new contract, a faster installation schedule or a change to the company’s financial guidance.
Article Brief
What the rebound does—and does not—show
3 Points18s Read
- The closeCBRS ended October 5 at $181.55, up 9.08%, but remained below its $185 IPO price.
- The contractOpenAI’s 750MW commitment remains; the additional 1.25GW is only an option.
- The near-term testCerebras’s Q3 core-revenue midpoint implies about 2.4% sequential growth.
Reassurance is not a new order
Altman addressed speculation about the partnership in an October 2 post, calling Cerebras a close partner. His statement is meaningful because OpenAI is central to Cerebras’s future workload. It did not disclose a revised purchase commitment, deployment date or division of work between Cerebras and other chip suppliers. Different systems can serve different models or workloads at the same customer; one supplier’s use does not, by itself, cancel another’s contract.
The binding relationship is already visible in Cerebras’s June-quarter filing. OpenAI committed to buy 750 megawatts of inference capacity and related services over a multiyear term. It funded an approximately $1 billion secured working-capital loan to help build the required infrastructure and received a warrant tied to contractual milestones. OpenAI also has an option, not a commitment, for another 1.25 gigawatts. Calling the full potential two gigawatts a guaranteed order would overstate the agreement.
That structure creates two tests. The partnership can be commercially intact while a particular product tier uses other hardware. Separately, Cerebras must deliver capacity on schedule and recognize revenue as service is supplied. The filing says the company recognized $56.8 million of revenue under the OpenAI arrangement in the second quarter. That was about 31.5% of the quarter’s $180.1 million GAAP revenue, a TECHi calculation using the two disclosed figures. It is a measure of revenue recognized in that period, not OpenAI’s share of all future business.
A $25.4 billion backlog is not this quarter’s revenue
Cerebras reported $25.4 billion of remaining performance obligations at June 30, with a significant amount attributable to OpenAI. That large balance helps explain why a reassurance from the customer can move the stock. But the filing describes revenue expected to be recognized over years, subject to the timing of deployments and customer requests. It also says some data-center pass-through costs are included in the transaction price. RPO should not be treated as cash already collected, a quarterly sales forecast or a measure of future profit.
The near-term hurdle is less dramatic. Cerebras’s August 12 results release guided third-quarter core, non-GAAP revenue to $214 million–$216 million, versus $209.9 million in the second quarter. At the $215 million midpoint, that is about 2.4% sequential growth. The calculation does not use GAAP revenue: Cerebras’s second-quarter GAAP total was $180.1 million, and its core measure excludes specified warrant, stock-compensation and pass-through effects. Mixing those two bases would make the growth look much larger than management’s actual core guidance implies.
There is a credible bullish response. Cloud and other services GAAP revenue reached $126.0 million in the second quarter, up 281% year over year; Cerebras said more than 600 megawatts of data-center capacity was live or under contract for delivery by the end of 2027. A modest Q3 core revenue step may reflect an installation schedule rather than weaker demand. The 600-megawatt figure is a company-wide capacity measure with a different scope and timetable from OpenAI’s 750-megawatt commitment, so subtracting one from the other would not reveal an exact shortfall.
The counterargument is equally concrete. Scaling capacity requires sites, power, manufacturing and financing before the contracted revenue appears. Cerebras reported a 14% GAAP gross margin in Q2, versus 41% on its adjusted core measure. Investors should watch whether deployments improve the GAAP economics, not just whether a partner continues to speak favorably about the technology. TECHi’s earlier first-earnings analysis raised the related customer-concentration question; the newly reported OpenAI revenue gives that question a measurable baseline.
What the next update needs to show
The first check is third-quarter core revenue against the $214 million–$216 million guide, followed by GAAP revenue and the bridge between the two. Then look for newly delivered capacity, revenue recognized under the OpenAI agreement and any change to the $25.4 billion RPO balance. A rise in backlog without a corresponding delivery and revenue ramp would tell a different story from a rise in both.
The share register matters too. TECHi’s Cerebras lockup analysis explains why stock newly eligible for sale can add supply even when the operating outlook improves. Eligibility is not proof of actual selling, and Monday’s close cannot isolate its effect. The company and market-data providers currently place the next earnings event in November; the company should confirm the exact date before readers treat an estimated calendar entry as final.
Monday’s 9.08% gain is real. Its meaning is conditional. Altman’s comment reduces one worry about the relationship, while Cerebras’s own filing shows the financial work that remains: turn committed megawatts and a multiyear backlog into delivered service, recognized revenue and better GAAP margins. The next report, not the size of one rebound, will show whether that conversion is accelerating.
