Q
QuantAbundance
4 min read

Cerebras Q3 2026: core revenue $230.9M above guidance, cloud up 272%, Q4 guided at $310M to $330M

Cerebras posted Q3 core revenue of $230.9M against a $214M to $216M guide, with cloud revenue up 272% in a year and $26.1B of remaining performance obligations. GAAP revenue was $210.8M and the net loss $161.2M; Q4 core revenue is guided at $310M to $330M with core operating margin near breakeven.

Get the full $CBRS dossier ($19) → Code YOUTUBE takes 30% off until 1 December 2026.

CBRSCerebrasEarningsQ3 2026AI InferenceAI HardwareWafer-Scale

$CBRS reported third-quarter core revenue of $230.9M, up 70% from a year ago and above the $214M to $216M it had guided, with core cloud revenue of $144.9M, up 272%. GAAP revenue was $210.8M and the GAAP net loss $161.2M. Remaining performance obligations reached $26.1B, and the company guided fourth-quarter core revenue to $310M to $330M with a core operating margin around breakeven. This recap walks the print, the guidance and the balance sheet against our /stocks/cbrs thesis.

The TL;DR. The mix is moving from hardware sales to cloud: GAAP cloud revenue more than tripled while hardware revenue fell about 29% (our arithmetic). Core margins improved for the quarter and the guide points higher, but on a GAAP basis the company still lost $224.1M at the operating line.

This recap is drawn from the press release Cerebras filed with the SEC on November 12, 2026 after the close (8-K item 2.02, exhibit 99.1). The conference call and the 10-Q filed the same evening are not covered. "Core" figures are the company's non-GAAP measures: they exclude customer warrant amortization, stock-based compensation and data center pass-through revenue and costs.

The print

Metric, US$Q3 2026Q3 2025Change
Core revenue, non-GAAP$230.9Mn/a+70%
Core cloud and other services revenue$144.9Mn/a+272%
Core gross margin41%n/aup
Core operating margin-14%n/a
GAAP revenue$210.8M$135.7M+55%
GAAP cloud and other services revenue$141.9M$38.9M+265%
GAAP hardware revenue$68.9M$96.8M-29%
GAAP gross margin20%41%
GAAP operating loss-$224.1M-$26.5M
GAAP net loss-$161.2M-$22.2M
Net loss per share-$0.67-$0.41

Changes and the GAAP gross margins are our arithmetic from the release's tables. Against the second quarter, core revenue rose from $209.9M.

What moved the numbers:

  • Cloud over hardware. Cloud and other services were 67% of GAAP revenue, against 29% a year ago (our arithmetic). The company sells through direct system sales, Cerebras Cloud and third-party clouds.
  • The gap between core and GAAP. GAAP gross margin of 20% against 41% core: the difference is mostly non-cash customer warrant amortization, stock-based compensation and pass-through data center items.
  • Operating expenses. Research and development of $177.8M and total operating expenses of $266.2M, against $81.6M a year ago.
  • Below the line. $63.9M of other income narrowed the loss before taxes to $160.2M.

The business

  • Backlog. $26.1B of remaining performance obligations at September 30, from $25.4B at the end of June.
  • Customers. A cloud deal with a top three hyperscaler, details next year; Oracle Cloud Infrastructure to offer Cerebras inference in 2027; more neoclouds and the Bilt platform.
  • Capacity. 700 gross megawatts of data center capacity under contract for delivery by the end of 2027, new capacity in Norway and Finland, and manufacturing with Flex and Sanmina expected to exit 2026 at more than 10 times its start-of-year level.
  • Products. CS-4 soon in full production; CS-5, built on the Wafer Scale Engine 5, targeted for early summer 2027.

The guidance

Company guidance, core (non-GAAP)Q4 2026Full year 2026
Revenue$310M to $330M$940M to $960M
Gross margin47% to 49%42% to 44%
Operating margin-1% to +1%-10% to -8%

The CEO said the company expects to triple revenue in 2027. These are company targets, not results.

The cash

  • Liquidity. $7.8B of cash, cash equivalents, restricted cash and short-term investments, plus $850M of debt capacity.
  • Balance sheet. Total assets of $11.94B against $2.33B at December 31, with property and equipment up to $1.51B from $0.44B as the cloud build continues.
  • Customer warrants. $1.09B of customer warrants on the balance sheet (current and non-current), the item behind much of the gap between core and GAAP margins.

The read across AI inference

The standing QA thesis on /stocks/cbrs is a specialist wafer-scale inference accelerator with a large contracted book and high customer concentration. This print adds names (a hyperscaler, Oracle) to a customer list the thesis flagged as narrow, and shows the shift from hardware to cloud. The concentration question is not answered by the release itself: it does not break revenue down by customer.

Its closest correlated names in the QA universe are $LAZR (0.66), $RKLB (0.60) and $AAOI (0.54), over a short history since the May IPO. More names sit in the AI inference theme. Earlier the same day, $NBIS printed, see the Nebius Q3 2026 recap.

For the full teardown in writing, eight quarters of filings and the level map, see the Cerebras dossier.

What worked

  • Core revenue of $230.9M, above the $214M to $216M guide, the third beat in a row according to the company.
  • Cloud revenue up more than threefold in a year.
  • Core gross margin of 41%, guided to 47% to 49% in Q4.
  • New customers: a top three hyperscaler and Oracle Cloud Infrastructure.

What broke

  • GAAP losses. An operating loss of $224.1M and a net loss of $161.2M.
  • Hardware. Hardware revenue down about 29% in a year.
  • Spending. Research and development roughly tripled to $177.8M.
  • The tape. A $97.75 close on November 12, down 31% over a month and 61% over three months in the data QA tracks. The CBRS perpetual on Hyperliquid, which trades around the clock, marked $113.78 at 22:12 UTC after the release, against $100.58 the day before.

This recap is education, not investment advice.

The full dossier

Included in Pro

The CBRS teardown in writing, its SEC filings in one table, peers with numbers and the live level map, in one read.

See the dossier

Related bubbles

Related research

Go deeper

Want more of these in your Google results? Add QuantAbundance as a preferred source.

Add to Preferred Sources