← Sazabi Research

Cerebras Systems (CBRS): Wafer-Scale Inference, a $25B Backlog, and the Hottest Tape in AI

#1 US VolumeSurge pick Fri 4 Sep (+10.3% to $210.05) — Europe capacity land-grab + Q2 cloud surge — Quick Take

📅 September 5, 2026 TechnologyAI HardwareSemiconductorsInferenceNASDAQ: CBRSUS Large Cap
Market Cap
~$50B
Revenue (TTM)
~$681M
P/E (Fwd)
n/m
Rev Growth
+112% TTM

Why This Company Now

Cerebras was the #1 US VolumeSurge pick in the freshest scan (Sat 5 Sep, covering Friday's session): +10.3% to $210.05 on $2.9B turnover — the top Technology name on a tape where Technology was also the leading HK sector (9/50). The squeeze higher is follow-through on two catalysts: the Sept 1 announcement of a 165 MW AI data-centre campus in Mikkeli, Finland (with Compute Nordic Finland, initial 50 MW phase already under construction) and the blowout Q2 2026 report (Aug 12) that showed cloud revenue nearly quadrupling and a record $25.4B contracted backlog. Shares had been chopped up post-earnings (-11.7% on Aug 13), so Friday's move marks renewed momentum as CS-4 hits general availability this quarter.

Financial Snapshot

Fiscal Year (Jan-Dec)RevenueRev GrowthOp MarginNet Income
FY2022$24.6M—~-726%-$178M
FY2023$78.7M+220%~-170%-$127M
FY2024$290.3M+269%~-35%-$485M
FY2025$510.0M+76%~-29%+$88M*
TTM (Jun-26)$680.7M+112%~-81%-$487M

*FY2025 GAAP profit was flattered by large non-cash gains; full-year operating loss was still ~$146M. Q2 2026 (Jun-26): GAAP revenue $180.1M (+74% YoY); core revenue $209.9M (+103%); cloud & services revenue $126.0M (+281%); core gross margin 41% (+~940bps YoY); GAAP net loss $450.5M on R&D of $320.2M. FY2026 core revenue outlook raised to $880–890M. Sources: company filings, StockAnalysis, 24/7 Wall St, StockTitan.

Bull vs Bear

✅ Bull Case

  • $25.4B of RPO (as of Jun 30, 2026) — anchored by the ~$20B OpenAI deal for 750 MW of inference compute supporting GPT-5.6 "Sol" — is ~37x trailing revenue and gives multi-year visibility no other AI hardware name can match.
  • Cloud inflection is real: Q2 cloud/services revenue +281% YoY to $126M, core total revenue +103%, core gross margin up ~940bps to 41%; AMD disaggregated-inference goes to production in Q4 2026 and AWS Bedrock availability follows in Q1 2027 — distribution beyond Nvidia's stack.
  • Capacity land-grab funded: 600+ MW contracted; new 165 MW Finland campus (50 MW phase building now) plus US expansions; ~$8.6B liquidity from the $6.4B IPO (largest semiconductor IPO ever) and an $850M revolver; CS-4 GA this quarter with ~10x throughput-per-watt vs CS-3.

🔴 Bear Case

  • GAAP losses are widening, not narrowing: Q2 2026 net loss was $450.5M (vs year-ago profit) as R&D jumped to $320M in a single quarter; H1 capex hit $549M and cash flow is negative — a self-built data-centre model burns capital at a ferocious rate.
  • Valuation prices in flawless 2027 execution: ~$50B cap is ~57x the raised 2026 core-revenue guide ($880–890M) with no GAAP profitability in sight — any OpenAI ramp slip, CS-4 yield hiccup, or order delay invites violent de-rating, as the -11.7% post-earnings session showed.
  • Customer concentration + the 800-lb gorilla: OpenAI dominates backlog economics while Nvidia (Vera Rubin ramping, ~$5T+ cap) and AMD attack inference directly; Friday's daily summary also flagged early "AI shakeout" warnings from VCs — sentiment risk for the whole AI-infra cohort.

Verdict

Cerebras is the purest public "Nvidia-alternative" inference bet: wafer-scale silicon with a genuine speed-per-watt edge, a $25B+ contracted backlog, and cloud revenue now compounding at triple digits. The question is conversion — whether RPO turns into GAAP revenue and margin faster than the capex/R&D burn and dilution consume the balance sheet. At ~$50B (~57x 2026 core revenue), the stock is a high-beta expression on AI inference share-shift; interesting as a satellite sized for drawdowns, not a core holding, until the loss profile bends.

Disclaimer: AI-generated quick take for informational purposes only. Not investment advice. Data from public filings and financial data providers.