Strong Results and Raised Guidance Reinforce Bullish View; Capacity and Disaggregated Inference Open Incremental Opportunity

Overview

Cerebras exceeded expectations in Q2 on core revenue, gross margin, and operating margin, and raised its FY26 outlook. Morgan Stanley believes fast-inference demand continues to exceed available compute supply, while the company's wafer-scale processor architecture, expanding cloud capacity, and disaggregated inference partnerships with AMD and AWS support its long-term growth thesis; it maintains its Overweight rating.

Core views

Capacity is the most important near-term operating driver. Data-center availability remains a bottleneck, but more than 600MW of operational or contracted capacity through the end of 2027 improves visibility into the revenue ramp.Disaggregated inference can have GPUs or other accelerators handle prefilling, while Cerebras systems handle decoding; if commercialized successfully, both throughput and unit economics could improve materially.AWS is expected to begin contributing more broadly to hyperscale-customer revenue in 2027, while Morgan Stanley's current model remains conservative on this contribution, creating upside potential.The technology roadmap continues to advance, with CS4 expected to launch shortly and CS5 targeted for the second half of 2027; the ability to serve leading frontier models helps alleviate market concerns about suitability for large-scale workloads.Valuation uses 13x 2028 adjusted revenue, benchmarked against AI semiconductor peers, and the target-price increase mainly reflects higher 2028 revenue expectations.

Analysis framework

The assessment is based on variances between actual quarterly performance and analyst expectations, combined with management guidance, capacity deployment, customer/partner progress, and long-term revenue and margin modeling; valuation uses a revenue-multiple approach based on comparable AI semiconductor companies.

Methodology notes

• Valuation methodsRevenue Multiple Method

  13x valuation of 2028 adjusted revenue

  Applies a 13x multiple to 2028 adjusted revenue of $6.3bn; this multiple is broadly consistent with small- and mid-cap AI semiconductor companies with high AI revenue exposure and growth rates above the market.

• Scenario AnalysisBull/Base/Bear Scenarios

  Capacity deployment and hardware demand determine revenue growth and target-price ranges

  The bull case assumes deployment of more than 1GW of capacity by 2028; the bear case assumes capacity delays, lower utilization, and weak hardware demand.

• Earnings AnalysisActual Results Versus Expectations

  Revenue, gross margin, operating margin, and EPS beats

  Assesses earnings quality and subsequent forecast revisions by comparing actual quarterly metrics and company guidance with Morgan Stanley's prior estimates.

Asset mapping & comparison

Structured mapping from thesis to named assets (strengths, weaknesses, peers, risks).

• CBRS.US

  Directly Covered Security

  Strengths

  Strong demand for fast inference; differentiated wafer-scale processor architecture; improved capacity availability; partnerships with AMD and AWS provide commercialization expansion opportunities.

  Weaknesses

  Current profitability remains in a ramp-up phase; business expansion depends on large-scale data-center construction and customer deployment timing.

  Comparison

  Valuation uses a 13x multiple of 2028 adjusted revenue, near the average level for AI semiconductor peers.

  Risks

  Capacity deployment delays, customer concentration, intensifying competition, weaker-than-expected hardware demand, and short-term volatility from lock-up expirations.

Key data

• Q2 Core Revenue$209.9mnUp 9.7% quarter over quarter and 103% year over year, above Morgan Stanley's $194.3mn estimate.
• Q2 Core Gross Margin40.6%Approximately 380 basis points above Morgan Stanley's 36.8% estimate.
• Q2 Non-GAAP EPS-$0.05Better than Morgan Stanley's -$0.21 estimate.
• Core Cloud and Other Services Revenue$127.7mnUp 287% year over year.
• Remaining Performance Obligations$25.4bnBalance at quarter-end.
• September-Quarter Core Revenue Guidance$214mn-$216mnMidpoint of $215mn, up 2.4% quarter over quarter and 58% year over year, above Morgan Stanley's $205.9mn estimate.
• FY26 Core Revenue Forecast$886mnPreviously $863mn.
• FY27 Revenue Forecast$2.88bnPreviously $2.71bn.
• FY28 Revenue Forecast$6.70bnPreviously $6.46bn.
• FY26 Core Gross Margin Forecast41.8%Previously 39.4%.
• Secured Capacity>600MWOperational or contracted capacity expected to be delivered through the end of 2027; Morgan Stanley's model is approximately 560MW.
• Target Price$279.00Raised from $273, based on 13x 2028 adjusted revenue of $6.3bn.

Impact & implications

The report's key implication is that Cerebras's investment thesis is shifting from pure high-growth expectations to three verifiable drivers: capacity delivery, improved inference efficiency, and hyperscale-customer commercialization. If capacity comes online faster than model assumptions, or AWS and disaggregated inference generate scaled revenue earlier, revenue and margin forecasts could see further upside; conversely, valuation is sensitive to execution misses and near-term trading volatility.

Risks

• Delays in data-center and power-capacity deployment could cause a slower-than-expected revenue ramp and lower utilization.
• Customer concentration is high, and changes in purchasing cadence by key customers could materially affect performance.
• Intensifying competition in AI inference and accelerators could pressure pricing, market share, or gross margins.
• The timing and scale of commercialization for AMD/AWS disaggregated inference and hyperscale-customer partnerships remain uncertain.
• The company must execute large-scale capacity buildouts amid rapid expansion, with high capital requirements and operating complexity.
• Staggered lock-up expirations in August could increase short-term share-price volatility.

What to watch

• Actual delivery, commissioning, and utilization progress for more than 600MW of capacity, and whether it exceeds the approximately 560MW model assumption.
• Delivery against September-quarter revenue guidance and 38%-40% core gross-margin guidance.
• Customer adoption, orders, and revenue contribution after the AWS channel becomes broadly available in Q1 2027.
• Commercialization progress for AMD and AWS Trainium disaggregated inference solutions, including evidence of throughput gains and gross-margin improvement.
• Order, RPO, and customer-demand validation of the target for more than 3x FY27 core revenue growth.
• The impact of the CS4 launch and CS5 R&D progress on technology competitiveness and customer deployments.