Cerebras Q3 2026 Earnings & 272% Cloud Surge: Factor Attribution of Wafer-Scale Moats, 41% Gross Margin, and AI Hardware Inversion
Cerebras Systems Q3 2026 delivered $230.9M in Core revenue, up 70% YoY, driven by a 272% surge in Core Cloud revenue to $144.9M and 41% gross margin. We evaluate wafer-scale engine architecture, warrant dilution, and factor exposures across the Six-Factor Framework.
Executive Summary: The Pure-Play Inference Inflection
On November 12, 2026, Cerebras Systems (NASDAQ: CBRS) reported its third-quarter 2026 financial results, posting record GAAP revenue of $210.8 million (up 55% year-over-year) and record Core revenue of $230.9 million (up 70% year-over-year), exceeding consensus midpoint expectations by $16.0 million. The standout driver of the quarter was the company's cloud segment: Core cloud and other services revenue accelerated 272% year-over-year to $144.9 million, now constituting 62.8% of total Core net sales. Furthermore, Core gross margin expanded to 41.0%, outperforming quarterly guidance by 200 basis points and marking a definitive inflection from the cyclical hardware troughs of mid-2026.
For institutional allocators and equity long/short managers, Cerebras represents a unique study in asset-specific factor dispersion within the artificial intelligence compute complex. Unlike fabless GPU designers that compete on general-purpose floating-point operations or traditional original design manufacturers that absorb low-margin contract assembly, Cerebras operates a vertically integrated wafer-scale architecture (CS-3 powered by the Wafer Scale Engine 3, or WSE-3). By packaging 4 trillion transistors, 900,000 AI-optimized cores, and 44 gigabytes of on-chip SRAM onto a single contiguous 215mm x 215mm silicon wafer, Cerebras bypasses memory bandwidth bottlenecks to deliver orders-of-magnitude faster token generation for real-time generative inference.
+-----------------------------------------------------------------------------------+
| CEREBRAS Q3 2026 CONSOLIDATED HIGHLIGHTS |
+------------------------------+--------------------+--------------------+----------+
| Metric | Q3 2025 Reported | Q3 2026 Reported | YoY Δ |
+------------------------------+--------------------+--------------------+----------+
| GAAP Total Revenue | \$135.7M | \$210.8M | +55.3% |
| Core (Non-GAAP) Revenue | \$135.7M | \$230.9M | +70.1% |
| Core Cloud & Services Sales | \$38.9M | \$144.9M | +272.4% |
| Core Hardware Sales | \$96.8M | \$86.0M | -11.2% |
| GAAP Gross Margin | 40.6% | 20.0% | -2,060bps|
| Core Gross Margin | 40.7% | 41.0% | +30 bps |
| Core Cloud Gross Margin | 16.5% | 46.9% | +3,040bps|
| Core Operating Margin | (19.5%) | (14.3%) | +520 bps |
| Core Adjusted EBITDA | (\$18.2M) | \$2.1M (YTD) | Positive |
| Contracted Power Capacity | 120 MW | 700 MW | +483.3% |
+------------------------------+--------------------+--------------------+----------+
Yet underneath this explosive top-line acceleration sits an intricate structural tension between customer concentration, customer warrant asset amortization ($36.4 million in Q3), capital leaseback structures, and high fixed-cost infrastructure depreciation. In this factor attribution, we deconstruct Cerebras Systems across the Parallax Six-Factor Framework—Value, Quality, Momentum, Defensive, Size, and Tactical—and map how portfolio managers should position wafer-scale hardware plays alongside broader neocloud allocations such as /blog/nebius-q3-2026-earnings-45b-capex-guidance-cut-factor-attribution and /blog/coreweave-q3-2026-earnings-132b-backlog-capex-debt-factor-attribution.
Deconstructing Q3 2026 Financial Results: GAAP vs. Core Reconciliation
To accurately evaluate Cerebras within systematic valuation and quality models, quantitative teams must disentangle GAAP figures from management's Core operating metrics. In Q3 2026, the discrepancy between GAAP and Core performance spanned revenue, gross margins, and operating income, driven by three major accounting mechanisms:
- Customer Warrant Amortization: Cerebras historically granted performance-based warrant assets to foundational anchor clients (notably G42) to secure multi-year hardware and cloud commitments. Under GAAP accounting, non-cash amortization of these customer warrants is recorded as a direct contra-revenue reduction. In Q3 2026, warrant amortization reduced GAAP revenue by $36.4 million ($17.1 million allocated against hardware, and $19.3 million against cloud and other services).
- Pass-Through Data Center Revenues and Costs: Cerebras engages in pass-through colocation contracts where power and shell infrastructure expenses are rebilled at cost to specific hosting tenants. GAAP revenue reflects $16.4 million of gross pass-through activity with zero net margin contribution, which management strips out to reflect baseline software, hardware, and inference service demand.
- Stock-Based Compensation & IPO Payroll Tax: GAAP operating expenses expanded to $266.2 million (compared to $81.6 million in Q3 2025), largely bloated by $16.5 million in cost-of-revenue stock compensation and front-loaded post-IPO equity vesting across R&D ($177.8 million).
GAAP to Core Revenue & Margin Walk (Three Months Ended September 30, 2026)
-------------------------------------------------------------------------
GAAP Revenue: \$210.8M
(-) Data Center Pass-Through Revenue: (\$16.4M)
(+) Customer Warrant Asset Amortization: +\$36.4M
Core Revenue: \$230.9M
GAAP Gross Profit: \$42.1M (20.0% margin)
(-) Pass-Through Data Center Net Margin: (\$0.5M)
(+) Customer Warrant Amortization Add-Back: +\$36.4M
(+) Stock-Based Compensation in Cost of Sales: +\$16.5M
Core Gross Profit: \$94.6M (41.0% margin)
The Core margin expansion from 38.3% in 9M 2025 to 42.5% in 9M 2026 highlights an accelerating mix shift: cloud inference contracts carry structural pricing power, while older legacy hardware purchase agreements signed at discounted initial pricing are completing their final deliveries. As hardware revenues transition to current-generation wafer-scale ASPs in Q4 2026, management guided for Core gross margin to leap another 700 basis points to between 47% and 49%, with full-year 2026 Core revenue reaching $940 million to $960 million.
Six-Factor Attribution Analysis
Using the Parallax Six-Factor Framework, we calibrate Cerebras Systems' relative factor deciles against the North American Information Technology and Communications sector universe as of November 13, 2026.
+---------------------------------------------------------------------------------------+
| CEREBRAS SYSTEMS (CBRS) FACTOR SCORECARD |
+--------------------+----------------+----------------+--------------------------------+
| Factor Pillar | Pre-Q3 Decile | Post-Q3 Decile | Decile Shift & Primary Catalyst|
+--------------------+----------------+----------------+--------------------------------+
| Value | Decile 1 | Decile 1 | Neutral (Premium P/S, EV/Rev) |
| Quality | Decile 3 | Decile 5 | +2 Deciles (Cloud Margin Ramp) |
| Momentum | Decile 8 | Decile 9 | +1 Decile (Relative Strength) |
| Defensive | Decile 2 | Decile 2 | Neutral (Customer Beta Risk) |
| Size | Decile 4 | Decile 4 | Neutral (Mid-Cap Growth) |
| Tactical | Decile 7 | Decile 8 | +1 Decile (Earnings Surprise) |
+--------------------+----------------+----------------+--------------------------------+
1. Value Factor (Decile 1 - Extreme Premium)
On standardized trailing and forward valuation multiples, Cerebras resides in the most expensive decile of the market. Trading at an annualized forward enterprise-value-to-revenue multiple exceeding 16.5x (and over 38x forward Core gross profit), CBRS screens as deeply antithetical to traditional Graham-and-Dodd Value screens.
The multiple reflects a scarcity premium: Cerebras is the only publicly traded pure-play competitor to Nvidia in commercial ultra-high-speed inference clusters. However, multi-factor models penalize CBRS severely on book-to-market, free-cash-flow yield (-4.8% on GAAP basis), and EV/EBITDA. While enterprise contract prepayments provide working capital support, Value-sensitive systematic portfolios must treat CBRS as a pure growth-duration factor exposure rather than an intrinsic value play.
2. Quality Factor (Decile 3 -> Decile 5 - Operating Inflection)
Quality metrics for Cerebras underwent an observable upward re-rating in Q3 2026, rising two full deciles from Decile 3 to Decile 5. Historically, Cerebras screened as a lower-tier quality entity due to significant operating losses (GAAP operating margin was -106% in Q3 2026) and heavy customer concentration.
However, decomposed Quality signals reveal two major structural tailwinds:
- Cloud Margin Convexity: Core Cloud gross margin expanded from 16.5% in Q3 2025 to 46.9% in Q3 2026—a 3,040 basis point surge. Unlike neocloud providers that lease third-party GPU clusters and face aggressive hardware amortization and cost-of-capital drag, Cerebras manufactures its own silicon, capturing vertical hardware rents directly through its API inference offerings.
- Balance Sheet De-Risking: Post-IPO liquidity and institutional credit facilities eliminated near-term insolvency risks. With zero outstanding high-yield long-term debt and positive year-to-date Core Adjusted EBITDA ($2.1 million), the company's Altman Z-Score and distance-to-default metrics transitioned out of the distress quadrant.
The primary factor overhang capping Quality at Decile 5 is customer warrant accounting and client concentration. Despite signing two new tier-one hyperscalers in Q3, revenue remains weighted toward foundational Middle Eastern sovereign AI agreements and contracted neocloud clusters.
3. Momentum Factor (Decile 8 -> Decile 9 - Relative Acceleration)
Cerebras exhibits elite 3-month and 6-month price and fundamental momentum. Relative price strength has decisively decoupled from broader semiconductor indices such as the SOX, echoing the cross-asset divergence detailed in /blog/semiconductor-dispersion-nvidia-governance-sox-factor-attribution.
Fundamental momentum metrics are equally pronounced:
- Full-year 2026 Core revenue guidance was raised to $940M–$960M (up from prior $880M consensus targets).
- Core Q4 revenue guidance of $310M–$330M implies sequential top-line growth of 38.6% and year-over-year expansion of over 95%.
- Contracted pipeline for high-speed AI inference expanded by more than 10x sequentially during the third quarter.
These operational revisions trigger strong positive revisions in quantitative earnings-momentum (SUE - Standardized Unexpected Earnings) screens.
4. Defensive Factor (Decile 2 - Elevated Volatility and Beta)
From a risk and capital preservation perspective, Cerebras scores in Decile 2. Its beta versus the S&P 500 exceeds 2.35, with annualized 30-day realized volatility surpassing 62%.
The asset's defensive vulnerabilities stem from:
- Supply Chain Single-Source Dependency: WSE-3 chips require specialized packaging, extreme wafer-level testing, and dedicated foundry allocation at TSMC (Taiwan Semiconductor Manufacturing Co.). Any geopolitical friction or advanced packaging bottlenecks (e.g., CoWoS and custom substrate interconnects) immediately jeopardizes volume deliveries.
- Power and Data Center Interconnection Timelines: In the Q3 earnings conference call, management acknowledged that sequential cloud revenue growth in Q4 would be flattish (following 13% sequential growth in Q3) due strictly to data center capacity constraints. Although Cerebras has contracted 700 MW of power, physical cluster energization will not scale meaningfully until December 2026 and early 2027.
- Macro Capex Sensitivity: If hyper-scaler AI infrastructure expenditure slows or shifts from frontier model training to lower-spec edge deployment, Cerebras' capital-intensive wafer systems face severe order push-outs.
5. Size Factor (Decile 4 - Mid-Cap Liquidity Profile)
With a market capitalization situated in the $8B to $12B bracket, Cerebras operates as a liquid mid-cap. The Size factor contributes positively to idiosyncratic alpha generation: it sits outside the mega-cap concentration of the Magnificent Seven, yet possesses sufficient average daily volume (ADV > $120M) to accommodate institutional block execution without excessive market-impact slippage.
6. Tactical Factor (Decile 7 -> Decile 8 - Guidance & Inference Demand Surge)
The Tactical factor captures short-term sentiment, order-flow catalysts, and technical positioning over a 5-to-30 day horizon. Following the Q3 release, CBRS rallied strongly in after-hours trading, driven by:
- Short Squeeze Dynamics: High short interest accumulated ahead of lockup expirations and IPO valuation debates faced an immediate catalyst in the 200 bps Core gross margin beat and full-year guidance raise.
- Inference Shift Catalyst: Industry benchmarks indicating that Cerebras delivers up to 1,800 tokens per second on Llama-3.1-70B models establish a compelling competitive wedge against Nvidia H200 and B200 clusters, which typically achieve 150 to 250 tokens per second in standard serving configurations.
- Q4 Gross Margin Step-Function: Management's specific guidance for a 700 bps sequential gross margin leap to 48% provides a tangible short-term fundamental anchor for quantitative breakout strategies.
Architectural Moat: Wafer Scale Engine vs. Commodity GPU Clusters
To evaluate the durability of Cerebras' Quality factor score, fundamental analysts must evaluate whether wafer-scale integration constitutes a sustainable technological moat or a temporary manufacturing anomaly.
+----------------------------------------------------------------------------------------+
| ARCHITECTURAL COMPARISON: WAFER SCALE VS. DISAGGREGATED GPUS |
+--------------------------+------------------------------+------------------------------+
| Dimension | Cerebras CS-3 (WSE-3) | Modern Multi-GPU Node (8x) |
+--------------------------+------------------------------+------------------------------+
| Silicon Form Factor | Contiguous 215mm x 215mm | 8 Disaggregated Reticle Dies |
| Transistor Count | 4.0 Trillion | ~1.6 Trillion Combined |
| On-Chip SRAM Cache | 44 Gigabytes | ~2 Gigabytes Combined |
| Memory Bandwidth | 21 Petabytes / second | ~64 Terabytes / second (HBM) |
| Interconnect Fabric | Silicon-Native 2D Mesh | External NVLink / PCIe / CXL |
| Primary Optimization | Ultra-Low Latency Inference | General-Purpose Training |
| Memory Wall Vulnerability| Bypassed (Model in SRAM) | Constrained by HBM Latency |
+--------------------------+------------------------------+------------------------------+
In traditional distributed GPU architectures, token generation during autoregressive inference is severely memory-bandwidth bound. Each generated token requires reading billions of model weights from High-Bandwidth Memory (HBM) into compute registers. As illustrated in our analysis of high-speed interconnect infrastructure in /blog/astera-labs-q3-2026-earnings-scorpio-switch-factor-attribution, multi-chip clusters spend a substantial portion of their energy budget simply shuttling bytes across PCIe, Retimer, and optical interfaces.
By fitting entire medium-scale language models or distributed mixture-of-experts (MoE) active layers directly into 44 GB of on-chip SRAM operating at 21 PB/sec bandwidth, Cerebras eliminates external memory round-trips entirely. This architectural advantage explains why Core Cloud gross margins expanded by over 3,000 basis points in a single year: when token throughput per megawatt increases by 4x to 8x, the unit economics of AI-as-a-Service shift decisively in favor of the wafer operator.
Portfolio Construction: Sizing CBRS in Institutional Mandates
How should a systematic risk manager integrate Cerebras Systems into an institutional multi-factor equity mandate?
In accordance with our core portfolio construction guidelines outlined in /blog/core-satellite-factor-allocation-framework, high-beta, high-valuation assets with accelerating fundamental momentum belong strictly in the tactical satellite sleeve rather than the core factor benchmark.
Institutional Portfolio Allocation Schematic
-------------------------------------------------------------------------
[Core Multi-Factor Equity Portfolio (70% - 85%)]
├── Quality-Tilted Mega-Cap Platforms (Nvidia, Microsoft, TSMC)
├── Low-Volatility Defensive Satellites
└── Value-Screened Cash-Flow Generators
[Tactical Thematic Compute Sleeve (15% - 30%)]
├── Wafer-Scale AI Inference Pure-Play: Cerebras Systems (1.5% - 3.0%)
├── Interconnect & PCIe Architecture: Astera Labs (2.0% - 3.5%)
├── Specialized Neocloud Infrastructure: CoreWeave / Nebius (3.0% - 5.0%)
└── Liquid Power & Data Center Cooling: Vertiv / AAON (4.0% - 6.0%)
Risk Controls and Hedging Framework
- Factor Crowding Constraint: CBRS exposure should be strictly capped at 2.5% to 3.0% of portfolio active risk. Position sizing must be dynamically adjusted for its 60%+ realized volatility using inverse-volatility weighting or trailing average-true-range (ATR) risk stops.
- Momentum Reversal Protection: Given that Value is ranked in Decile 1, CBRS is acutely vulnerable to factor rotation shocks. Quantitative risk engines should implement systematic stop-losses if 50-day moving average volume collapses or if Core gross margin progression stalls below 45% in Q4.
- Cross-Asset Pairs Trading: Long CBRS vs. Short Low-Moat Contract Server Assemblers provides an attractive factor-neutral tactical structure, isolating proprietary wafer-scale IP while hedging broad hardware capex cyclicality.
Conclusion
Cerebras Systems' Q3 2026 financial report confirms that the market for AI compute is transitioning from indiscriminate training cluster buildouts to latency-critical, high-throughput inference serving. With Core cloud sales surging 272% to $144.9 million and Core gross margins accelerating toward 48% guidance, Cerebras has demonstrated authentic operating leverage. While customer concentration and premium valuation multiples place CBRS in Decile 1 for Value and Decile 2 for Defensive, its elite Momentum (Decile 9) and inflecting Quality (Decile 5) make it an essential tactical vehicle for institutional investors navigating the next generation of AI infrastructure.
Sources
- 1.https://www.cerebras.ai/press-release/cerebras-systems-delivers-record-revenue-in-third-quarter-2026-raises-full-yea-3402195
- 2.https://www.stocktitan.net/sec-filings/CBRS/8-k-cerebras-systems-inc-reports-material-event-ef2cd93f6837.html
- 3.https://in.investing.com/news/stock-market-news/earnings-call-transcript-cerebras-tops-guidance-in-q3-2026-shares-jump-after-hours-93CH-5701337