The answer first
The technology is relevant. The stock is not yet attractive.
Cerebras has one of the most differentiated architectures in public AI infrastructure. Its wafer scale design offers credible speed advantages in selected inference workloads. The problem is price, customer concentration, dilution, and the capital required to deliver contracted capacity.
What must go right
Convert the $25.4 billion performance obligation into recognized revenue.
Deploy 750 megawatts for OpenAI on time and economically.
Lift core gross margin above 45% while revenue accelerates.
Reduce reliance on OpenAI and a small group of large customers.
Scenario map
Reference price $209.85. Probabilities: 25% bear, 50% base, 25% bull. Probability weighted value: $179.
Valuation is the gating issue
Capital structure matters
| Metric | Basic | Fully diluted |
|---|---|---|
| Enterprise value | $43.4B | $57.6B |
| EV to 2026 revenue | 48.9x | 64.9x |
| EV to 2027 revenue | 14.7x | 19.5x |
Options, restricted units, performance units, and customer warrants materially change the economic valuation.
Reverse DCF burden
Required annual revenue growth through 2035 using a 30% terminal free cash flow margin.
If the terminal margin is 25%, required growth rises to roughly 37%.
Peer context
| Company | Forward sales | Gross margin | Economic contrast |
|---|---|---|---|
| Cerebras | 14.7x basic 2027 EV | 40.6% core | Faster growth, thinner margins, extreme concentration |
| NVIDIA | 12.0x | 74.2% | Dominant software and platform economics |
| AMD | 11.3x | 55.7% | Broader product base and positive cash flow |
| Broadcom | 12.2x | 76.3% | Custom silicon plus strong free cash flow |
| Nebius | 8.8x | 74.3% | Rapid cloud growth and heavy capital needs |
| CoreWeave | 2.5x | 67.4% | Large debt and negative free cash flow |
Conviction scorecard
Five weighted pillars
Pass, no initial position
Horizon: 18 to 24 months.
Key risk: delayed or uneconomic conversion of the OpenAI capacity commitment.
Business Quality decomposition
| Component | Score | Read |
|---|---|---|
| Quality of each revenue dollar | 4.0 | Margins and returns remain weak, with extreme customer concentration. |
| Quantity of revenue dollars | 8.5 | Exceptional historical and expected revenue growth. |
| Average | 6.3 | Growth offsets incomplete economic quality. |
| Margin trajectory modifier | 0 | Margins are mixed rather than clearly expanding. |
| Final Business Quality | 6.3 |
Operating evidence
Hardware: differentiated, not fortress economics
Q2 GAAP hardware revenue was $54.1 million. Core hardware revenue was approximately $82.1 million, with a 38.8% core gross margin.
The moat is architectural IP, specialized system design, and wafer scale software. The challenge is NVIDIA’s integrated platform, hyperscaler silicon, and limited independent total cost evidence.
Cloud: the growth engine, with capital intensity
Q2 GAAP cloud and services revenue was $126.0 million, up 281% year over year. Core cloud margin declined from 52.9% in Q1 to 41.8% in Q2.
The OpenAI contract validates demand but concentrates execution, financing, and dilution risk in one relationship.
Management critique
| Flag | Claim | Independent read |
|---|---|---|
| RED | $25.4B RPO demonstrates durable demand | Real, but heavily tied to OpenAI and long delivery periods. |
| YELLOW | CS 4 is up to 30 times faster | Workload specific company testing without broad independent cost validation. |
| RED | Liquidity is $8.6B | Includes restricted assets. Unrestricted cash and investments were closer to $7.9B. |
| GREEN | Q2 core revenue guidance near $194M | Actual core revenue was $209.9M, a material beat. |
| GREEN | Q2 core gross margin guidance at 36% to 38% | Actual core gross margin was 40.6%. |
What changes the decision
Reassessment policy
Research archive
Primary sources
Q2 2026 earnings call transcript
The complete source list is included in the PDF and workbook.