Cerebras AI Vendor Risk Profile
AI supercomputing company building wafer-scale engine (WSE) chips and Cerebras Inference cloud. Offers the world's largest AI chip and competes with Nvidia for training and inference workloads. Filed for IPO in 2024.
Risk overview
Risk score: 35/100
Risk tier: Moderate
Safety rating: 65/100 (higher is better)
Lower risk scores indicate lower assessed risk.
Last verified: Apr 19, 2026 Aging
Risk dimensions
| Dimension | Risk score |
|---|---|
| Data Handling | 28/100 |
| IP Exposure | 26/100 |
| Jurisdiction | 21/100 |
| Security | 40/100 |
| Regulatory Compliance | 50/100 |
| Transparency | 70/100 |
| Business Stability | 42/100 |
| Dependency Chain | 34/100 |
| Agent Governance | Not assessed |
Analyst summary
Rating: Acceptable
Cerebras builds wafer-scale AI systems and operates a fast inference cloud, with clean data-handling terms and SOC 2 Type II. The business is heavily concentrated in its G42 (UAE) relationship, and an in-flight CFIUS review has delayed its IPO, creating real near-term uncertainty for enterprise procurement.
Bottom line: Acceptable for commercial workloads that can tolerate concentration and CFIUS overhang; not a default safe pick.
Strengths
- Unique wafer-scale CS-3 hardware and fast open-model inference cloud
- No training on customer data, SOC 2 Type II, DPA available
- On-premises CS-3 deployment possible for sovereign and classified workloads
- Viable Nvidia-alternative for training and high-throughput inference
Concerns
- G42 (UAE) represented approximately 87% of 2023 revenue per S-1 filing
- IPO delayed by CFIUS review of the G42 commercial relationship
- Enterprise compliance footprint still maturing relative to hyperscalers
- Model menu focused on open-source models (Llama family); limited proprietary coverage
Best for
- Research and HPC teams needing wafer-scale training throughput
- Latency-focused open-model inference workloads
- Workloads willing to underwrite near-term IPO and concentration risk
Avoid if
- You require a FedRAMP-authorized inference path today
- Your procurement cannot accept a vendor with 80%+ revenue concentration
- You are a US government contractor sensitive to CFIUS-reviewed foreign commercial ties
Citations
- Data Handling — Data Retention Period
Cerebras retains API request logs for up to 30 days for abuse monitoring and operational purposes, after which they are deleted.
- Data Handling — Hipaa Baa Available
HIPAA Business Associate Agreements are available for Cerebras Inference enterprise customers processing protected health information.
- Data Handling — Trains On User Data
Cerebras does not use Customer Content submitted to the Cerebras Inference API to train or improve its foundation models.
- Governance — Financial Stability
Cerebras faces significant revenue concentration in G42, which accounted for 87% of 2023 revenue per S-1 filing.
- Governance — Government Scrutiny
Cerebras Systems' IPO has been delayed pending CFIUS review of its commercial relationship with G42, an Abu Dhabi-based AI firm.
- Governance — Strategic Investors
G42 has committed to purchase $1.43 billion of Cerebras systems and services, making it Cerebras' largest customer by revenue.
- Jurisdiction Profiles — Incorporation Country
Cerebras Systems, Inc. is headquartered in Sunnyvale, California and incorporated in Delaware.
- Security Compliance — Soc2 Type2
Cerebras Systems maintains SOC 2 Type II compliance covering its Inference Cloud service.