# Cerebras Systems (CBRS) — Investment Analysis

## Company & Research Metadata

- **Company:** Cerebras Systems
- **Ticker:** CBRS
- **Exchange:** NASDAQ
- **Sector / Industry:** Information Technology / AI semiconductors and inference infrastructure
- **Template version:** 2.7
- **Research date:** 2026-10-21 (comparison refresh; fundamental review preserved from 2026-10-18)
- **Valuation / score price snapshot:** $97.82 — 2026-11-03 regular-session close, America/New_York
- **Tracking tier:** On the Radar
- **Owned:** No
- **Primary operating period reviewed:** Q2 2026 and TTM through June 2026
- **Historical period reviewed:** FY2023–FY2025 and TTM June 2026
- **Latest investor presentation reviewed:** No separate presentation identified in the reviewed analysis
- **Latest earnings call reviewed:** Q2 2026 earnings call, 2026-08-12
- **Latest post-earnings management commentary reviewed:** August 2026 CS-4 announcement

- **Price snapshot date:** 2026-11-03

- **Price snapshot time / timezone:** Regular-session close, America/New_York (including exchange half-days)
- **Competitor / analyst-Q&A audit:** 2026-10-28; original operating, price and valuation snapshots retained.

## What's Changed Since Previous Review

**Current Review:** 2026-10-28

**Previous Review:** 2026-10-21 (comparison refresh; fundamental review preserved from 2026-10-18)

**Update Trigger:** Competitor relevance and analyst-Q&A correction

| Area | Previous View | Current View | Investment Impact |
|---|---|---|---|
| Competition | Preceding operating and comparable-company classifications | Product/customer/geographic boundaries corrected; direct rivals separated from substitutes, partners and valuation anchors | Corrects the competitive comparison; does not by itself re-underwrite the valuation |
| Analyst questions | Prior source-linked Q&A summaries | Complete questions and sentence-form management answers; all 6 existing rows retained | Improves readability while preserving the recorded exchange meaning |
| Valuation and scores | Approved operating/price snapshots, scenarios, scores and decisions | Preserved | No price refresh, rerank or change to valuation assumptions in this correction pass |

**Thesis:** No new fundamental thesis determination is made in this correction pass. The prior investment decision is preserved; the corrected competitive scope should govern the next substantive review.

## Investment Snapshot

Cerebras is one of the more technologically differentiated challengers to Nvidia in AI computing. Its wafer-scale architecture offers a compelling alternative for extremely fast AI inference, and commercial adoption is accelerating.

However, **a superior technology does not automatically make a superior investment.** At approximately 44 times trailing revenue, the stock still prices in a remarkable amount of future growth, despite having fallen sharply from its post-IPO highs.

The central investment question is whether Cerebras can turn its enormous contracted AI infrastructure opportunity into durable, profitable cash flow before competing architectures improve or customer demand shifts.

| Metric | Current View |
|---|---|
| Price Snapshot | $97.82 — 2026-11-03 regular-session close, America/New_York |
| Rating / Zone Status | Buy |
| Bear / Base / Bull Fair Value | $29.22 / $121.12 / $258.76 |
| Scenario Probabilities | 25% / 50% / 25% |
| Probability-Weighted Fair Value | $132.56 |
| Probability-Weighted Expected Return | +35.5% |
| Bear-Case Return | -70.1% |
| Strong Buy Zone | ≤ $32.15 |
| Buy Zone | >$32.15–$106.05 |
| Avoid Chasing Above | > $147.29 |
| Quality Score / Tier | 64.3 / Tier C |
| Opportunity Score / Rank | 59.2 / #25 of 76 |
| Risk Level | High |
| Tracking / Ownership | On the Radar / Not Owned |

**Decision rationale:** Cerebras has exceptional growth potential, but current shareholders are accepting substantial customer-concentration, execution, competitive, and dilution risks for a relatively small probability-weighted return in my model.

These fair values are my estimates, not management guidance or analyst consensus. They are particularly sensitive to assumptions about 2029 revenue, profitability, dilution, and terminal valuation.

## Investment Thesis

The strongest argument for Cerebras is that AI inference is becoming a more important part of AI computing economics.

Training a model is an expensive but relatively infrequent undertaking. Inference happens whenever users or software agents interact with that model. If AI agents increasingly perform complex tasks involving hundreds or thousands of model calls, the speed and cost of producing tokens could become critical commercial differentiators.

Cerebras is attempting to own the premium segment of this market.

Its wafer-scale architecture is optimized to deliver very high token-generation speeds. Rather than trying to replace all GPU computing, Cerebras can succeed by capturing workloads where latency has an unusually high economic value.

There are three possible routes to long-term per-share value creation.

First, Cerebras can sell accelerated computing systems to customers operating their own infrastructure. Second, it can operate its own cloud and sell inference services. Third, it can collaborate with companies such as AWS and AMD to integrate Cerebras hardware into larger computing ecosystems.

The investment opportunity would become particularly attractive if the third route turns Cerebras into a widely adopted complement to existing AI infrastructure rather than a niche GPU alternative.

### What the Market May Be Mispricing

The bullish interpretation is that investors underestimate the value of ultra-fast inference and the potential for Cerebras to expand from a hardware supplier into a much larger infrastructure-services business.

The bearish interpretation is that investors overestimate the economic value of its performance advantage, underestimate how quickly competitors can respond, and place too much confidence in long-duration contracts.

My assessment is that **the technical differentiation is more established than the eventual business economics**. That distinction explains why I am constructive on the company but less enthusiastic about the stock at its present valuation.

## Business Overview

Conventional semiconductor manufacturers fabricate numerous chips on a silicon wafer, then separate those chips for use in computers and servers.

Cerebras takes a fundamentally different approach. Its Wafer-Scale Engine integrates a very large number of processing cores and memory into a single wafer-scale processor. This minimizes some of the communication overhead that conventional multi-chip architectures must manage.

The architecture is particularly interesting for inference, where memory movement and generating each successive token can constrain performance.

Cerebras currently has two principal commercial revenue streams.

| Segment / Brand / Business Unit | Role | Revenue / Growth | Economics / Position | Strategic Takeaway |
|---|---|---|---|---|
| Hardware systems | CS-series computing systems and associated infrastructure | $54M of Q2 2026 GAAP revenue | Large individual transactions; establishes installed base | Validates architecture and supplies customer infrastructure |
| Cloud and other services | Access to Cerebras-powered inference infrastructure | $126M of Q2 2026 GAAP revenue; approximately 70% of total | Usage and contract-driven revenue; requires significant capital investment | Greater opportunity for recurring demand and infrastructure scale |

In Q2 2026, reported cloud and other services revenue reached approximately $126 million, compared with $54 million from hardware. Cloud represented about 70% of quarterly GAAP revenue, demonstrating how rapidly the business mix has shifted.

The customer relationships are also important.

OpenAI provides access to frontier-model workloads and contracted infrastructure demand. G42 and MBZUAI have been important customers and strategic relationships. AWS potentially gives Cerebras access to enterprise developers through Amazon Bedrock. AMD offers a route to integrating Cerebras with existing GPU-based infrastructure.

Cerebras is also winning contracts with application companies, including Cognition, Lovable, Figma, Block, and CrowdStrike.

The challenge is that these are not economically identical customer relationships. A large hyperscaler deployment, a subsidized or warrant-linked strategic contract, and an ordinary commercial inference customer can generate very different returns on invested capital.

### Business Takeaway

Cerebras is transitioning from a specialized hardware company into a vertically integrated AI infrastructure provider. That expands its revenue opportunity substantially, but it also introduces the capital intensity and utilization risks associated with operating data centers.

## Industry, TAM & Market Share

The total market for AI compute is enormous and expanding, but Cerebras's directly addressable opportunity is narrower than the total industry.

The relevant question is not how much companies spend on all AI hardware. It is how much inference spending will eventually require a level of speed and cost efficiency that Cerebras can deliver competitively.

Demand is supported by growth in coding agents, interactive AI assistants, real-time security applications, automated research, and increasingly sophisticated enterprise workflows.

Cerebras currently has a relatively small revenue base compared with established AI computing suppliers. That creates significant potential headroom, but the company must translate speed advantages into favorable cost per token, sufficient customer demand, and sustained system utilization.

### Market-Share Analysis

A consistent market-share dataset covering dedicated inference chips, GPUs, proprietary hyperscaler accelerators, and private competitors is not available. Reporting market-share percentages from different market definitions would therefore be misleading.

#### Competitive Scale / Position

| Company / Platform | Primary Position | Relative Scale | Growth Direction | Investment Relevance |
|---|---|---|---|---|
| Nvidia | Broad AI hardware and software ecosystem | Vastly larger | Strong incumbent | Leading competitor with entrenched ecosystem |
| AMD | GPU alternative with large enterprise and cloud relationships | Much larger | Expanding | Competitor and infrastructure partner |
| Google TPU | Proprietary AI acceleration | Large internal infrastructure deployment | Expanding | Hyperscaler alternative |
| AWS Trainium / Inferentia | Cloud-native AI acceleration | Large distribution advantage | Expanding | Cloud alternative and complementary route to distribution |
| Groq | Specialized low-latency inference | Private; directly relevant niche rival | Developing | Competes on inference speed |
| Cerebras | Wafer-scale inference and AI systems | Emerging | Rapid commercialization | Subject company |

### Industry Takeaway

The market opportunity is sufficiently large for Cerebras to become a meaningful business without displacing Nvidia. But the industry is also attracting some of the world's best-capitalized technology companies.

Cerebras's opportunity depends less on winning a universal accelerator market-share contest and more on building a sustainable position in workloads where its architecture offers measurable value.

## Customer Economics

This is an area where the investment thesis still requires considerably more evidence.

Cerebras reported approximately $25.4 billion in remaining performance obligations at June 30, 2026. That is an extraordinary figure relative to current annual revenue and demonstrates substantial contracted demand.

However, **RPO is not the same as cash received, profit, or guaranteed shareholder value**.

The contracts must be fulfilled over time. Cerebras needs data centers, equipment, electricity, operational capability, and customer utilization. The associated costs and shareholder dilution matter enormously.

The company also recognizes certain customer warrants as reductions of reported revenue. These arrangements complicate the interpretation of both revenue growth and customer profitability.

Most importantly, revenue remains highly concentrated. The three largest disclosed customer categories represented approximately 76% of Q2 2026 GAAP revenue.

The indicators I would prioritize are:

| Metric | What it reveals | Current assessment |
|---|---|---|
| Cloud revenue growth | Commercial demand and deployments | Very strong |
| Cloud gross margin | Cost efficiency of serving inference | Improving year over year, but volatile |
| Revenue concentration | Dependency on major contracts | High risk |
| Revenue per deployed MW | Infrastructure productivity | Insufficient disclosure |
| Utilization and cost per token | Competitive economics | Insufficient independent evidence |
| Customer retention / expansion | Durability beyond initial contracts | Limited history |

Management reported approximately 41.8% **core cloud gross margin** in Q2 2026, up considerably year over year. That is encouraging.

Management also expects higher margins as it replaces expensive rented infrastructure with owned systems and improves throughput. The longer-term ambition is gross margins above 60%.

That ambition is plausible in a favorable utilization environment, but it is not yet demonstrated at the company's intended scale.

### Customer-Economics Takeaway

Customers clearly value the product. What remains unproven is whether Cerebras can deliver that value at attractive returns after accounting for infrastructure spending, customer incentives, and dilution.

## Competitive Landscape & Moat

### Competitive Landscape

| Competitor | Market Share | Positioning | Advantages vs. Company | Disadvantages vs. Company | Current Direction |
|---|---|---|---|---|---|
| Nvidia | Not comparable | Direct: AI training/inference systems and GPU compute | CUDA ecosystem, scale, flexibility, distribution | Cerebras has high-speed inference specialization | Strong |
| AMD | Not comparable | Direct/partner: AI GPUs and compute systems | GPU portfolio and broader installed base | Cerebras offers specialized decode performance | Partner and rival |
| Google TPU | Not comparable | Substitute: TPU cloud compute; not a merchant chip sold on identical terms | Hyperscale infrastructure and internal demand | Cerebras offers an independent specialized platform | Strengthening |
| AWS Trainium | Not comparable | Partner and compute alternative: Trainium; joint prefill/decode deployment with Cerebras | Bedrock distribution and prefill capability | Joint solution complements Cerebras decode rather than simply displacing it | Commercial partnership; alternative architectures remain |
| Groq | Not disclosed | Direct segment: low-latency inference compute/services | Inference-focused architecture and specialization | Cerebras offers wafer-scale approach and large contracts | Competitive |
| Broadcom custom AI silicon | Not directly comparable | Indirect alternative: hyperscaler custom-chip development | Custom-design capability for hyperscalers | Design partner/supplier rather than a like-for-like independent inference cloud | Custom silicon is a make-versus-buy alternative |

### Moat Assessment

Cerebras's moat comes from an unusual combination of hardware architecture, interconnect design, systems engineering, compiler software, and practical expertise in running very large models.

Its August CS-4 announcement claimed up to 30 times faster inference than selected GPU-based alternatives and significant improvements in throughput per watt. These are company-reported benchmarks, not evidence of a universal 30-fold total-cost advantage.

There is a distinction between token speed and economic throughput. Generating tokens extremely quickly for one user is not necessarily equivalent to generating the most tokens per dollar across thousands of simultaneous users.

This is why its AMD collaboration is strategically interesting. Using GPUs for the *prefill* stage and Cerebras for *decode* could allow the two architectures to complement one another.

If successful, this could broaden Cerebras's market substantially. It could also reduce the need for customers to choose one architecture exclusively.

The opposing argument is that Nvidia, AMD, and hyperscalers can improve inference-specific software, memory architectures, networking, and scheduling. Nvidia's ecosystem advantage is especially difficult to replicate.

**Moat direction: Strengthening, but not yet proven durable.**

### Competitive Takeaway

Cerebras has real intellectual-property and engineering differentiation. I would not yet call it a wide-moat business because switching costs, ecosystem adoption, profitability, and the persistence of its performance advantage remain insufficiently established.

## Financial Profile

### Historical Financials

Financial reporting periods follow the issuer's fiscal year: completed FY rows plus standalone current-year quarters and cumulative H1/9M for FY2026. Revenue, FCF and SBC are flows; cash is a period-end balance. FCF = operating cash flow minus property/equipment capex from SEC company facts; quarterly FCF after Q1 and any quarterly SBC derived from YTD are visibly marked. Missing values are N/D, not zero. Unit: USD millions. Data extracted from SEC-tagged filings as of 2026-11-07. Company-specific FCF definitions and acquisition effects may require further reconciliation.

| Period | Revenue | Growth | Gross Margin | Operating Margin | FCF | FCF Margin | SBC | Diluted Shares | Cash | Debt |
|---|---:|---:|---:|---:|---:|---:|---:|---:|---:|---:|
| FY2023 | 78.7 | 219.9% | 33.5% | -170.1% | -85.6 | -108.7% | 26.6 | 44M | 152 | 1 |
| FY2024 | 290.3 | 268.6% | 42.3% | -35.0% | 428.5 | 147.7% | 58.6 | 49M | 337 | 41 |
| FY2025 | 510.0 | 75.7% | 39.0% | -28.6% | -392.8 | -77.0% | 49.8 | 172M | 1,108 | 262 |
| Q1 FY2026 | $193.406 | N/D | 44.6% | -7.8% | −$119.635 (derived: CFO - capex) | -61.9% | $9.593 | 62.806 | $1,716.016 | N/D |
| Q2 FY2026 | $180.11 | N/D | 14.2% | -265.0% | −$476.726 (derived: current YTD - prior YTD) | -264.7% | $377.008 (derived: current YTD - prior YTD) | 150.968 | $6,742.157 | N/D |
| H1 FY2026 | $373.516 | N/D | 29.9% | -131.8% | −$596.361 (derived: CFO - capex) | -159.7% | $386.601 | 107.132 | $6,742.157 | N/D |

#### Supplemental historical reporting context (retained)

#### Supplemental historical reporting context (retained)

#### Supplemental historical reporting context (retained)

The TTM comparison is calculated using the corresponding prior four quarters. The TTM share figure is an accounting-period weighted average affected by the IPO and should not be used to value today's equity.

### Financial Takeaways

#### Latest Quarter: Q2 2026

| Metric | Q2 2026 |
|---|---:|
| GAAP revenue | $180.1M; +74% YoY |
| Core non-GAAP revenue | $209.9M; +103% YoY |
| GAAP gross margin | 14.2% |
| Core gross margin | 40.6% |

The discrepancy between GAAP and core results is important. The company excludes certain effects from its core presentation, including customer-related warrant economics. Those warrants represent a real economic concession even if the accounting recognition occurs over time.

Similarly, Q2's exceptionally high stock-based compensation largely reflects IPO-triggered vesting, so the period should not be treated as a normal steady-state expense run rate. But dismissing all SBC would also understate shareholder costs.

#### Guidance and Growth

| Metric | Management outlook |
|---|---:|
| FY2026 core revenue | $880M–$890M |
| FY2026 core gross margin | 41%–43% |
| FY2026 core operating margin | -19% to -17% |
| FY2027 core revenue expectation | More than 3x FY2026 |
| Longer-term core gross margin aspiration | Above 60% |

The 2027 expectation implies core revenue above approximately $2.7 billion if the 2026 midpoint is achieved.

This would be remarkable growth. It would also require a major step-up in capacity deployment and successful recognition of contracted business.

Cerebras's balance sheet was transformed by its May 2026 IPO. At June 30, it had approximately $7.9 billion in cash and investments excluding restricted cash under standardized financial presentation, against approximately $1.5 billion in debt and debt-like obligations.

This provides a substantial funding cushion. Nevertheless, it should not be mistaken for permanently available excess cash.

Capital expenditures have accelerated dramatically, and free cash flow over the trailing twelve months was approximately negative $680 million. Cash requirements could remain substantial as the company builds capacity to service its contracted commitments.

The positive FY2024 free cash flow was heavily influenced by working-capital movements and customer funding; it was not evidence of normalized profitability.

**Financial direction: Mixed.** Growth and liquidity have improved sharply. GAAP profitability, free cash flow, and per-share economics remain areas of concern.

## Management & Capital Allocation

Andrew Feldman, Cerebras's co-founder and CEO, has led a remarkably unconventional semiconductor technology from development into meaningful commercial adoption.

Management deserves credit for demonstrating that wafer-scale computing can work outside the laboratory and for securing major customer relationships.

The company has also been willing to rethink its commercial model, evolving from hardware sales toward operating inference infrastructure and integrating with other processors.

However, management's capital-allocation record as a public company is still very short.

Three issues warrant particular scrutiny.

**Customer-linked warrants and dilution.** The OpenAI arrangement includes warrants for up to approximately 33.4 million Class N shares at a nominal exercise price, subject to specified conditions. Approximately 10 million were exercised in July 2026. AWS and other partners have additional warrant arrangements. These incentives may support customer acquisition, but they transfer part of the resulting value to strategic customers rather than ordinary shareholders.

**Financial controls.** The June 2026 filing identified material weaknesses in internal financial controls involving accounting expertise, IT controls, segregation of duties, and several important accounting functions. Management has remediation plans, but the issues were not yet resolved.

**Governance.** The multi-class share structure gives certain shares enhanced voting rights, limiting the influence of ordinary public shareholders.

### Management Takeaway

I view management as **technically impressive and commercially ambitious**, with a capital-allocation and governance record that has not yet earned a premium assessment.

The next several years must establish whether massive infrastructure expansion produces attractive *per-share* returns, not merely impressive revenue growth.

## Valuation

### Current Valuation

| Metric | Value |
|---|---:|
| Price Snapshot | $97.82 — 2026-11-03 regular-session close, America/New_York |
| Shares outstanding | 237.56M |
| Market Capitalization | $30.22B |
| Enterprise Value | $23.79B |
| Trailing GAAP revenue | $680.67M |
| Price / Sales | 44.4x |
| EV / Sales | 34.9x |
| Forward P/E | Not meaningful |
| Trailing FCF | -$680.21M |

Only the price row is refreshed here. Other quoted valuation statistics retain their earlier source dates and definitions.

### Comparable Company Analysis

| Company | Relevance | Public / Private | Valuation Metric 1 | Valuation Metric 2 | Growth / Margin / Risk Context |
|---|---|---|---:|---:|---|
| **Cerebras** | Subject company | Public | P/S 44.4x | EV/Sales 34.9x | Very high growth; unprofitable |
| Nvidia | Major compute competitor | Public | P/S 17.1x | EV/Sales ~17.0x | Diversified, highly profitable AI leader |
| AMD | Competitor and partner | Public | P/S 20.9x | EV/Sales ~20.7x | Larger semiconductor platform |
| Groq | Specialized inference rival | Private | N/A | N/A | Relevant architecture competition |
| Google TPU | Cloud-compute substitute; internal Alphabet platform | Internal platform | N/A | N/A | Part of Alphabet |
| AWS Trainium | Partner/complement in joint inference; alternative compute architecture | Internal platform | N/A | N/A | Part of Amazon |
| Broadcom | Secondary custom-silicon valuation anchor; not direct cloud equivalence | Public | P/S 17.6x | EV/Sales ~18.0x | Different business mix and margin profile |

Cerebras trades at a substantial premium to larger, more mature semiconductor competitors.

That premium is not inherently irrational. The company expects much faster growth than those businesses from a small starting base.

However, it makes comparison based on trailing revenue particularly unforgiving.

The market is effectively valuing Cerebras on its ability to transform into a multibillion-dollar AI infrastructure company over the next several years.

### Valuation Scenarios

I use a **three-year forward enterprise-value model** rather than a conventional near-term earnings multiple.

The model estimates revenue and enterprise value in 2029, adds estimated net cash at that time, divides by projected diluted shares, and discounts the resulting share value to October 2026 at 15% annually.

The revenue forecasts below use the company's *core* revenue definition. That is not perfectly comparable with the GAAP revenue used in the current trading multiples, so the resulting valuation should be treated as an approximate scenario framework rather than precise intrinsic value.

| Scenario | Key Assumptions | Fair Value | Probability | Probability Rationale | Return From Price Snapshot |
|---|---|---:|---:|---|---:|
| **Bear** | 2029 core revenue $2.0B; 5x EV/revenue; $2B net cash; 270M diluted shares | $29.22 | 25% | Growth falls short, business remains capital-intensive and valuation compresses | -70.1% |
| **Base** | 2029 core revenue $5.5B; 9x EV/revenue; $3B net cash; 285M diluted shares | $121.12 | 50% | Strong expansion with moderate mature-market multiple and warrant/SBC dilution | +23.8% |
| **Bull** | 2029 core revenue $9.0B; 13x EV/revenue; $5B net cash; 310M diluted shares | $258.76 | 25% | Broad inference deployment, strong adoption, favorable economics and sustained technology advantage | +164.5% |
| **Probability-Weighted** | Weighted average across scenarios | **$132.56** | **100%** | Scenario-weighted outcome | +35.5% |

The bear case assumes growth falls substantially short of management's expectations, the business remains capital-intensive, and investors eventually value it closer to an infrastructure or specialized hardware provider.

The base case requires strong 2027 growth and continued expansion through 2029, but assumes a more moderate mature-market revenue multiple. It also assumes dilution from warrants and stock compensation.

The bull case assumes Cerebras becomes a widely deployed inference platform with strong enterprise adoption, favorable economics, and a sustained technology advantage.

The bull case is possible. I am unwilling to make it the base case without stronger evidence of unit economics and customer diversification.

The probability-weighted calculation is:

(0.25 × $29.22) + (0.50 × $121.12) + (0.25 × $258.76) = **$132.56**

### Valuation Takeaway

- **Primary valuation method:** Discounted 2029 enterprise-value scenarios.
- **Cross-check methods:** Current and peer revenue multiples, liquidity and dilution analysis.
- **Key valuation dependencies:** 2029 revenue, utilization and margins, terminal multiple, shareholder dilution and net cash.
- **Valuation confidence / Risk Level:** Low to moderate confidence / High risk.
- **Margin-of-safety conclusion:** Insufficient at the analyzed $127.21 share price.

The valuation is much more sensitive to future execution than to near-term earnings.

A long-term valuation of $120–130 per share is plausible if Cerebras continues growing rapidly but eventually commands a lower sales multiple than the market currently awards.

The present price therefore appears **reasonable for an optimistic business trajectory, but not compelling enough for a new investment under a margin-of-safety discipline**.

#### Valuation Zones

| Zone | Price boundary |
|---|---|
| Strong Buy | At or below $32.15 |
| Buy | $32.16–$106.05 |
| Hold / Watch | $106.06–$147.29 |
| Avoid | Above $147.29 |

These boundaries follow the established research methodology. The unusually low Strong Buy threshold is a consequence of the severe bear-case downside, not a prediction that the shares will fall to that level.

## Analyst Questions

The latest reviewed earnings call was Q2 2026. The material exchanges focused overwhelmingly on customer concentration, partner monetization, capacity, and the economics of disaggregated inference.

| Analyst / Firm | Core Question | Management Response | Research Read-Through |
|---|---|---|---|
| Timothy Arcuri / UBS | Will OpenAI and AWS dominate 2027 revenue? How fast can manufacturing scale? | OpenAI remains significant, while AWS and others should gradually diversify revenue. Manufacturing capacity is expanding rapidly. | Customer concentration persists despite new partnerships. |
| Joshua Buchalter / TD Cowen | How will AWS commercialization work? How does AMD integration fit with its broader AI roadmap? | Management expected Bedrock availability in Q1 2027 and said AMD Helios and Cerebras will provide combined prefill/decode infrastructure. | Promising distribution and technical partnership, with economics still emerging. |
| Kyle Buser / Barclays | Does Cerebras retain revenue when deploying AMD equipment in its own cloud? Could AWS-related solutions eventually run in Cerebras infrastructure? | Cerebras retains the cloud-service revenue under the described deployment model. Additional hosting configurations are possible. | Supports the infrastructure-platform thesis, but details are incomplete. |
| Quinn Bolton / Needham | Does improved throughput create additional revenue under existing OpenAI agreements? Could future standalone improvements reduce the need for disaggregation? | Better throughput improves potential revenue and margins. Management declined to disclose detailed OpenAI economics and emphasized continuing architecture improvements. | One of the most important unanswered unit-economics questions. |
| Joe Moore / Morgan Stanley | How commercially mature is disaggregated inference? How does GPU generation affect results? | Management said solutions were running in labs, with commercial deployment expected in Q4 2026. It said stronger GPU generations should improve performance. | Commercial proof remains a near-term catalyst. |
| Vijay Rakesh / Mizuho | Will contracted power and manufacturing expansion accelerate revenue in 2027? | Management said capacity is central to growth and that it is aggressively pursuing additional sites. | Execution and deployment timing are critical. |

### Analyst-Q&A Takeaway

The questions show a clear disconnect between the technological story and the financial evidence investors still need.

Management articulated a compelling case for larger and faster infrastructure deployments, but several responses did not quantify the actual revenue, costs, or incremental economics of the most important partnerships.

The most material outstanding question is not whether Cerebras can serve models quickly. It is **how much sustainable gross profit and free cash flow each new dollar of deployed capacity can generate**.

## Thesis Monitoring

| Driver | Current State | Positive Trigger / Catalyst | Negative Trigger / Falsifier | What It Changes |
|---|---|---|---|---|
| OpenAI contracts | Large concentration and backlog | Timely ramp and strong economics | Delays, reduced commitments or weak margins | Revenue and customer economics |
| AWS distribution | Launch expected Q1 2027 | Material enterprise revenue beyond OpenAI | Limited customer adoption | Diversification and platform relevance |
| AMD disaggregated inference | Commercial deployment expected Q4 2026 | Demonstrated throughput economics | Technical promise fails to monetize | Addressable market and differentiation |
| CS-4 adoption | New system announced | Successful production deployment | Benchmark advantages fail to translate to customer ROI | Technology moat |
| Data-center capacity | 600MW live or contracted for delivery by end 2027 | On-time, economically productive capacity | Delays, overruns, weak utilization | Contract conversion and capital returns |
| Gross margins | Core GM around 41% | Sustained improvement toward management targets | Structural margin pressure | Profitability |
| Cash flow | Substantial negative FCF | Improving cash conversion as deployments scale | Persistent high burn and new financing | Liquidity and per-share value |
| Share dilution | Major warrants and equity compensation | Strong per-share revenue and FCF growth | Value creation disproportionately captured by new shares | Shareholder economics |
| Financial controls | Material weaknesses outstanding | Verified remediation | Continued deficiencies or restatement | Risk and management assessment |

The most important upcoming checkpoints are the next quarterly earnings report, expected in November 2026, and initial commercial evidence from the AMD and AWS deployment programs.

## Quality & Opportunity Scorecards

### Quality Score

| Component | Score | Weight | Weighted Contribution | Evidence |
|---|---:|---:|---:|---|
| Business durability and economics | 58 | 25% | 14.5 | Strong product demand, immature long-term economics |
| Moat and competitive position | 70 | 25% | 17.5 | Differentiated architecture and valuable IP |
| Growth runway and durability | 85 | 20% | 17.0 | Exceptional opportunity and contracted demand |
| Financial quality and balance sheet | 44 | 15% | 6.6 | Strong liquidity, negative FCF and dilution |
| Management, governance and capital allocation | 58 | 15% | 8.7 | Technically strong, governance and control concerns |
| **Quality Score** | | **100%** | **64.3** | **Tier C** |

- **Quality Tier:** C
- **Speculative override:** None

This is a price-independent assessment. Cerebras has above-average technology and growth potential, offset by below-average financial durability and substantial execution uncertainty.

I have not applied a separate speculative override because the company is already commercial and generating meaningful revenue.

### Opportunity Score

| Component | Raw Input | Normalized Score | Weight | Weighted Contribution |
|---|---|---|---|---|
| Quality | 64.3 | 64.3 | 30% | 19.3 |
| Probability-weighted upside | +35.5% | 100.0 | 35% | 35.0 |
| Bear-case downside | -70.1% | 0.0 | 20% | 0.0 |
| Risk / valuation confidence | 33 | 33.0 | 15% | 5.0 |
| **Opportunity Score** |  | **59.2** | **100%** | **59.2** |

- **Opportunity Rank:** #25 of 76
- **Zone Status:** Buy

## Portfolio Context

- **Tracking tier:** On the Radar
- **Owned:** No recorded position
- **Tracking source of truth:** ontheradar.md
- **Portfolio source of truth:** portfolio/holdings.md
- **Scoring independence:** Ownership, account, share count, position size and cost basis do not alter Quality Score, Opportunity Score, rank, rating, valuation or valuation zones.

## Investment Decision History

| Timestamp | Price | Rating | Zone Status | Buy Zone | Strong Buy Zone | Avoid Chasing Above | Base FV | P/W FV | P/W Return | Quality | Opportunity / Rank | Change / Rationale |
|---|---:|---|---|---|---|---|---:|---:|---:|---:|---|---|
| 2026-10-18 (2026-10-16 close) | $127.21 | Hold / Watch | Hold / Watch | $32.16–$106.05 | ≤$32.15 | $147.29 | $121.12 | $132.56 | +4.2% | 64.3 | 44.7 / Unranked | Initial analysis |
| 2026-10-21 (America/New_York) | $118.24 | Hold / Watch | Hold / Watch | >$32.15–$106.05 | ≤ $32.15 | > $147.29 | $121.12 | $132.56 | +12.1% | 64.3 | 50.2 / #35 | Common close and shared-universe rank; prior fundamental assumptions and zones preserved |

#### Final Investment Assessment

Cerebras is an unusually promising AI infrastructure company. Its architecture is differentiated, its customer roster is improving, and its growth ambitions are supported by considerable contractual demand.

The opportunity is also unusually difficult to value. Much of the prospective revenue lies several years in the future, customer economics have not been proven at scale, and the number of shares participating in future profits may rise substantially.

At **$127.21**, I would maintain a **Hold / Watch** stance rather than initiate a full position solely because the stock has fallen sharply since its IPO.

The key events that could change this assessment are successful AWS and AMD commercialization, evidence of improving gross margins without excessive capital spending, greater revenue diversification, and tighter visibility into how contract growth translates into free cash flow.

**My conclusion: attractive technology, potentially exceptional long-term business, but insufficient margin of safety at the current valuation.**

---

## Material Sources

| Source | Type | Date / Period | What It Supports |
|---|---|---|---|
| [Cerebras Q2 2026 earnings release](https://investors.cerebras.ai/news-releases/news-release-details/cerebras-systems-fast-inference-cloud-business-nearly-quadruples) | Company IR | 2026-08-12 | Results, business developments and guidance |
| [Cerebras Form 10-Q](https://www.sec.gov/Archives/edgar/data/2021728/000162828026056357/cbrs-20260630.htm) | Regulatory filing | 2026-06-30 | Backlog, customers, warrants, SBC, risks and controls |
| [Q2 2026 earnings-call transcript](https://stockanalysis.com/stocks/cbrs/transcripts/686677-q2-2026/) | Earnings-call transcript | 2026-08-12 | Management commentary and material analyst Q&A |
| [CS-4 announcement](https://investors.cerebras.ai/news-releases/news-release-details/cerebras-unveils-cs-4-30-times-faster-gpu-based-solutions) | Company IR | 2026-08-18 | Product roadmap and claimed performance |
| [Cerebras financial statements](https://stockanalysis.com/stocks/cbrs/financials/) | Financial data | FY2023–TTM June 2026 | Revenue, margins, balance sheet and FCF |
| [Cerebras cash-flow statements](https://stockanalysis.com/stocks/cbrs/financials/cash-flow-statement/) | Financial data | FY2023–TTM June 2026 | Capital expenditures and cash generation |
| [Cerebras valuation statistics](https://stockanalysis.com/stocks/cbrs/statistics/) | Market data | 2026-10-16 | Price, shares, multiples and market cap |
| [Nvidia valuation](https://stockanalysis.com/stocks/nvda/statistics/) | Market data | 2026-10-16 | Comparable-company valuation |
| [AMD valuation](https://stockanalysis.com/stocks/amd/statistics/) | Market data | 2026-10-16 | Comparable-company valuation |
| [Broadcom valuation](https://stockanalysis.com/stocks/avgo/statistics/) | Market data | 2026-10-16 | Secondary valuation anchor |
| [CBRS common regular-close snapshot](https://stockanalysis.com/stocks/cbrs/history/) | Market data | 2026-10-20 16:00 EDT | Refreshed price, scenario returns and normalized Opportunity Score / shared rank |
| [AWS–Cerebras joint inference announcement](https://press.aboutamazon.com/aws/2026/3/aws-and-cerebras-collaboration-aims-to-set-a-new-standard-for-ai-inference-speed-and-performance-in-the-cloud) | Regulatory filing / primary product evidence | 2026-03-13; checked 2026-10-28 | Trainium prefill/Cerebras decode complement; avoids classifying a joint solution as pure rivalry |
| [SEC XBRL company facts — CBRS](https://data.sec.gov/api/xbrl/companyfacts/CIK0002021728.json) | Regulatory filing (as tagged) | FY2026, through 2026-06-30 | Current-year standalone-quarter and year-to-date source facts for FCF, SBC, revenue and other reported metrics |