How much will manufacturing capacity grow in 2027, following the 10x increase in 2026?

Manufacturing capacity will be well over 10x by end of 2026, with contracted facilities already 3x-4x more for 2027 growth, and more contracting to come. “We already have contracted facilities 3x or 4x more for growth in 2027, and we still have some time to contract for more.”

How will the Amazon deal economics work, and when will it be available?

The service will be deployed in Amazon data centers, delivered through Amazon's Bedrock API service, with deployments currently being organized. It is expected to be live in Q1 2027. “It will be available. It is deployed in Amazon data centers. It will be delivered through Amazon's API service, Bedrock... We expect the service to be live in Q1.”

If Cerebras buys AMD Helios racks for its cloud, does Cerebras keep all the revenue from customers renting the disaggregated solution, or is there revenue sharing?

Cerebras keeps all the revenue from the disaggregated solution when it purchases and stands up AMD Helios racks in its cloud. “Yes, to the first part of the question.”

CBRS

Cerebras (CBRS) Q2 FY26 earnings call: guidance and Q&A

Earnings call· Jun 2026 (Q2 FY26)· Information Technology· Aug 12, 2026

Source: earnings call transcript

Executive summary

Cerebras Q2 FY26 — Record Core Revenue and Strategic Partnerships Drive Future Growth

Cerebras Systems delivered a strong Q2 FY26, exceeding guidance across all core metrics, driven by exceptional demand for fast inference and significant capacity expansion. The company is building foundational capabilities through strategic partnerships with AWS and AMD for disaggregated inference, expanding its market reach and improving economics. With a substantial $25.4 billion RPO, Cerebras is poised for massive revenue growth in 2027 and beyond, despite near-term gross margin pressures from temporary capacity rental.

Highlights

Core revenue reached $209.9 million, marking a 103% year-over-year increase and exceeding guidance.
Core cloud and other services revenue surged by 287% year-over-year to $127.7 million.
Core gross margin improved significantly to 40.6%, up 940 basis points year-over-year, beating expectations.
Core operating margin improved to negative 16% from negative 42% a year ago, an improvement of 2,600 basis points.
Secured over 600 megawatts of data center capacity by the end of 2027, with a pipeline measured in gigawatts.

Concerns

Core gross margin sequentially declined to 40.6% in Q2 FY26 from 46.5% in Q1 FY26, primarily due to higher costs from temporarily renting back systems.
Q3 FY26 core gross margin is expected to be the low point, projected in the range of 38% to 40%.

Guidance & targets

Category	Target	Timeline	Confidence
Core Revenue	$214 million to $216 million high materiality	Q3 · FY26	High
Core Gross Margin	38% to 40% medium materiality	Q3 · FY26	High
Core Operating Margin	minus 25% to minus 23% medium materiality	Q3 · FY26	High
Full Year Core Revenue	$880 million to $890 million high materiality	FY26	High
Full Year Core Gross Margin	41% to 43% medium materiality	FY26	High
Full Year Core Operating Margin	negative 19% to negative 17% medium materiality	FY26	High
Core Revenue Growth	more than triple high materiality	FY27	High
Core Revenue Growth	continue to grow at multiples high materiality	FY28 and FY29	High
Margin Expansion	significant margin expansion high materiality	FY27 and beyond	High
Core Gross Margin	60% plus high materiality	Long-term	High

Orderbook & backlog

Remaining Performance Obligations (RPO) $25.4 billion June 30, 2026

Does not reflect any backlog of business from AWS or any other hyperscaler at this time. Provides visibility and high confidence in future revenue growth.

Product announcements

Product	Type	Timeline	Details
CS4	launch	Next week
CS5	roadmap	Second half of 2027

Deals & partnerships

OpenAI Strategic partnership for delivery of GPT-5.6 Sol and providing fast inference cloud services.

Cerebras serves GPT-5.6 Sol 10x faster than others, providing scale and frontier insight.

Amazon Web Services (AWS) Collaboration to offer Cerebras solutions through AWS' Bedrock platform.

Solutions will be deployed in Amazon data centers and delivered via Bedrock API. Solutions generally available in Q1 2027.

AMD Collaboration to build disaggregated inference solutions combining AMD Helios racks with Cerebras CS systems.

Helios racks handle prefill, Cerebras systems handle decode. Solutions are running in labs and expected to be deployed and available in Q4 2026.

Figma New agreement for AI coding services.

Expands Cerebras' footprint in the AI coding category.

Cognition New agreement for AI coding services.

Expands Cerebras' footprint in the AI coding category.

Lovable Major win in Europe for AI coding services.

Extends Cerebras' presence in Europe for AI coding.

Block New agreement to leverage fast inference for custom-made agents.

Supports agentic flows, where the value of speed compounds.

AlphaSense New agreement to leverage fast inference for custom-made agents.

Supports agentic flows, where the value of speed compounds.

GSK New agreement to leverage fast inference for custom-made agents.

Supports agentic flows, where the value of speed compounds.

CrowdStrike Win for an AI-based security application.

Application only exists if AI is fast, enabling LLMs to secure enterprise traffic invisibly without delay or disruption.

U.S. government Partnerships for delivery of stacked memory solutions and integrated wafer-scale optical solutions.

Part of Cerebras' invention engine, focusing on pioneering advances.

Capital programs

Data Center Capacity Expansion Underway

Funding: Primarily Cerebras' own hardware (lower BOM cost); meaningful portion of fit-out reimbursed by largest customer

Start: Over the last 7 months

Benefit:Over 600 MW live or under contract by end of 2027; pipeline measured in gigawatts

Aggressive push to secure and build out data centers in Alabama, Dallas, Denver, Minneapolis, Santa Clara, Stockton, France, Finland, Manitoba, Montreal, Norway, Saskatchewan, and Toronto. Focus on inference allows flexibility in site selection, not requiring gigawatt training clusters.

Manufacturing Capacity Expansion Underway

Benefit:Increase by more than 10x in 2026; 3x-4x more contracted for 2027

Building new factories with Flex and Sanmina to support exceptional growth expected in the years ahead.

Risks & headwinds

Data center capacity bottleneck Ongoing

The faster we and our customers bring on new data centers, the faster we grow.

Mitigation:Aggressive push to secure and build out data centers, repeatable process for site selection and deployment, leveraging inference needs (not gigawatt training clusters) for flexibility, and a pipeline of new opportunities measured in gigawatts.

Temporary reduction in core gross margin Q2 FY26 and Q3 FY26

Q2 core gross margin 40.6% vs 46.5% in Q1 FY26. Would have been ~500 bps higher without rented capacity. Q3 expected to be low point (38%-40%).

Mitigation:Expects Q3 to be the low point, with significant improvement in Q4 FY26 as more data centers with lower-cost Cerebras-owned systems come online. Long-term target of 60%+ gross margin.

What to watch in Q3 FY26

Core Gross Margin trajectory

Q4 FY26

Current 40.6% (Q2 FY26), 46.5% (Q1 FY26)

Target Improvement from Q3 low point (38%-40%)

Why it matters

Management expects Q3 to be the low point and Q4 to improve significantly as owned systems replace rented capacity, impacting profitability.

Looking forward, we expect Q3 to be the low point for core gross margin before improving significantly in Q4 '26 as we bring on more data centers filled with lower-cost Cerebras-owned systems.

Find out whether CBRS delivers this. Inve re-checks your case every morning.

Q&A highlights

How will customer concentration evolve in 2027 with AWS ramping, given OpenAI's current large share? What about other hyperscalers?

OpenAI will remain a meaningful portion of revenue next year but will shrink as a percentage over time as AWS and other customers (coding, security) grow. Cerebras has a history of expanding its customer base from initial wins (government, sovereign cloud, frontier lab, now AWS). Discussions with other hyperscalers are ongoing.

“OpenAI will shrink as a percentage of our revenue over time. But I think you can expect for next year them to still be a meaningful portion of our revenue.”

asked by Timothy Arcuri · answered by Andrew Feldman

2 min read 5 chapters

Detailed narrative

Capacity Expansion & Supply Chain

Cerebras has aggressively expanded its data center footprint, securing over 600 megawatts by the end of 2027 with a gigawatt pipeline, leveraging its ability to scale inference capacity across diverse locations globally. Manufacturing capacity is set to increase by over 10x in 2026 through new factories with Flex and Sanmina.

This expansion is supported by strong relationships with TSMC for 5nm wafer supply, which is less constrained and more cost-effective than leading-edge nodes, reinforcing confidence in future growth plans.

Disaggregated Inference Strategy

The company is pioneering disaggregated inference solutions with AMD (Helios) and AWS (Trainium), separating prefill (GPUs) and decode (Cerebras WSE). This approach significantly increases throughput by 5x with AMD while maintaining Cerebras' industry-leading speed. This improves inference economics by producing more high-value tokens per system at lower cost and power, and enables customers to leverage existing GPU investments, expanding the market for both Cerebras and its partners.

Product Roadmap & Innovation

Cerebras continues rapid engineering execution, planning to double system speed annually for several years, starting with a 15x performance advantage. The CS4 system will be unveiled next week at the Supernova Conference, and the CS5 is on track for launch in H2 2027.

Long-term innovation includes significant partnerships with the U.S. government for stacked memory and integrated wafer-scale optical solutions, focusing on advancements in chip architecture, packaging, I/O, and power delivery to reduce cost per token.

Customer Growth & Market Expansion

Cerebras is expanding its customer base beyond OpenAI, with AWS solutions generally available in Q1 2027 via Bedrock, providing global enterprise reach. The company signed 6 deals over $30 million in Q2, growing in AI coding (Figma, Cognition, Lovable) and agentic flows (Block, AlphaSense, GSK).

New markets like security (CrowdStrike win) are emerging, enabled by fast AI for real-time traffic inspection without user disruption, creating massive new opportunities.

Financial Performance & Outlook

Q2 FY26 saw record core revenue of $209.9 million (up 103% YoY) and strong core gross margin of 40.6% (up 940 bps YoY), both exceeding guidance. The company raised its full-year 2026 guidance for core revenue ($880M-$890M), gross margin (41%-43%), and operating margin (-19% to -17%).

With $25.4 billion in RPO and $8.6 billion in cash, Cerebras is well-positioned for substantial revenue and margin expansion in 2027 and beyond, targeting 60%+ gross margins.

Summary of the company's earnings call; see how call summaries are prepared. Not investment advice.