J Keynes's $CBRS thesis: Cerebras fast inference scales on wins

J Keynes

Growth Investment Research

12 Nov, 23:56Latest

Cerebras' new hyperscaler wins and improving cloud economics strengthen the case that fast inference can become a large business. Microsoft joins AWS and Oracle; Q3 core cloud revenue rose 272% year over year, and cloud gross margin reached 46.9%, helped by higher pricing. AMD GPU prefill with Cerebras decode now serves cloud customers. Management still expects 2027 revenue to triple; delivery depends on data-center buildout, while Feldman says wafer supply is sufficient and yields are high. x.com x.com x.com x.com x.com

CBRS

11 Nov, 22:59

Cerebras's defect-tolerant manufacturing and in-field repair challenge the claims that its wafers are uneconomic and wear out after two years. Cerebras says independent voltage tuning across 84 die regions and spare cores help yields meet or exceed those of GPUs 1/58 the size; remaining redundancy can restore the compute fabric after failures without removing the wafer. The engineering detail strengthens the case for shares to return to $200 before too long. x.com

CBRS

This signal is no longer available. Showing the latest signals.

Andrew Feldman

13 Nov, 18:14

Co-Founder & CEO of Cerebras

We will be bringing up about 75% of our data center capacity outside the US in 2027. We have large builds underway in Canada. We have builds underway in France, in Norway, in Finland. youtube.com

AlvaSupportsCerebras’s expected revenue tripling hinges heavily on international construction and commissioning, making overseas delivery milestones central to the ramp.

Dow Jones

13 Nov, 17:40

Real-Time Financial News Wire

Morgan Stanley analysts say ... The company has fallen behind on delivering cloud capacity for a major customer, potentially driving small penalties and raising the possibility of a cancellation if the delay is prolonged. tradingview.com

AlvaChallengesWeakens confidence in Cerebras’s expected 2027 revenue tripling: the risk extends beyond delayed revenue to losing part of the contracted business.

Dow Jones

13 Nov, 17:54

Real-Time Financial News Wire

Analysts Matt Bryson and Tuo Ma say that rising GPU rental prices, increased costs and pricing from competitors, and contracted cloud capacity beyond the OpenAI partnership all bode well for 2027. tradingview.com

AlvaSupportsIf this view holds, Cerebras would have support for its 2027 revenue ramp from both pricing conditions and demand beyond OpenAI.

zerohedge

13 Nov, 13:20

Financial Blog and News Aggregator

Cerebras Systems (CBRS) gains 15% after the chipmaker gave a sales guidance well above estimates, while unveiling new deals with Oracle and Microsoft. x.com

AlvaSupportsThe Oracle deal adds evidence that Cerebras’s fast-inference business can expand across multiple hyperscalers.

Cerebras Systems Inc.

12 Nov, 21:12

AI infrastructure company

Core cloud revenue grew 272% x.com

AlvaSupportsStrengthens the case that Cerebras’s fast inference is converting into a large cloud business.

Reuters

13 Nov, 11:49

Global News Agency

On the earnings call, CEO Andrew Feldman said delivery delays under Cerebras' $20 billion multi-year OpenAI deal stemmed from industry-wide shortages of land and power, rather than chip manufacturing constraints tradingview.com

AlvaMixedThis makes data-center readiness a critical execution risk for Cerebras’s expected revenue tripling in 2027, while supporting management’s wafer-supply assurance.

Reuters

13 Nov, 10:32

Global News Agency

Cerebras jumped more than 16% after the chip and cloud company raised its 2026 revenue forecast and signed Microsoft as a customer. tradingview.com

AlvaSupportsAdding Microsoft broadens Cerebras’s hyperscaler customer base, strengthening the case for fast inference becoming a large business.

Reuters

12 Nov, 21:44

Global News Agency

Cerebras sees Q4 2026 core revenue of $310 mln to $330 mln ... Company expects Q4 2026 core gross margin of 47% to 49% ... Raises FY26 core revenue outlook to $940 to $960 mln, vs analysts' estimate of $887 mln tradingview.com

AlvaSupportsThe stronger growth outlook and expected margin expansion strengthen the financial basis for Cerebras’s return-to-$200 case.

Artificial Analysis

11 Nov, 23:46

AI Benchmarking Company

Take away its tools, and Claude Fable 5.2 scores below Claude Haiku 5.5 with tools on VIALS-AA; across models tool use nearly doubles scores, from ~25% to ~47% on average x.com

AlvaSupportsStrengthens the case for tool-using research agents on life-sciences tasks, giving substance to the kind of workload Cerebras could accelerate.

Eric Vishria

11 Nov, 19:47

General Partner at Benchmark

[Quoting @1538525334781370368: "It got Claude running on AMD in just 3 days with basically no human intervention. Claude ported or wrote almost 100% of the kernels."] If, otoh, the moat on your HW is actually HW, not only is your moat intact, but customers' ability and willingness to try your HW is dramatically increased. x.com

AlvaSupportsIf this holds for Cerebras, easier software integration would lower the hurdle for Anthropic to evaluate its hardware for research agents.

Ronnie Chatterji

11 Nov, 08:01

Chief Economist of OpenAI, Distinguished Professor at Duke University

We released a report on research acceleration, it's one of my most interesting reports right now, looking at how AI helps AI researchers, but also helps people across the company. If you think about three times the human hours, we're going to open up a lot of opportunities here. pdrl.fm

AlvaSupportsOpenAI frames AI-assisted research as potentially providing three times the human hours, supporting the thesis that agents can expand capacity for successor-model research.

Andrew Feldman

8 Nov, 14:05

Co-Founder & CEO of Cerebras

A person can do it by hand in 37 seconds. The assistants took up to 7 minutes and 40 seconds. So we built a prototype assistant on @Cerebras, tuned for this task. It booked the same table in 22 seconds, faster than a person could. x.com

AlvaSupportsThe reported 22-second booking supports Cerebras’s speed case for consumer agents, alongside parallel searches and task-specific tuning; the result covers one task.

SemiAnalysis

6 Nov, 04:59

Independent Semiconductor and AI Research Firm

OpenAI just announced that it will deprecate GPT-5.6 Sol Ultra Fast. This is the same model that Cerebras publicly announced it ran at ultra-fast speeds for. x.com

AlvaChallengesRetiring this Cerebras-powered ultrafast service creates a concrete checkpoint for whether its OpenAI workloads carry over to replacement models.

Ask Alva about this thesis

Ask Alva about this thesis