CBRS
Cerebras Systems Inc.
Buzzberg insights
Attention & activity
- unique authors
- 7
- vs usual
- +60%
- unique authors
- 39
- vs usual
- +108%
Author calls
- Long
- 100.0%Authors: 3
- Short
- 0.0%Authors: 0
- Avoid
- 0.0%Authors: 0
- Long
- 83.3%Authors: 5
- Short
- 16.7%Authors: 1
- Avoid
- 0.0%Authors: 0
Author positions
- Long
- 100.0%Authors: 3
- Short
- 0.0%Authors: 0
- Trimmed
- 0.0%Authors: 0
- Closed
- 0.0%Authors: 0
- Long
- 100.0%Authors: 4
- Short
- 0.0%Authors: 0
- Trimmed
- 0.0%Authors: 0
- Closed
- 0.0%Authors: 0
Targets
1 dot = 1 author
Feed & authors
- ideas
- 787
- Authors
- 171
- Sources
- 132
News
Last 24 hours · Nov 18, 01:01
Events & watchlist
What authors are watching
Call performance
Performance since each author’s call.
| Author | Direction | Return |
|---|---|---|
| x256xxMay 15 | Short | +63.1% |
| photonbullAug 05 | Short | +52.1% |
| u/Fearless-Elephant-81Sep 30 | Short | +41.9% |
| Alexander GoodSep 30 | Short | +41.8% |
Best calls over the past year · return since each call
Ownership
Tracked fund positions and author disclosures; coverage is not market-wide.
Company facts
- Market cap
- $24.54B 2026-11-17
- Revenue
- $210.81M Q3 FY2026
- Revenue growth · YoY
- +55.3% Q3 FY2026
- Gross margin
- 20.0% Q3 FY2026
- Net income
- -$161.22M Q3 FY2026
- Free cash flow
- -$838.59M Q3 FY2026
Financial details
- Revenue
- $210.81M
- Revenue growth · YoY
- +55.3%
- Gross margin
- 20.0%
- Operating margin
- -106.3%
- Net income
- -$161.22M
- Free cash flow
- -$838.59M
- Operating cash flow
- -$341.41M
- Diluted EPS · GAAP
- -$0.67
- Capital expenditure
- $497.19M
Reported 2026-11-12 · Period ended 2026-09-30
Market cap · polygon · 2026-11-17
FMP · Q3 FY2026
Who holds CBRS
Company filings, tracked 13F funds and positions disclosed by Buzzberg authors.
Tracked fund positions · 13F
5 holders · 12 tracked managers| Fund | Reported equity weight | Weight change | Report date | Filings |
|---|---|---|---|---|
| Altimeter Capital Management, LP | 10.32% | -5.93 pp | SEC filings 2026-11-16 | |
| Atreides Management, LP | 5.77% | -0.40 pp | SEC filings 2026-11-16 | |
| Tiger Global Management | 3.92% | +1.15 pp | SEC filings 2026-11-16 | |
| Renaissance Technologies | 0.15% | +0.15 pp | SEC filings 2026-11-12 | |
| Citadel Advisors | 0.01% | -0.15 pp | SEC filings 2026-11-16 |
Weight is the position’s share of the fund’s reported portfolio, not its ownership of the company.
Changes compare portfolio weights between reports, not share purchases or sales. 13F filings can arrive up to 45 days after quarter-end.
Insiders & strategic holders
Reviewed insider and strategic ownership filings are not available yet.
Who holds CBRS
22 authors shownLatest disclosed long positions within 90 days · May include options · Not a current portfolio
No disclosed holders in this category.
Events & watchlist
Confirmed company dates and the developments authors are watching.
What authors are watching
Source-backed expectations and catalysts; timing may not be confirmed. · Nov 18, 01:01
When does the Microsoft agreement begin contributing revenue?
Management said revenue from the Microsoft cloud agreement should begin in the first half of 2027, with svtrivo citing Q1 2027. Contract value and deployment scope were not disclosed. Call summaries said the overall opportunity pipeline expanded more than tenfold in the quarter.
Does Oracle's Cerebras-powered inference service launch in summer 2027?
Cerebras said Oracle is buying hardware for its own data centers, with shipments beginning in Q4 2026 and continuing through at least the first half of 2027, ahead of an inference service expected to launch in summer 2027. Contract value, end-customer pricing and post-launch utilization have not been disclosed.
Can Cerebras bring OpenAI capacity deliveries back on schedule by H2 2027?
The 750 MW take-or-pay OpenAI agreement is behind schedule, which management attributes to land, power and building readiness. Delay credits have already been deducted from RPO, and late delivery can trigger service credits or, if severe, cancellation of capacity tranches. Cerebras expects to return to schedule in the second half of 2027, with added capacity starting to come online in December.
CS-4 ramp and CS-5 early-summer 2027 target are next milestones
Cerebras said CS-4, promising up to 30x faster inference, is approaching full production, while CS-5 is targeted for early summer 2027 at up to 10,000 tokens per second per user. Management said it has sufficient TSMC wafer commitments for 2027; whether the roadmap ships on schedule remains open.
Last 24 hours
Nov 17, 01:01 → Nov 18, 01:01
Cerebras CEO confirms OpenAI monetizes capacity at about $200 per watt
At TomorrowX, thematictrader relayed that CEO Andrew Feldman confirmed with Gavin Baker that OpenAI is selling Cerebras inference at $200/Watt, while svtrivo shared a quoted post describing roughly $200 million per megawatt and called the direction positive. Neither post disclosed contract terms or revenue contribution.
Cerebras says delays are land, power and shell as capacity flows
At TomorrowX, thematictrader reported Feldman describing forecasts far above what vendors had planned and delays to promised data-center buildouts. Capacity is only now starting to be delivered and should flow through Q1 to Q4 of next year, while Gavin said the 40% sequential uptick is just the beginning and expects similar or better in the first quarter. New contracts are expected to help from Q2.
Cerebras denies any chill with OpenAI and expects revenue to rise
frenchie_ recapped the first day of TomorrowX, listing Cerebras' denial of any cold relationship with OpenAI and its expectation of higher revenue from Q3 to Q4, then again in 2027. The post placed this among the day's concrete signals, separate from broader conference commentary on other companies.
Feldman says orbital compute at scale is at least a decade away
pequityresearch said a private Q&A with Andrew Feldman at TomorrowX covered orbital compute, token pricing, IP advantage and AI in chip design. Feldman had strong words on orbital compute and does not believe it will happen for at least a decade, at least at the scale being discussed. The interviewer plans to publish the session for paid subscribers this week.
Weekly brief
Nov 11, 01:01 → Nov 18, 01:01
Cerebras Q3 core revenue tops consensus as cloud services jump 272%Major development
Core revenue reached $230.87 million, up 70% year over year and above the $215.82 million consensus, while GAAP revenue was $210.8 million and GAAP EPS was -$0.67. Cloud and other services rose 272% to $144.91 million; hardware revenue fell 29% to $85.96 million but rose 5% sequentially. Remaining performance obligations were $26.1 billion, with $7.8 billion of liquidity and an unused $850 million revolver.
Cerebras lifts FY26 core revenue outlook, guides Q4 far above estimatesMajor development
Management raised 2026 core revenue guidance to $940–960 million from $880–890 million and guided Q4 core revenue to $310–330 million against a $271.2 million estimate. Q4 core gross margin is guided to 47–49% and core operating margin to -1% to +1%, with cloud revenue roughly flat sequentially because added capacity arrives too late to contribute. Management expects total revenue to roughly triple in 2027.
Cerebras adds Microsoft as hyperscaler with revenue starting in 2027Major development
Cerebras announced a cloud agreement with Microsoft during its Q3 earnings call, saying the deal was signed in October and that revenue should begin in the first half of 2027; svtrivo cited Q1 2027. The win brings Cerebras to three hyperscalers landed in six months, alongside AWS and Oracle. Contract value was not disclosed, and call summaries said the opportunity pipeline expanded more than tenfold in the quarter.
Cerebras says OpenAI capacity deliveries are behind scheduleMajor development
Feldman acknowledged Cerebras is behind schedule on initial capacity deliveries under its OpenAI agreement, blaming land, power and completed data-center buildings rather than chips or materials. The take-or-pay agreement covers 750 MW, and potential delay credits were already deducted from the $26.1 billion RPO as of Sept. 30. Management expects to return to schedule in the second half of 2027.
Oracle to buy Cerebras hardware for its own data centers
Management said Oracle is purchasing Cerebras hardware for deployment in its own US data centers, with shipments expected to begin in Q4 and continue through at least the first half of 2027. An Oracle-powered inference service is expected to launch in summer 2027. The partnership had been announced earlier; this week's disclosure added procurement structure and deployment timing.
Cerebras contracts about 700 MW, mostly outside US in 2027
Cerebras has roughly 700 gross megawatts of data-center capacity under contract for delivery by the end of 2027, and management said more than 75% of the capacity coming online that year will be outside the United States. The company also cited additional capacity agreements in Norway and Finland, and additional capacity is scheduled to begin coming online in December.
Cerebras scales production with Flex and Sanmina toward 10x capacity
Cerebras said it is scaling manufacturing with Flex and Sanmina and expects to exit 2026 with more than ten times the manufacturing capacity it had at the start of the year. Management also said it has sufficient TSMC wafer commitments for 2027. Q4 growth is expected to be hardware-led, with new hardware orders driving the quarter.
CS-4 nears full production as Cerebras targets CS-5 for summer 2027
Cerebras says CS-4 promises up to 30x faster inference and is approaching full production, while CS-5 is targeted for early summer 2027 at up to 10,000 tokens per second per user. svtrivo noted that the year-by-year doubling plan is backed by the modular Nexus rack, which can take the next two generations without a full redesign.
Cerebras runs production inference with AMD GPUs handling prefill
Cerebras introduced production-scale heterogeneous disaggregated inference, pairing AMD GPUs for prefill with Cerebras systems for decode, and said the approach can raise tokens delivered per megawatt by up to fivefold. TheValueist's read-through argued this validates a complementary role for AMD while challenging Nvidia's capture of premium inference economics, though Cerebras' revenue base is far smaller.
Bilt selects Cerebras to power Neighborhood OS and Concierge
Bilt chose Cerebras to power its AI-driven Neighborhood OS and Bilt Concierge, which serve more than 8.2 million members and 50,000 merchants; financial terms were not disclosed. svtrivo, a Bilt user and $CBRS shareholder, said the deal explained Cerebras' restaurant-reservation speed demos, since inference speed is what makes agentic booking usable at that scale.
Law firms investigate Cerebras investor claims after results
Pomerantz LLP said it is investigating claims on behalf of Cerebras investors, while Bronstein, Gewirtz & Grossman said it is investigating potential claims and encouraging shareholders to connect. The notices appeared in mid-November and invited affected investors to contact the firms.
Cerebras shares jump after results, premarket gains reported near 12–15%
Post-earnings coverage differed on the size of the move: zerohedge reported a 15% premarket gain after guidance came in well above estimates and new Oracle and Microsoft deals were unveiled, Newsquawk cited 12%, donaldjdean noted an 11.5% premarket gain at $109.00, and unusual_whales reported an 18.4% rise after results.
Sentiment
Chart data and price change
Window: 2026-10-20 – 2026-11-18 UTC
Price: 118.24 → 104.18 (-11.89%), from 2026-10-20 to 2026-11-17.
Unique authors discussing this ticker as a share of all author–ticker pairs across X, YouTube, Reddit and Substack. Each author counts once per ticker in the selected period. Mindshare as of 2026-11-18T03:11:17.717783+00:00; refreshed daily.
| Date | Price | Sentiment | Mindshare |
|---|---|---|---|
| 118.24 | -0.150 | 0.102% | |
| 117.98 | +0.150 | 0.109% | |
| 108.94 | +0.350 | 0.089% | |
| 110.82 | — | 0.000% | |
| — | +0.325 | 0.279% | |
| — | +0.700 | 0.128% | |
| 108.37 | -0.083 | 0.160% | |
| 119.31 | +0.174 | 0.421% | |
| 112.56 | -0.199 | 0.736% | |
| 115.16 | -0.100 | 0.071% | |
| 109.18 | +0.600 | 0.039% | |
| — | +0.000 | 0.207% | |
| — | — | 0.000% | |
| 96.36 | -0.270 | 0.292% | |
| 97.82 | +0.030 | 0.240% | |
| 91.85 | -0.775 | 0.102% | |
| 81.71 | +0.150 | 0.211% | |
| 89.68 | -0.058 | 0.210% | |
| — | — | 0.000% | |
| — | +0.100 | 0.173% | |
| 96.70 | -0.050 | 0.079% | |
| 93.66 | +0.000 | 0.201% | |
| 101.08 | +0.261 | 0.122% | |
| 97.75 | +0.384 | 0.826% | |
| 108.72 | +0.262 | 0.662% | |
| — | +0.125 | 0.383% | |
| — | +0.350 | 0.282% | |
| 100.99 | +0.367 | 0.137% | |
| 104.18 | +0.053 | 0.346% | |
| — | +0.150 | 0.289% |
Author price targets
- Median target
- USD 150.00
Where authors set their targets
- Lowest
- —
- Highest
- —
No comparable targets in this selection. Try another filter.
Authors and targets
Call performance
All sources
All author categories
| # | Author | Direction | Source | Published | Entry / latest | Return | Original idea |
|---|---|---|---|---|---|---|---|
| 01 | svtrivoSemiconductor Industry Commentator | Long | 95.20104.18 | +9.4% | |||
Original ideaCerebras shareholder sees Bilt inference use case | |||||||
| 02 | Mike VolpiPartner, Index Ventures; former Cisco SVP | Long | youtube | 101.56104.18 | +2.6% | ||
Original ideaCerebras is an existing portfolio holding. | |||||||
No data yet
Price change since each call, adjusted for long/short direction. Results calculated:
Author views
- Long
- 100.0%Authors: 3
- Short
- 0.0%Authors: 0
- Avoid
- 0.0%Authors: 0
One latest Long / Short / Avoid view per author.
- Long
- 83.3%Authors: 5
- Short
- 16.7%Authors: 1
- Avoid
- 0.0%Authors: 0
One latest Long / Short / Avoid view per author.
The bull thesis
Gavin Baker: Cerebras speed is decisive for latency-sensitive and cyber workloads
Gavin Baker said Atreides is an investor in Cerebras, which has the fastest available token generation through wafer-scale inference. He argued speed is decisive in cyber defense and latency-sensitive AI workloads, and pointed to Cerebras' partnership with CrowdStrike to bring fast inference to security.
Bulls point to 27x revenue RPO and near-breakeven Q4 margins
ren_stocks called it a killer earnings, citing core revenue of $230.9 million versus roughly $215 million expected, cloud and services up 265% to $141.9 million, RPO of $26.1 billion — about 27x this year's guided revenue — and Q4 core gross margin guided to 47–49% with operating margin near breakeven. He saw a possible bottom after a 70% post-IPO decline.
svtrivo stays bullish on Cerebras technology despite worst-performing holding
svtrivo said $CBRS remains a meaningful position; he participated in the IPO and it is currently the worst performer among his holdings, which he said does hurt. He said he still views the company's technology and long-term vision with hope and remains bullish, framing it as a personal view and position, not investment advice.
Fast inference market seen growing to $550B with CBRS the pure play
stocksavvyshay cited Morgan Stanley's expectation that inference spending grows about 8x by 2030 while fast inference grows from $4 billion in 2026 to $550 billion by 2030 as AI agents make speed critical, leaving $CBRS as the only public pure play on that tier.
Non-HBM wafer-scale design framed as a supply-chain advantage
Damnang argued the wafer-scale, non-HBM approach could make sense given current supply constraints, pointing to Cerebras' claim that memory shortages and price increases persist through 2028, raising competitors' costs. stocksavvyshay listed CBRS among SRAM inference beneficiaries. Damnang added Cerebras is moving toward 3D stacked DRAM and can complement AMD by handling decode while GPUs handle prefill.
Bilt deployment cited as a real-world agentic use case for speed
svtrivo said the Bilt partnership explained Cerebras' restaurant-reservation speed demos: Bilt Concierge and Neighborhood OS run AI agents for real-time bookings, travel, rewards and local commerce across more than 8 million members and 50,000 merchants, where inference speed is not a nice-to-have. He is both a heavy Bilt user and a $CBRS shareholder.
Damnang calls the level fairly attractive while flagging high beta
Damnang said he would not call Cerebras cheap at its current price but sees it as a fairly attractive level, expecting the company to carve out a place as the compute market grows. He warned it is a high-beta stock with significant risk and plenty left to prove, and could be among the hardest hit in a bearish market.
Reporters say Microsoft and Oracle wins show Cerebras gaining on Nvidia
Bailey Lipschultz wrote that the sales forecast beat and customer deals with Microsoft and Oracle signaled Cerebras' push to challenge Nvidia in AI chips is paying off, while noting 14% short interest reflects negative sentiment.
Mike Volpi says Cerebras has done very well at speed
Mike Volpi disclosed that his firm is an investor in Cerebras and said the company has done very well at speed. He noted its SRAM-based architecture requires large models to be fitted carefully and suffers weaker memory bandwidth once work moves off-chip, describing engineering trade-offs rather than a broken thesis.
firstadopter no longer negative on Cerebras at $83
firstadopter said he had been publicly bearish on Cerebras and its valuation since the first day of the IPO, but in this instance believes Cerebras is right and Culper Research is wrong. He is no longer negative on the stock at $83, citing the point that less than half of one percent of deployed CS-3 wafers have been decommissioned.
Cerebras framed as a beneficiary of Nvidia's turmoil
deepvaluebagger argued $CBRS is a big benefactor of Nvidia's chaos, not because of its own earnings. Tatyana said Cerebras is well positioned to shine as a capacity-constrained Nvidia competitor that sells both AI-specific processors and cloud access, with Nvidia's leadership turmoil creating an opening, though she flagged schedule and OpenAI contract questions.
Cerebras ranked tier-5 in managed inference but with a permanent latency edge
jiahanjimliu placed Cerebras in Tier 5 of managed-inference competitiveness alongside TogetherAI and Modal, but said it sits there because it has a permanent advantage in vertical integration with low-latency wafer-scale chips, unlike lower tiers that pay revenue share to Chinese labs. No position was disclosed.
Bear thesis & objections
thematictrader stayed away from Cerebras over yield issues and competition
thematictrader said he never had a strong long-term view on Cerebras and was not interested anywhere near the IPO given glaring yield issues and competition. He called the recent earnings report nice and noted it clearly caught many investors offsides.
Damnang flags wafer-scale limits behind the no-HBM marketing
Damnang argued that Cerebras' claim of not needing HBM is marketing wordplay, noting that running Kimi K3 would require 59 wafers by the company's own calculations. He said Cerebras will eventually have to compete for memory supply as it moves toward 3D stacked DRAM, and that proving efficiency as models and systems grow remains the biggest question.
Damnang says Cerebras is not cheap and carries high beta
Kramer cites Cerebras' own lock-up unwind as a cautionary template
michaelmottcm cited Cerebras as the template for lock-up risk: a 30 million share deal priced at $185 in May that found a floor near the IPO price, then sold off at the end of September as lock-up shares unwound steadily. He used it to frame the staggered lock-up schedule for Anthropic's offering.
jonahlupton says hot IPOs like CBRS spiked then fell 50–80%
jonahlupton argued that direct listings like Palantir may price lower and benefit public shareholders more, citing a broken IPO model. He noted many recent IPOs — including $FIG, $CRCL, $CRWV and $CBRS — were up 50–300% in the first week or two, then fell 50–80% in the following months.
TheValueist says land, power and shells, not chips, now gate revenue
TheValueist's read-through of the Q3 call stressed that data-center readiness, not accelerator availability, increasingly determines when demand becomes revenue, with delays attributed exclusively to land, power and shells and cloud revenue flat in Q4 despite demand exceeding supply. He also said large cloud backlogs offer no unconditional protection against deployment delays.
Late delivery can trigger service credits or cancel capacity tranches
Late delivery carries contract consequences beyond timing: capacity delays can trigger service credits or, if severe, cancellation of capacity tranches, and Cerebras' filing said delays shorten the OpenAI contract term on a day-for-day basis. Potential delay credits were already deducted from the $26.1 billion RPO.
Nov 18
Nov 17
Nov 17

