
Cerebras Systems (NASDAQ: CBRS): The Pullback Is Noise. The Inference Demand Is the Signal.
Coach Kyle | October 12, 2026

The Bottom Line
Cerebras is down more than 60% from its IPO-day high. It’s trading near $150, below its $185 IPO price, after a rough run of headlines: an OpenAI/Nvidia report, insider selling and a staggered unlock of insider shares. None of that changes the core story. The company just grew core revenue 103%, has $25.4 billion of contracted backlog, and Wall Street expects revenue to more than triple in 2027 as EPS turns solidly positive. ARK is buying the dip, and we like it here too.
G.L.A. View: BULLISH for long-term investors. Our call rests on next year’s earnings and revenue estimates and one simple fact: demand for fast inference still looks close to infinite.(Not individualized investment advice)
Snapshot (as of Oct. 12, 2026, intraday)

Metric | Value |
|---|---|
Share price | $150.56 (-8.29% today) |
Market cap | $35.8B |
52-week range | $148.50 – $386.34 |
IPO price (May 14, 2026) | $185.00 |
Cash and investments (Q2 2026 balance sheet) | $7.92B in cash and investments; $918.2M OpenAI working-capital loan outstanding |
Analyst consensus | Strong Buy — 8 of 8 bullish |
Average price target | $291 (~93% upside) |
Next earnings | Q3 2026 — Nov. 11, after the close |

What Cerebras Does
Cerebras builds the Wafer-Scale Engine, a single chip the size of a full silicon wafer. It puts compute and memory on one piece of silicon instead of linkingdozens of GPUs together. That design removes data-traffic bottlenecks and makes Cerebras the speed leader in AI inference, the moment a model actually answers you. Cerebras sells that speed two ways: as hardware (the CS-3 and the new CS-4) and as a fast-inference cloud service. The cloud service is now the faster-growing half of the business.
The customer list is getting deep. Per the Q2 call, it includes OpenAI, AWS, CrowdStrike, Figma, Cognition, Lovable, Block, AlphaSense and GSK.

Growth & Sales: Verified Hypergrowth
Annual revenue
Year | Revenue | YoY Growth |
|---|---|---|
FY2022 | $24.6M | — |
FY2023 | $78.7M | +220% |
FY2024 | $290.3M | +269% |
FY2025 | $510.0M | +76% |
FY2026E (consensus) | $887.1M | +74% |
FY2027E (consensus) | $2.95B | +233% |

Revenue grew 20x from 2022 to 2025, and Cerebras booked a profit of $237.8 million in FY2025.
Q2 2026: the most recent quarter
Core revenue: $209.9M, up 103% year over year, a record.
Cloud and services revenue: $127.7M, up 287%. That’s nearly 4x, driven by demand for Cerebras Fast Inference.
Core hardware revenue: $82.1M, up 17%.
GAAP revenue: $180.1M, up 74.3%.
Core operating margin: -16%, versus -42% a year ago. That’s a 2,600-basis-point improvement in one year.
Remaining performance obligations (contracted backlog): $25.4 billion.
Cash, investments and restricted cash: more than $8.6 billion.
For context, $25.4 billion is about 29x this year’s expected revenue. Management says it doesn’t yet include any business from AWS or other hyperscalers.

Earnings: Reading the Scorecard Correctly
Quarter | Revenue (actual vs. est.) | GAAP EPS (actual vs. est.) | 1-day stock move |
|---|---|---|---|
Q1 2026 (reported Jun. 23) | $193.4M vs. $181.2M — beat | -$0.04 vs. -$0.14 — beat | -19.6% |
Q2 2026 (reported Aug. 12) | $180.1M vs. $190.6M — miss | -$2.98 vs. -$0.21 — miss | -11.8% |
Q3 2026 (Nov. 11) | Est. $214.7M | Est. -$0.13 to -$0.15 | — |

What actually happened in Q2: The GAAP loss of $450.5M (-$2.98 a share) came mostly from $377.0M of stock-based compensation tied to the IPO, a non-cash charge concentrated around the listing. On its core metrics, Cerebras beat its own guidance for revenue, gross margin and operating margin, and it raised full-year guidance.
Gross margin: Core gross margin dipped to 40.6% in Q2 from 46.5% in Q1. The reason: Cerebras is renting back its own systems from cloud customers so it can serve overflow inference demand sooner. The margin hit comes from too much demand, which is the good kind of problem. Management expects Q3 to be the low point (38%–40%). Margins should step up in Q4 as Cerebras-owned data centers come online, then trend toward a 60%+ target.

Management guidance
Period | Guidance |
|---|---|
Q3 2026 core revenue | $214M – $216M |
FY2026 core revenue | $880M – $890M (raised) |
FY2026 core gross margin | 41% – 43% (raised) |
FY2026 core operating margin | -19% to -17% (raised) |
2027 | Core revenue to grow more than 3x |

Analyst Expectations & Estimates
Consensus estimates
FY2026E | FY2027E | |
|---|---|---|
Revenue | $887.1M | $2.95B |
EBITDA | $37.5M | $1.20B |
Gross profit | $372.2M | $1.43B |
Diluted EPS | -$0.50 | +$1.30 |
The quarterly ramp analysts are modeling
Quarter | Revenue Est. | EPS Est. |
|---|---|---|
Q3 2026 | $214.7M | -$0.15 |
Q4 2026 | $274.4M | -$0.25 |
Q1 2027 | $402.8M | -$0.08 |
Q2 2027 | $567.0M | +$0.11 |
Q3 2027 | $822.0M | +$0.47 |
Q4 2027 | $1.12B | +$0.77 |
Q1 2028 | $1.36B | +$0.93 |
Q2 2028 | $1.68B | +$1.31 |
Source: Perplexity Finance — CBRS consensus estimates
Analysts see quarterly revenue rising from about $215 million today to about $1.68 billion by mid-2028, nearly 8x in under two years. They have EPS turning positive in Q2 2027 and free cash flow turning positive in early 2028.

Price targets: every analyst is bullish
Firm | Rating | Price Target | Date |
|---|---|---|---|
UBS | Buy | $330 | Aug. 13 |
Citigroup | Buy | $320 | Aug. 13 |
Mizuho | Outperform | $300 | Aug. 13 |
Needham | Buy | $300 | Aug. 19 |
Rosenblatt | Buy | $300 | Aug. 19 |
Wedbush | Outperform | $290 | Aug. 13 |
Morgan Stanley | Overweight | $279 | Aug. 13 |
Freedom Capital | Upgraded to Buy | $209 | Oct. 2 |
Consensus: Strong Buy, 8 of 8 bullish, average target $291, median $300. Source: Perplexity Finance — CBRS analyst research
Freedom Capital upgraded the stock in the middle of the selloff, calling the OpenAI fears overblown and pointing to strong inference demand. Barclays’ Tom O’Malley called the decline “overdone” and said Cerebras “remains the preferred provider limited by supply”.finance.yahoo+1
Valuation on next year’s numbers
At a $35.8B market cap, CBRS trades at about 12x 2027 consensus revenue. That’s for a company expected to grow revenue 233% next year with $25 billion already under contract. It’s also a far cry from the 54x–61x trailing-sales multiples quoted in the bearish headlines.

The Restrictions: What’s Actually Weighing on the Stock
1. The IPO lockup: a staggered release, not a cliff
When Cerebras went public, insiders agreed to restrictions on when they could sell. Instead of one big unlock at 180 days, Cerebras is releasing those shares in 11 separate waves.
Date | Shares eligible for sale |
|---|---|
After Q1 and Q2 earnings | ~27.7M and ~36.4M |
Aug. 19 – Sept. 16 (every two weeks) | ~14.6M per wave |
Sept. 30 | ~19.4M (about 8% of shares outstanding) |
Oct. 14 (this Wednesday) | ~19.4M |
Oct. 28 | ~19.4M |
Final release | All remaining — 2nd trading day after Q3 earnings or Nov. 9, whichever comes first |

The pattern is clear. The stock fell 8.87% on the Sept. 30 unlock day alone. Today’s 8% slide to a new post-IPO low comes two trading days before the next wave.
Why this is good news for long-term holders: The overhang has an end date. By about mid-November, every insider share will be free to trade. Selling pressure from unlocks is temporary and mechanical, and it says nothing about the business.
2. Insider selling under pre-set plans
CEO Andrew Feldman and CTO Sean Lie sold more than $240 million of shares between Aug. 20 and Sept. 25, under trading plans adopted shortly after the IPO (CNBC). COO Dhiraj Mallick sold 396,000 shares for about $80.8 million under a Rule 10b5-1 plan adopted June 30 (Gate News). These are pre-scheduled sales by founders who became billionaires on IPO day. They read as diversification, not a verdict on the company.
3. Export controls
Cerebras’ 10-Q notes that its products are subject to U.S. export controls and generally need licenses to ship to certain countries. This is a long-standing rule for every AI chipmaker, not a new restriction. Cerebras’ growth engine (OpenAI, AWS, U.S. enterprises and new data centers in places like Finland) sits largely in approved markets.investors.cerebras
How the Negative Headlines Hit the Stock
Verified timeline
Date | Headline | Stock reaction |
|---|---|---|
Aug. 12 | Q2 GAAP miss on IPO stock comp | -11.8% next day |
Aug. 18 | Supernova event (CS-4 launch) | -12.7% on the day |
Sept. 29–30 | SemiAnalysis says OpenAI’s GPT-6.1 Sol “Ultrafast” tier runs on Nvidia GPUs + 19.4M-share unlock | -8.9% on Sept. 30 |
Week ending Oct. 2 | Insider sales + unlock + Nvidia report | -19.5% to $166.43, lowest close since IPO |
Oct. 2 (evening) | Sam Altman: “Cerebras is a close partner” | +~3% after hours |
Oct. 5 | Rebound on Altman comments | +9% to $181.55 |
Oct. 12 | Fresh low ahead of Oct. 14 unlock | -8.3% to ~$150 |
Sources: Yahoo Finance, TheStreet, CNBC, Perplexity Financecnbc+1
Our interpretation
The OpenAI scare was about one speed tier of one model. Neither OpenAI nor Cerebras confirmed which hardware runs it. Within days, OpenAI’s CEO stepped in personally to say “we have a deep engagement pushing on the frontiers of speed”. The 750-megawatt OpenAI agreement, worth more than $10 billion, runs through 2028 and hasn’t changed. Barclays’ read is that Cerebras is limited by supply, not demand, which matches what management said on the call.gate+2
This is a classic young-IPO pattern: a stock that reacts sharply to headlines, plus a steady stream of new shares hitting a small float. Headlines move the stock day to day. Earnings and backlog move it year to year.
ARK Is Still Buying the Pullback
Cathie Wood’s ARK has bought CBRS on nearly every major dip since the IPO, through the ARK Innovation ETF (ARKK) and the ARK Next Generation Internet ETF (ARKW).

Date | Shares | Value | Context |
|---|---|---|---|
Mid-May | 149,176 | ~$46.4M | Right after the IPO |
Jun. 23–24 | 124,949 | ~$28.3M | Around the Q1 report and 19.5% drop |
Jun. 25 | 111,989 | ~$20.4M | Kept adding |
Aug. 13 | 106,941 | ~$28.0M | Day after the Q2 selloff — largest single buy |
Aug. 18 | 35,000 | ~$8.8M | Supernova-day drop |
Aug. 20 | — | ~$14.5M | Continued accumulation |
Aug. 25 | 93,290 | ~$17.3M | Funded partly by AMD sales |
Sept. 8 | 32,561 | ~$6.5M | Swapped from Palantir |
Oct. 2 | 117,758 | ~$20.0M | Bought at the post-IPO low close of $166.43 |
Sources: IndexBox, TradingView/GuruFocus, Investing.com, Yahoo Finance, GuruFocus, Yahoo Finance, Investing.com, TipRanks, Investing.comfinance.yahoo+8
That adds up to about $190 million of reported purchases in under five months, and the latest buy came on the exact day the stock hit its post-IPO low. ARK has repeatedly sold AMD and Palantir to fund these buys, a clear signal of where it sees the next leg of AI-chip upside. CBRS made up about 2.3% of ARKK as of Oct. 9.blog.arkinvesttrades
Wood’s thesis is simple: Cerebras doesn’t need to dethrone Nvidia. It just needs to own the speed slice of a massive market. ARK has also argued that Cerebras’ on-chip memory design could make expensive high-bandwidth memory less .

Catalysts Ahead
Q3 earnings (Nov. 11): Guidance says gross margin should bottom here, setting up a “worst is behind us” quarter.
Final lockup release (around Nov. 9–13): The share-supply overhang ends.
Q4 2026: Launch of disaggregated inference with AMD Helios racks. More Cerebras-owned data centers come online, which should lift margins.
Q1 2027: Cerebras becomes available on Amazon Bedrock, giving it global enterprise reach.
Mid-2027: First revenue from additional hyperscalers.
Capacity: More than 600 MW of data center capacity secured, a pipeline “measured in gigawatts,” manufacturing capacity up more than 10x in 2026, and TSMC wafer supply in hand.
New capacity deals: A 165 MW data center with Compute Nordic in Finland.investors.cerebras

G.L.A. View: Bullish
This section is our opinion.
We’re bullish on CBRS for long-term investors, and our conviction rests on next year’s numbers:
Revenue is expected to more than triple. Consensus calls for $2.95 billion in 2027, up 233%, and management has guided to more than 3x growth as well.
Earnings turn positive. Consensus EPS goes from -$0.50 this year to +$1.30 in 2027, with EBITDA above $1.2 billion. The heavy IPO stock-comp charges fade, and the operating leverage shows up.
The backlog is real. $25.4 billion of contracted revenue gives us multi-year visibility, and that’s before AWS or any other hyperscaler is counted.
Inference demand still looks close to infinite. Every chatbot answer, coding assist and AI-agent task runs on inference, around the clock. Feldman calls it “unbound demand for fast inference,” and the evidence backs him up. Cloud revenue nearly quadrupled, Cerebras is renting back its own systems just to keep up, and it’s limited by supply, not demand. When customers pay a premium for faster responses, speed becomes a moat.
The price is right. CBRS trades at about 12x next year’s sales, every analyst is bullish, and the average target is near $291. The pullback gives long-term investors an entry point well below where ARK and other big buyers have been adding.
How we’d approach it: We expect volatility to stay high through the lockup schedule and Q3 earnings. For our long-term approach, that’s an opportunity, while the risks below still apply. Our approach is dollar-cost averaging through the Oct. 14 and Oct. 28 unlocks and into the Nov. 11 report, while the share overhang clears.

Risks to Watch
We’re bullish, but we’re honest about the risks:
Customer concentration: OpenAI is the biggest piece of the backlog. A real shift of its workloads to Nvidia would matter.
Margins: Gross margin needs to climb from about 40% toward the 60%+ target, as promised.
Execution: Tripling revenue requires data centers, manufacturing and supply to scale on time.
Competition: Nvidia’s ecosystem, balance sheet and supply remain formidable.
Near-term supply: Two more unlocks of about 19.4M shares each, plus a final release, before mid-November.
Disclaimer: This report is for educational and informational purposes only and is not personalized investment advice. Data is as of October 12, 2026, and subject to change. Do your own research and consider your own risk tolerance before investing.

G.L.A. INVE$T RE$EARCH | Coach Kyle
Balance-sheet clarification: cash and investments exclude restricted cash. The June 30 Form 10-Q reports the OpenAI working-capital loan and distinguishes quarterly from six-month stock compensation.
Original edition: Kyle’s Beehiiv newsletter. Author opinions, scenarios and dated estimates are retained; limited editorial clarifications do not verify every source claim.