Gross profit margin plummeted by 10 percentage points. How much did CBRS's business model transformation cost?
Author: David, Tide Research
Original title: First financial report from the CBRS listing: revenue doubled but gross margin guidelines plummeted, OpenAI's big order fulfillment path was too long
Tide's reading: Cerebras (CBRS) handed over its first quarterly report after the IPO. Q1 core revenue was US$191 million, up 92% year over year, exceeding market expectations. However, the Q2 core gross margin guide plummeted from 46.5% to 36%-38%, and the stock price fell more than 10% after the market. This company, which uses an entire wafer as a chip and bets on the AI inference circuit, has an OpenAI contract of over $20 billion and the AWS cooperation framework, leading to annual revenue of 855-865 million US dollars. The growth data is hard enough, and the valuation controversy is big enough.
Core concerns
Revenue exceeded expectations, and guidance even exceeded expectations. Q1 core revenue of $191.3 million (+92% YoY) was higher than the agreed estimate of approximately $181 million. Core revenue guidance for the full year was US$855-865 million (+69% YoY), higher than market expectations of US$828 million. Under GAAP, cloud and services revenue was US$82.8 million, up 178% year over year, making it the fastest growing sector.
The sharp drop in gross margin guidance was the biggest negative of the season. The Q1 core gross profit margin was 47%, up nearly 5 percentage points year over year. However, the Q2 guidance fell to 36%-38%, down about 10 percentage points from Q1; the full-year guidance was 38%-41%. Management attributed the reason to insufficient data center capacity: the company was temporarily leasing back systems from existing customers who had sold hardware to deploy capacity, and short-term costs worsened as a result. Shares fell more than 10% after the market.
There is a direction for improvement in customer concentration, but it is far from being solved. 86% of revenue for fiscal year 2025 came from two UAE related entities (MBZUAI 62% and G42 24%). OpenAI began contributing revenue in February 2026, and the AWS partnership is expected to be reflected financially only in 2027. True income diversification will not be verified until 2027.
The valuation price is until 2028. Based on about $200 after the market, CBRS corresponds to about 90 times the revenue of the past 12 months; even with a median value of US$860 million using the full-year guidance, the long-term P/S is still more than 50 times higher. The 10 coverage analysts had a median target price of $300 (range of $250-340), implied that OpenAI's contract of over $20 billion and the AWS deployment were fulfilled on time and in volume.
Short-term catalytic and repressive factors coexist. Catalysts: Accelerated deployment of OpenAI 750MW computing power, implementation of the AWS inference solution, and launch of new data center production capacity in the second half of the year. Constraining factors: The lockdown period included unconventional early lifting clauses (the market capitalization can be triggered when the market value exceeds 40 billion US dollars; the current market value has reached near this threshold), the gross margin recovery path is unclear, OpenAI itself is not profitable, and it has promised to reduce part of its computing power.
Financial Report Reveals Business Model Transformation: From Selling Chips to Selling Computing Power
Q1 What is most easily overlooked in financial reports is the change in revenue structure.
Under the core caliber, hardware revenue was US$116.6 million, accounting for 58% of total revenue; cloud and service revenue was US$79.8 million, accounting for 42%. In the same period a year ago, the ratio was roughly 70:30. Cloud service revenue increased 167% year over year, nearly three times that of hardware.
Management made this trend more clear during the conference call:
Hardware revenue will gradually decline in the next few quarters, as the company will deploy more hardware production capacity to its own cloud to fulfill inference computing power contracts with OpenAI and AWS rather than selling it directly to customers. Cerebras is changing from a “company that sells chips” to a “company that sells computing power.”
This transformation also directly explains why Q2 gross margin plummeted.
During the call, an analyst inquired about the details of production capacity deployment, and management revealed:
The company's current bottleneck is not TSMC's chip supply, but the physical space in the data center. In order to deliver computing power to OpenAI as soon as possible, Cerebras is “temporarily leasing back” the hardware system already sold from G42 (previously its largest customer and minority shareholder).
If you rent a third-party facility to deploy your own system, the cost structure will deteriorate in the short term. This is the main reason for the decline in gross margin from the 47% guideline to 36%-38%. The schedule given by management is for the new data center to be launched in the second half of the year, when the cost pressure will ease.
The financial structure of the OpenAI contract is also worth breaking up. On the surface, it was a multi-year computing power purchase of over $20 billion, but there are three layers of relationships underneath: OpenAI provided Cerebras with a $1 billion working capital loan (reflected in the Q1 balance sheet as $621 million in current loans and $362 million in non-current loans), and also obtained Cerebras's share warrants.
In other words, OpenAI simultaneously plays the triple role of Cerebras's largest customer, creditor, and potential shareholder. The risk in S-1 suggests that if Cerebras fails to deliver production capacity as agreed, OpenAI has the right to terminate the contract and trigger loan repayment.
The AWS cooperation framework uses a “decoupled reasoning” architecture: AWS's Trainium 3 chip is responsible for processing prompt input (prefill stage), and Cerebras' CS-3 system is specifically responsible for generating output at high speed (decode phase). This design allows Cerebras not to take on a complete inference link, but only to do the part where it has the greatest speed advantage. However, during the Q&A session, management declined to disclose the exact scale of the AWS partnership and stated that revenue contributions would not be reflected financially until 2027.
The common characteristics of the two large orders are: the contract is huge, but the path to fulfillment is very long, and it is highly dependent on the progress of Cerebras data center construction.
The annual revenue guidance of $855-865 million means that the next three quarters will require an average of about $220 million, and the growth rate will accelerate quarter by quarter. Management said “the year-on-year growth rate will increase in every quarter of 2026, and more revenue will be concentrated in the second half of the year.”
Bullish logic: nine investment banks are calling to buy at the same time, what are they buying
On June 8, the day the IPO silence period ended, nine underwriters simultaneously initiated coverage, and all gave ratings for buying or increasing their holdings. CBRS rose 18.3% in a single day on the same day. This kind of “open door” group opinion is not uncommon in US IPOs (underwriters naturally have interests bound), but their betting logic points to the same core proposition.
Proposition 1: The battleground of AI computing power is shifting from training to reasoning, and the rules of competition in inference scenarios are different from training.
Morgan Stanley analyst Joseph Moore gave an increase in holdings rating and a target price of $250 in the first coverage report on June 8. His core argument is: the training scenario competes for total computing power throughput, where the Nvidia GPU cluster absolutely dominates; the inference scenario competes for the speed and latency of a single response, because the model has to process millions of user requests per second, and speed directly affects service costs and user experience. Because the on-chip SRAM capacity far exceeds that of conventional GPUs, Cerebras' wafer-level chips do not require frequent transportation of data to external storage, and have structural advantages in inference latency. Moore stated that Cerebras is “the only company to commercialize wafer-level processors,” giving Nvidia a first-mover advantage.
Citibank analyst Atif Malik gave the highest price target of $340 in coverage. Mizuho added a technical detail to the June 8 research report: The WSE-3 chip has 44GB SRAM, which is several times that of Google's latest TPU and Groq LPU. This hardware gap cannot be bridged in the short term through architectural optimization.
Proposition 2: Two big deals moved Cerebras from a “tech story” to a “revenue story.”
OpenAI contracts exceed $20 billion, covering 750MW of inference computing power, and multi-year delivery. On a five-year amortization basis, this contract alone contributed approximately $4 billion in revenue per year, which is nearly five times the median value of the 2026 full-year revenue guidance. Although management of the AWS partnership declined to disclose the exact amount, the framework confirmed that Cerebras' reasoning capabilities will be made available to global enterprise customers through Amazon Bedrock.
The Q1 financial data provided early verification. OpenAI began deploying the Cerebras system in February, and cloud service revenue jumped from less than $30 million in the same period last year to nearly $80 million in a quarter. Management stated that “the year-on-year growth rate will increase in every quarter of 2026, and more revenue will be concentrated in the second half of the year,” and the full year's guidance of US$855-865 million is higher than the agreed estimate of US$828 million.
Proposition 3: The coverage density itself after the quiet period is over is a signal.
The 10 analysts had a median target price of $300, a minimum of $250 (Morgan Stanley), and a maximum of $340 (Citigroup). Based on the after-market price of $200, the median target price implied an upward margin of about 50%. Wedbush ($270 target), Needham ($300), Barclays ($280), TD Cowen ($275), and Craig-Hallum (buy) all launched coverage in the same week.
The underlying assumption of multi-headed logic boils down to one sentence:
If AI reasoning becomes a larger computing power market than training (many agencies predict that inference spending will exceed training in 2027), and Cerebras' speed advantage is real and sustainable, then it only needs to cut 3%-5% of the market where Nvidia accounts for 80% + share, which is enough to support current valuations.
Bearish logic: gross profit margin, customer concentration, and vulnerability to a $50 billion valuation
There are three propositions from the bulls, and the bears each have their own rebuttal.
Rebuttal 1: The moat with the speed advantage of reasoning is probably narrower than you might think.
Cerebras' speed advantage is based on on-chip SRAM capacity, but Nvidia isn't waiting. The B300 chip released by NVIDIA in March greatly increased HBM bandwidth, and Groq's LPU architecture also rapidly iterated on inference scenarios.
Looking at it another way: Cerebras customers are currently highly concentrated on OpenAI and AWS, while OpenAI is also one of Nvidia's largest GPU buyers, and the Trainium chip developed by AWS is also covering more and more inference scenarios. Cerebras' major customers are also betting on alternatives, which means that its speed premium will continue to face price negotiation pressure.
Objection 2: The decline in gross margin may not only be “temporary.”
Management attributed the Q2 gross margin reduction from 47% to 36%-38% to temporary rental costs due to insufficient data center capacity. However, the premise of this explanation is that “costs will improve after the launch of the new data center in the second half of the year.”
Considering that the revenue scale will jump in the second half of the year (management made it clear that revenue is loaded on the back end), and the climbing capacity of the new data center itself will require time and capital investment, this recovery path is not an easy one.
A deeper issue is the impact of the business model transformation itself on gross margins. Cerebras switched from selling hardware to selling cloud computing power, which meant that it had to bear the construction, operation, and depreciation costs of data centers. As depreciation costs for self-built data centers are included, there is uncertainty about whether the gross margin of cloud services can be maintained above 50%. The profit margin ceiling for this business model has yet to be tested.
Objection 3: Customer concentration is a “name changed but not solved” problem.
In 2024, the G42 family contributed 85% of Cerebras' revenue. In 2025, G42's share fell to 24%, but MBZUAI (Mohammed bin Zayed University of Artificial Intelligence) soared from nothing to 62%. The S-1 prospectus clearly states that these two companies are “related parties.” Together, the two UAE affiliates still account for 86% of revenue. Diversification of revenue sources is more a change of name than actual fragmentation.
Finally, CBRS's IPO lockdown included an unconventional clause:
If the company's market value continues to exceed $40 billion, the ban on insider shares can be lifted early. At an after-market price of $200, the current market capitalization is around $45 billion, and it is close to the trigger line. In terms of short positions, as of May 29, the shorting ratio was 17.15% of tradable shares, which is a high level. Once the lockdown period is lifted early and a large number of insider shares are released, compounding existing shorting pressure, stock prices may face concentrated sell-off.
Twitter:https://twitter.com/BitpushNewsCN
Compare the TG exchange group:https://t.me/BitPushCommunity
Compare TG subscriptions:https://t.me/bitpush
