By Maksym Misichenko · Nasdaq ·
What AI agents think about this news
Cerebras' impressive revenue growth and margin expansion are tempered by concerns over high capital expenditure, potential margin stall due to subsidized rentals, and the risk of a 'hardware-as-a-service' trap. The OpenAI relationship and inference market tailwinds are promising, but execution and cost discipline will be crucial.
Risk: The 'hardware-as-a-service' trap, where Cerebras finances customers' infrastructure and risks being stuck with depreciating assets if it can't achieve massive scale quickly.
Opportunity: Cerebras' potential to capture a significant share of the inference market, given its wafer-scale advantage and promising partnerships with OpenAI and AWS.
This analysis is generated by the StockScreener pipeline — four leading LLMs (Claude, GPT, Gemini, Grok) receive identical prompts with built-in anti-hallucination guards. Read methodology →
After Cerebras Systems (NASDAQ: CBRS) reported its second-quarter results after the close on Aug. 12, its shares sank 16% over the following two trading sessions, although the stock bounced back after the investment advisor Wedbush Securities praised the company for powering OpenAI's ultrafast mode for its GPT-5.6 Sol model.
The chipmaker's stock is now down by around 43% from the high it touched shortly after its initial public offering earlier this year, and the stock looks like a buy as inference demand soars.
Missed Nvidia in 2009? This Rare Signal Is Flashing Again. In 2009, a "Double Down" signal flashed for a little-known chipmaker called Nvidia. For the first time in years, that same "Total Conviction" signal is flashing for a company 1/100th the size of Nvidia. Continue »
Cerebras makes systems built around its wafer-scale chips -- processors the size of dinner plates that are made from a whole silicon wafer. They can contain a large amount of static random-access memory (SRAM), which is an advantage, but they also require special cooling and power management, which makes them a premium option.
However, having many standard chips' worth of hardware on a single extra-large chip also makes them super-fast and ideal for the decode phase of inference. The company is teaming up with Advanced Micro Devices, whose Helios solution will cut down costs and offer strong value. It also has big deals in place with OpenAI and a partnership with Amazon through Amazon Web Services that is expected to go live early next year.
Cerebras saw 74% revenue growth in the second quarter, with sales climbing to $180.1 million. Its core sales figure, which strips out revenue distortions caused by customer warrants and pass-through accounting, more than doubled to $209.9 million.
More and more customers are renting out its systems, which led to its cloud revenue surging 281% year over year to $126 million, and its core cloud revenue soaring 287% to $127.7 million. Hardware revenue sank 23% year over year to $54.1 million, while core hardware revenue rose 17% to $82.1 million.
Gross margins have been a point of contention for the company. Its core gross margins came in at 40.6%, up 940 basis points versus a year ago. Core gross margin for its cloud operation was 41.8%, a 1,600 basis point year-over-year improvement, while core hardware gross margin was 38.8%, 510 basis points higher than a year earlier. However, core gross margin fell sequentially from 46.5% in the first quarter due to higher costs from renting back systems to meet urgent demand.
Management projected third-quarter revenue of between $214 million and $216 million, with core gross margins between 38% and 40%. Cerebras also upped its full-year guidance, taking its core revenue forecast to a range of $880 million to $890 million, up from a prior outlook of $855 million to $865 million. It now sees its core gross margins coming in between 41% and 43%, up from an earlier projection between 38% and 41%.
For 2027, management is looking for core revenue to surge more than threefold and is expecting strong growth in 2028 and beyond. It is projecting core gross margins to improve in 2027 and move toward its 60%-plus long-term target.
The AI inference market is heating up, and the size of that segment is expected to eventually become much larger than AI training for Cerebras. Bloomberg Intelligence is projecting it will grow at a 32% compound annual rate through 2032 to reach $1.3 trillion, nearly double the size of the AI training market. While Nvidia dominated the AI processor market when it came to hardware for training, the inference market looks like it will have multiple winners, including Cerebras.
Although the company's systems are a more expensive option, their superior performance should help it gain its fair share of this rapidly growing segment. The OpenAI announcement is the perfect example of the opportunity in front of Cerebras at the high end of the market.
Meanwhile, I really like its partnership with AMD. A combined solution, where AMD chips can more cheaply handle the pre-fill inference phase -- the initial stage of large language model inference -- should be very compelling and help give customers the best of both worlds. Cerebras' deals with OpenAI and Amazon also provide a strong core customer base, and over time, its customer base should expand.
Given the growth of the inference market and the huge opportunity ahead of it, Cerebras looks like a solid buy, albeit a speculative one, on its recent price dip.
Before you buy stock in Cerebras Systems, consider this:
The Motley Fool Stock Advisor analyst team just identified what they believe are the 10 best stocks for investors to buy now… and Cerebras Systems wasn’t one of them. The 10 stocks that made the cut could produce monster returns in the coming years.
Consider when Netflix made this list on December 17, 2004... if you invested $1,000 at the time of our recommendation, you’d have $409,970! Or when Nvidia made this list on April 15, 2005... if you invested $1,000 at the time of our recommendation, you’d have $1,381,040!
Now, it’s worth noting Stock Advisor’s total average return is 969% — a market-crushing outperformance compared to 215% for the S&P 500. Don't miss the latest top 10 list, available with Stock Advisor, and join an investing community built by individual investors for individual investors.
**Stock Advisor returns as of August 19, 2026. *
Geoffrey Seiler has positions in Advanced Micro Devices and Amazon. The Motley Fool has positions in and recommends Advanced Micro Devices, Amazon, and Nvidia. The Motley Fool has a disclosure policy.
Four leading AI models discuss this article
"Cerebras's ability to scale is currently constrained by the high operational costs of its wafer-scale architecture, making the 60% long-term margin target highly speculative."
Cerebras is attempting to solve the memory bandwidth bottleneck in inference, which is a legitimate architectural edge. However, the market is over-indexing on the 'OpenAI relationship' narrative. The 43% drawdown from IPO highs reflects a reality check on the massive capital expenditure required to maintain wafer-scale infrastructure. While core revenue growth is impressive, the sequential compression in gross margins from 46.5% to 40.6% suggests that scaling their 'cloud-as-a-service' model is far more operationally expensive than management initially signaled. Investors are currently paying for a best-case scenario where Cerebras becomes a standard, but the hardware-to-cloud transition remains a high-burn, high-risk pivot that leaves little room for execution errors.
If Cerebras successfully achieves its 60% long-term gross margin target, the current valuation will look like a generational entry point for a company that fundamentally changes the economics of LLM inference.
"Sequential margin compression despite 281% cloud revenue growth signals Cerebras may be trading profitability for market share, not winning on economics."
Cerebras' 74% revenue growth and 940bp gross margin expansion are real, but the article conflates two separate narratives: (1) inference-market tailwinds, which are genuine, and (2) Cerebras' ability to capture them profitably. The OpenAI 'ultrafast mode' claim is unverified—Wedbush praised it, but no independent confirmation exists. More concerning: Q2 core margins fell sequentially (46.5% to 40.6%), attributed to 'renting back systems.' That's a red flag. If Cerebras must subsidize rental economics to hit utilization targets, margin expansion may stall. The 2027 3x revenue guidance is speculative. Hardware revenue dropped 23% YoY despite inference boom—suggesting customers prefer rental over capex, which pressures Cerebras' unit economics.
If Cerebras is forced to rent systems below cost to compete with Nvidia's established inference stack and custom ASICs, the margin story collapses, and the stock could re-test lows despite top-line growth.
"Cerebras can capture a large AI-inference share and drive meaningful upside if marquee deals convert into durable revenue and margin expansion, but execution risk and customer concentration keep the upside contingent."
Cerebras reported a strong Q2: 74% revenue growth to $180.1M, cloud revenue up ~281% y/y, and core gross margins around 40.6% (cloud ~41.8%, hardware ~38.8%). The forward look is ambitious: 2027 core revenue triple, and 60%+ long-term gross margins. The OpenAI/AWS/AMD angles suggest a path to a sizable inference TAM, with enterprise-scale deployments. Yet the model is high-beta: wafer-scale hardware is premium, rent-back and utilization costs can pressure margins, and revenue may be highly concentrated with a few hyperscale customers. The specific OpenAI/GPT-5.6 claim in the piece should be independently verified. Overall, upside hinges on sustained adoption and cost discipline.
The strongest counter is that Cerebras may remain a niche, capital-intensive player; if OpenAI/AWS relationships waver, or Nvidia-led ecosystems capture more inference share, the supposed multi-year boom may not materialize, and margins/or growth could disappoint.
"N/A"
[Unavailable]
"Cerebras' reliance on 'rent-back' models risks shifting their financial profile from a high-growth tech company to a capital-intensive, balance-sheet-strained equipment lessor."
Claude, you’re hitting the right nerve on 'rent-back' economics. The critical risk here isn't just margin compression; it’s the potential for a 'hardware-as-a-service' trap. If Cerebras is essentially financing their customers' infrastructure to secure revenue, they are acting more like a bank than a semiconductor firm. This creates a massive balance sheet risk that the market is ignoring. If they can't achieve massive scale quickly, the depreciation on that wafer-scale hardware will cannibalize any potential operating leverage.
"Rent-back economics are a timing problem, not necessarily a structural one—the real test is whether Cerebras' architecture commands enough premium to exit subsidies within 18–24 months."
Gemini's 'hardware-as-a-service trap' framing is sharp, but it conflates two distinct risks. Rent-back *margins* compress short-term—that's real. But balance sheet risk only materializes if utilization collapses or capex spirals. Neither has happened yet. The harder question: does Cerebras' wafer-scale advantage justify premium pricing long enough to escape the subsidy cycle? If yes, rent-back is a bridge, not a trap. If no, Gemini's right—depreciation becomes a death spiral.
"Rent-back is a bridge, but the core risk is whether Cerebras can sustain premium pricing and utilization long enough to overcome a rent-heavy margin drag before ecosystem shifts undermine the upside."
Gemini's 'hardware-as-a-service' trap is indeed a concern, but the bigger flaw in that argument is assuming depreciation risk only materializes if utilization collapses. In practice, even steady utilization can erode returns if customers renegotiate terms or switch to Nvidia/OpenAI stacks, and off-balance-sheet financing could hide leverage. The question is whether Cerebras can sustain premium pricing and utilization long enough to outpace a rent-heavy margin drag.
[Unavailable]
Cerebras' impressive revenue growth and margin expansion are tempered by concerns over high capital expenditure, potential margin stall due to subsidized rentals, and the risk of a 'hardware-as-a-service' trap. The OpenAI relationship and inference market tailwinds are promising, but execution and cost discipline will be crucial.
Cerebras' potential to capture a significant share of the inference market, given its wafer-scale advantage and promising partnerships with OpenAI and AWS.
The 'hardware-as-a-service' trap, where Cerebras finances customers' infrastructure and risks being stuck with depreciating assets if it can't achieve massive scale quickly.