Cerebras Systems Shares Rise for Three Straight Days, Return to $240: Strategic Partnership Reached With CrowdStrike, CS-3 Inference Speed Exceeds Nvidia by 15 Times

T radingK ey - Nv idia com petitor Cereb ras Sys tems ( C BRS) shares rose for three consec utive day s, retu rning to a one-m onth hig h. As of press t ime, th e stock was up 3.89 % to $2 16.69.

Ce rebras t oday an nounced that it h as teame d up wit h cyber s ecurity giant Cr owdStri ke ( CRWD) to es tablish a strateg ic partn ership. Cerebras represen ts the w orld's f astest A I infere nce comp uting so lution, while Cr owdStri ke repre sents th e indust ry-leadi ng AI-na tive sec urity pl atform. The two companies are joi ning for ces to i ntegrate extreme inferenc e speed with AI -native security capabil ities fo r enterp rise cus tomers d eploying AI at sc ale.

[Source: Trading View]

Un der the partner ship, C rowdStr ike wil l levera ge Cere bras's i nference speed to power i ts Falco n AI De tection and Res ponse ( AIDR); m eanwhile , Cerebr as will standard ize the deployme nt of th e CrowdS trike Fa lcon pla tform to secure i ts own o peration s. The l ogic of both co mpanies is clear : AI is fundame ntally r eshapin g the cy bersecur ity land scape, w ith atta cks beco ming fas ter, mor e target ed, and harder t o detect , while defenders need se curity s ystems c apable o f thinki ng and a cting at machine speed.

Naor Penso, Chief I nformati on Secur ity Offi cer of C erebras, stated in the annou ncement: "Infer ence is the key to reali zing AI' s value, and cybe rsecurit y is one of the s cenarios that dem ands the highest computin g speed. When an attack o ccurs, s ecurity defenses cannot b e held h ostage b y sluggis h AI com puting q ueues. E very mil lisecond counts—i t direct ly deter mines wh ether AI proactive ly stops an attac k or is left to perform a post-mo rtem ana lysis of the eve nt."

What Kind of Company Is Cerebras?

Cerebras's core selling point is: the world's fastest AI inference.

According to company data, the speed of its CS-3 system in specific model inference scenarios can be up to 15 times that of Nvidia's ( NVDA) GPUs; based on calculations by third-party research firm SemiAnalysis, in the Llama 3 70B inference scenario, the total cost of the CS-3 is 32% lower than that of Nvidia's B200, with end-to-end latency reduced by up to 21 times. Cerebras is currently the only manufacturer mass-producing wafer-scale chips and leads in inference speed benchmarks, though Nvidia still dominates the overall AI computing power market.

This speed advantage is particularly critical in the era of AI agents. AI inference workloads are projected to grow at a compound annual growth rate (CAGR) of 79% through 2030, far outpacing the 25% growth rate of training. Inference scenarios require latency to be under 50 milliseconds; an Akamai study shows that 64% of surveyed organizations require response times of less than 250 milliseconds, and 50% consider latency to be the greatest scaling challenge. Entering the agent era, AI models need to process and complete multiple steps, with latency from each step accumulating sequentially—thereby further unlocking growth potential for fast inference solutions.

Meanwhile, the company has signed agreements worth over $20 billion with OpenAI, and multiple GPT models are currently running on the CS-3 system. Starting in the first quarter of 2026, data center costs related to the OpenAI contract will be passed through to OpenAI with a 3% markup.

In addition, Cerebras has partnered with Amazon to combine the latter's Trainium chips with the CS-3, aiming to offer faster inference solutions. Any customer using Amazon Web Services (AWS) can run AI models leveraging Cerebras's computing power—Amazon brings computing capacity and an existing customer base, while Cerebras provides one of the most powerful inference chips.

However, it is worth noting that in its first post-IPO earnings report, Cerebras delivered a financial performance characterized by rapid revenue growth but persistent losses.

During the period, the company's total revenue was approximately $194 million, representing a 94% year-over-year increase. Among this, cloud services revenue surged 178% year-over-year, serving as the primary growth engine, while hardware (CS-3) revenue grew 59% year-over-year. The company expects hardware revenue to continue to decline over the next few quarters, as most of its hardware capacity will be deployed to Cerebras Cloud to fulfill its major contracts.

However, it remains unprofitable, posting a net loss of $2.48 million and a non-GAAP operating loss of $3.5 million. More importantly, the company expects its cloud services profit margins to decline by 10 to 15 percentage points over the next few quarters as it invests heavily in expanding computing capacity.

Overall, with its unique wafer-scale chip architecture, Cerebras has carved out a technological path distinct from Nvidia's in the AI inference track, validating its commercial value through partnerships with OpenAI, Amazon, and CrowdStrike. Yet, behind the high growth lies the reality of persistent losses, heavy debt, and pressured profit margins. Against the backdrop of certain, explosive growth in AI inference demand, whether Cerebras can achieve profitability while scaling up will be the core variable determining its investment value.