Oracle will offer cloud services running on Cerebras Systems’ AI computers in its data centers starting in 2027, the companies announced on Tuesday at the annual AI World conference. The move gives enterprise customers another option for heavy computing tasks, formalizing an Oracle Cerebras deployment designed to run large models on whole silicon wafers rather than standard graphics cards [1]. As part of a multi-vendor computing strategy, Oracle Cloud Infrastructure is expanding its hardware lineup with alternative accelerators to meet growing customer demand for artificial intelligence inference [2].
Datacenter Rollout for 2027
Oracle Cloud Infrastructure plans to begin hosting the Cerebras systems in 2027. The vendor did not say how much Cerebras hardware it will use in its facilities. Cerebras listed on public markets in May [1].
The planned deployment gives Cerebras another prominent cloud partner as it competes against established semiconductor suppliers in the computing industry. According to reporting from Bloomberg, the young challenger has steadily expanded its presence by signing major technology organizations to its platform. Cerebras already counts OpenAI and Amazon Web Services as customers, and the Oracle contract provides an additional high-volume hosting environment for its wafer-level systems. By securing multiple large clients, the chipmaker aims to show that non-traditional processor designs can handle demanding production workloads together with standard graphics clusters in modern data centers [1].
To support this hardware expansion, Oracle plans to spend $70B on capital projects in the year to May 2027. That planned expenditure follows $55.7B spent during the previous year [1].
Wafer Scale Design in Oracle Cerebras Systems
Cerebras takes an unusual engineering approach by building each chip from a whole silicon wafer, the round semiconductor disk that fabrication facilities normally cut into many smaller individual processors. Conventional chipmaking methods slice that silicon disk into hundreds of distinct individual dies, which are then packaged separately into standard central processing units or graphics cards. By keeping the entire silicon wafer intact, Cerebras builds a giant single processor containing millions of processing cores connected across the unbroken silicon surface. The company says this unbroken design lets its computing systems answer artificial intelligence requests much faster than rival processors that rely on traditional modular packages [1].
Data moves through the wafer without passing through external cables. Cerebras claims this architecture eliminates networking bottlenecks between separate chips. Fast data movement helps models run without latency delays [1].

The resulting computing systems will focus on artificial intelligence inference, which involves generating answers and processing live queries once an algorithm has completed its initial training phases. While training large foundation models requires months of uninterrupted raw compute power, running daily inference for millions of end users requires rapid response times and consistent data throughput. The Oracle Cerebras machines will give business clients an alternative way to serve complex language models without competing for constrained supplies of conventional graphics processors. That physical design advantage helps cloud customers scale interactive applications that require near-instantaneous token generation [1].
Revenue Projections and Market Share
Shares in Cerebras rose about 9.5% on Tuesday following the announcement. The stock gain reflected investor interest in its expanding roster of commercial cloud partnerships [1].
Financial forecasts compiled by Bloomberg indicate that Cerebras could experience steep revenue growth over the next twelve months as its cloud deployments take effect. Analysts expect the company to triple its annual revenue next year to almost $3B, up from less than $1B recorded this year. Reaching that three-billion-dollar target would represent rapid expansion for a young hardware manufacturer that completed its public listing in May. Sustained revenue growth is considered important for funding the capital-intensive research, specialized fabrication tools, and complex wafer packaging needed to maintain technological parity with market incumbents [1].
Investors still worry about its position in a market where Nvidia dominates. Nvidia’s computing products bring in close to $100B in sales each quarter. That massive scale gives the market leader strong leverage across global supply chains [1].

Why Oracle Cerebras Matters for Inference
Executive remarks from Oracle show the strategic motivation behind bringing non-traditional processor architectures into commercial data centers. “AI inference demands choice,” said Mahesh Thiagarajan, president of Oracle Cloud Infrastructure, when explaining why the cloud provider decided to partner with Cerebras to expand hardware variety for business customers. Thiagarajan pointed out that enterprise clients operate with diverse technical requirements, meaning cloud providers can’t rely on a single hardware blueprint to satisfy every enterprise use case. By offering wafer-scale machines in addition to standard server racks, Oracle aims to give software engineers greater flexibility when tuning large generative workloads for cost and response time [1].
Inference tasks require low response times and high memory bandwidth. Serving millions of live requests quickly requires processors that move data without internal delays. Wafer-scale chips keep memory directly next to the computing cores [1].
The Oracle Cerebras integration also provides engineering teams with an alternative platform as they transition experimental artificial intelligence software into high-volume production. Development teams can test whether their custom neural network architectures run more efficiently on wafer-scale systems compared to traditional graphics clusters. Providing multiple hardware tiers allows Oracle to cater to varied enterprise budgets and performance requirements without relying on a single silicon provider. As corporate artificial intelligence spending expands, cloud operators that provide diverse hardware options may attract enterprises seeking to optimize operational costs and avoid hardware shortages [1].
Alternative Chips Across Cloud Platforms
The Cerebras agreement was not the only silicon announcement at the conference. Oracle announced deals with Nvidia and Intel during the same event. The company also buys Arm’s own chips for its cloud instances [1].
Oracle is also broadening its hardware catalog through collaborations with other artificial intelligence leaders in the technology sector. According to reporting summarized by Techmeme, Oracle Cloud Infrastructure is expanding its lineup with Google TPUs, OpenAI’s Jalapeño chip, and Cerebras hardware. In addition to raw computing infrastructure, CNBC reporter Seema Mody noted that Oracle announced an expanded collaboration with OpenAI that adds ChatGPT and Codex to its enterprise application stack. Integrating these conversational models directly into business software allows corporate customers to automate programming tasks, draft internal documentation, and analyze large databases while keeping proprietary records inside managed corporate cloud environments [2].
Other major cloud providers have also deployed specialized accelerators to reduce their dependence on third-party semiconductor vendors. PerEXP Teamworks previously reported on how Microsoft unveils Maia 300 AI chip for cloud workloads to power its internal software services and Azure cloud platform. Through the Oracle Cerebras deployment and other hardware partnerships, Oracle is building a multi-vendor catalog that lets enterprise buyers select the specific silicon architecture best suited for their artificial intelligence workloads [2].
- ONLINE NEWS Dina, C. (2026, October 27). Oracle will offer Cerebras AI computers in its cloud from 2027. The Next Web. [Article Link]
- ONLINE NEWS Mody, S. (2026, October 27). Oracle adds ChatGPT and Codex to its app stack and expands Oracle Cloud Infrastructure lineup with Google TPUs, OpenAI’s Jalapeño chip, and Cerebras hardware. CNBC / Techmeme. [Article Link]
1 comment