Quick Read
- OpenAI signed a $10 billion deal with Cerebras Systems in January to boost AI inference capacity.
- The agreement will add 750 megawatts of low-latency AI compute to OpenAI’s platform.
- Deployment of the new capacity is scheduled from 2026 to 2028.
- OpenAI’s annual recurring revenue (ARR) reached over $20 billion in 2025, up from $2 billion in 2023.
- CFO Sarah Friar stated compute capacity tripled to 1.9 gigawatts in 2025, directly correlating with revenue growth.
In a strategic move to bolster its rapidly expanding artificial intelligence capabilities and ensure faster, more responsive services, OpenAI has finalized a substantial $10 billion agreement with chipmaker Cerebras Systems. This landmark deal, announced in January, is set to significantly enhance OpenAI’s AI inference infrastructure, directly impacting the performance of its flagship ChatGPT service and underpinning the company’s ambitious growth trajectory.
The collaboration comes as OpenAI reports an unprecedented surge in its annual recurring revenue (ARR), which soared to over $20 billion in 2025, a dramatic increase from $2 billion just two years prior. According to OpenAI’s chief financial officer, Sarah Friar, this financial expansion is intrinsically linked to the company’s aggressive scaling of compute capacity, which tripled year-over-year to approximately 1.9 gigawatts in 2025. Friar emphasized that the ability to serve customers and generate revenue directly correlates with available compute resources, highlighting it as the “scarcest resource in AI.”
Scaling for Unprecedented Demand
OpenAI’s explosive growth is largely attributed to ChatGPT’s evolution from a novel consumer curiosity to an essential piece of digital infrastructure. Friar noted that daily and weekly active users are at all-time highs, with ChatGPT now deeply embedded in diverse professional and personal workflows, including education, software development, marketing, and finance. This widespread adoption has driven OpenAI’s commercial strategy, expanding from consumer subscriptions to enterprise plans and usage-based pricing for developers through its API platform. The company’s recent foray into advertising and commerce, positioning ChatGPT as a decision-making platform, further underscores the need for robust, low-latency compute infrastructure to support complex, real-time interactions.
The deal with Cerebras Systems is a critical component of OpenAI’s strategy to meet this escalating demand. It will inject an additional 750 megawatts of low-latency AI compute into OpenAI’s platform. This additional capacity is vital for improving the responsiveness of services like ChatGPT, ensuring that user inputs are processed and answered with greater speed and efficiency. The phased deployment of this new capacity is slated to commence this year and is expected to reach full implementation by 2028.
Cerebras Technology: A Leap in AI Inference
Cerebras Systems, an emerging rival to industry giants like Nvidia, is known for its innovative approach to AI acceleration. The company’s technology is designed to significantly speed up AI inference by integrating massive compute power, memory, and bandwidth onto a single, exceptionally large chip. This unique architecture effectively bypasses bottlenecks commonly found in traditional hardware setups, which often hinder the rapid processing required for advanced AI models. Andrew Feldman, CEO and co-founder of Cerebras, articulated the transformative potential of this technology, stating that ‘real-time inference will transform AI, enabling entirely new ways to build and interact’ with models. This capability is crucial for OpenAI as it seeks to not only maintain but also extend its lead in the generative AI space.
For OpenAI, diversifying its compute providers marks a significant shift from its historical reliance on a single vendor. This strategic move aims to enhance resilience, optimize costs, and ensure a continuous supply of cutting-edge hardware necessary for its ambitious research and product development initiatives. The substantial investment in data centers and compute infrastructure is also a clear signal of OpenAI’s intent to vigorously compete with other tech titans such as Meta Platforms, Microsoft, and Google, all of whom are pouring vast resources into their own AI ecosystems.
Future Horizons: Beyond Subscriptions and APIs
Looking ahead, OpenAI’s CFO Sarah Friar outlined a future financial strategy focused on practical adoption, particularly in high-impact sectors like health, science, and enterprise applications. She envisions a future where enhanced AI intelligence directly translates into measurable outcomes, driving innovation in areas such as drug discovery, energy systems, and financial modeling. To support this vision, OpenAI is exploring new revenue models beyond its current subscriptions and API-based pricing, including licensing agreements, intellectual property (IP)-based partnerships, and outcome-based pricing. These diversified revenue streams will further necessitate robust and scalable inference capabilities, making the Cerebras deal a foundational element of OpenAI’s long-term strategic planning.
The $10 billion Cerebras deal underscores OpenAI’s critical understanding that sustained innovation and market leadership in AI are inextricably linked to controlling and expanding its underlying compute infrastructure, positioning it to not only meet current demand but also to pioneer future applications that require increasingly sophisticated real-time processing.