Nebius Group has signed a long-term agreement with Meta to provide large-scale artificial intelligence infrastructure as part of the social media company’s expanding data center strategy. The contract has a total potential value of up to $27 billion over five years.

Under the agreement, Nebius will supply approximately $12 billion in dedicated AI computing capacity across multiple locations. The infrastructure will be built on one of the first large-scale deployments of Nvidia’s Vera Rubin computing platform.

Nebius expects to begin delivering the capacity starting in early 2027.

Meta has also committed to purchasing additional computing resources from Nebius clusters as they become available. The total value of these additional purchases could reach $15 billion during the five-year contract period.

Nebius plans to offer the majority of the new infrastructure capacity to its broader AI cloud customer base. Any remaining unused capacity will be reserved for Meta under the agreement.

Expanding AI Cloud Infrastructure

The deal reflects increasing demand for large-scale computing infrastructure as technology companies train and deploy more advanced AI systems. Large models and AI agents require enormous amounts of computing power, driving companies to secure long-term infrastructure agreements.

Nebius has positioned itself as a dedicated AI cloud provider, focusing on high-performance computing environments optimized for machine learning workloads. The company operates specialized infrastructure designed for training and inference tasks used by AI developers and enterprises.

Arkady Volozh, founder and chief executive of Nebius, said the agreement strengthens the company’s collaboration with Meta while supporting the expansion of its AI cloud business.

“We are pleased to expand our significant partnership with Meta as part of securing more large, long-term capacity contracts to accelerate the build-out and growth of our core AI cloud business,” Volozh said.

The partnership also supports Nebius’s broader strategy of building large-scale AI data centers based on Nvidia’s next-generation accelerated computing platforms.

Nebius said its financial guidance for 2026 remains unchanged following the announcement.

Anthropic has introduced the Claude Partner Network, a new initiative designed to support organizations helping enterprises deploy its Claude AI models. The company said it will commit an initial $100 million to the program in 2026, funding partner training, technical assistance, and joint market development.

The program targets consulting firms, professional services companies, and AI specialists that guide enterprises through the adoption of AI systems. These partners typically help businesses identify high-value use cases for AI and implement production-ready applications within complex corporate environments.

Anthropic said the initiative reflects growing enterprise demand for structured support in adopting generative AI tools.

“Anthropic is the most committed AI company in the world to the partner ecosystem—and we’re putting $100 million behind that this year to prove it,” said Steve Corfield, Anthropic’s head of global business development and partnerships. “Our partners are instrumental in getting enterprises from proof of concept to production with Claude.”

Claude is currently available across all three major cloud providers: Amazon Web Services, Google Cloud, and Microsoft Azure, allowing partners to deploy the model across multiple enterprise environments.

Training, Certification, and Implementation Support

The Claude Partner Network will provide a range of resources designed to help organizations build services around Anthropic’s AI platform.

Partners will receive access to technical training through Anthropic Academy, along with dedicated engineering support for enterprise deployments. The company plans to expand its partner-facing team significantly, providing applied AI engineers and technical architects to assist with complex customer implementations.

Anthropic is also introducing its first technical certification, called Claude Certified Architect, Foundations. The certification is designed for solution architects developing production applications using Claude models.

Additional certifications for developers, architects, and sales professionals are expected to launch later this year. Partners joining the network will receive priority access to these programs.

In addition, Anthropic has released a code modernization starter kit intended to help enterprises migrate legacy software systems using Claude’s agentic coding capabilities. The company said modernizing older codebases and reducing technical debt is one of the most common enterprise AI use cases.

Growing Demand for Enterprise AI Deployment

Anthropic said the partner network will help enterprises move from experimental AI deployments to large-scale production systems. Many organizations have begun testing AI models but require technical and operational guidance to integrate them into existing workflows.

Members of the program will gain access to Anthropic’s partner portal, which includes sales playbooks, training materials, and co-marketing resources. Qualified partners will also appear in a public services directory where enterprises can find firms experienced in implementing Claude-based solutions.

The company said the Claude Partner Network is open to any organization involved in deploying Claude AI systems, and membership will be free of charge. Applications for the program opened today.

The partner program also comes as Anthropic expands its broader research and policy efforts around advanced AI. The company recently launched the Anthropic Institute, an initiative that brings together engineers, economists, and social scientists to study the societal, economic, and governance implications of increasingly powerful AI systems.

Amazon and Cerebras Systems have announced a partnership to integrate their computing chips into a new artificial intelligence inference service hosted on Amazon Web Services (AWS). The service aims to accelerate AI workloads such as chatbots, coding assistants, and other generative AI applications.

Under the agreement, Cerebras processors will be installed inside AWS data centers and connected with Amazon’s Trainium3 custom AI chips through proprietary networking technology. The companies did not disclose the financial terms of the partnership.

Cerebras, valued at roughly $23.1 billion, has developed a unique AI chip architecture designed to compete with Nvidia’s processors. Unlike many AI accelerators, Cerebras chips do not rely on high-bandwidth memory, a costly component used in Nvidia’s flagship GPUs.

Cerebras Chief Executive Andrew Feldman said the partnership could expand access to the company’s technology through AWS’s large customer base.

“Every customer large or small is on AWS, from individual developers to the largest banks in the world,” Feldman said. “This will make it easy as a click to get on Cerebras.”

Divide-and-Conquer Inference Architecture

The joint service will focus on inference, the stage where trained AI models generate responses to user prompts. Instead of running the entire inference workload on a single chip architecture, Amazon and Cerebras plan to split the process between two specialized systems.

Amazon’s Trainium3 chips will handle the “prefill” stage, which converts user input into tokens that AI systems process. Cerebras processors will manage the “decode” stage, where the model generates output tokens to produce responses.

According to Feldman, this “divide and conquer” approach is intended to improve speed and efficiency for AI inference workloads.

The strategy reflects a broader industry trend toward specialized AI infrastructure where different processors handle distinct parts of AI pipelines to improve performance and cost efficiency.

Competition in AI Infrastructure

The partnership comes amid intensifying competition in AI hardware and infrastructure. Nvidia currently dominates the AI accelerator market, but several companies are attempting to develop alternatives to its GPU-based architecture.

Analysts expect Nvidia to unveil a similar multi-chip strategy soon. Reports indicate the company may combine its GPUs with chips from startup Groq, which Nvidia acquired in a deal valued at $17 billion late last year.

Amazon said its service, expected to launch in the second half of the year, could offer better price-performance compared with GPU-based solutions.

The company added that its Trainium chip roadmap, including the upcoming Trainium4 processor, is designed to provide an alternative to third-party AI hardware by optimizing cost and efficiency for AWS customers.