IBM and Together AI Strike $240M Deal for Open-Source Inference

IBM and Together AI have formalized a multi-year, $240 million partnership to construct a dedicated AI inference cluster on IBM Cloud, marking a significant investment in open-source AI infrastructure.

The collaboration centers on deploying Nvidia HGX B300 systems combined with Spectrum-X Ethernet networking equipment. IBM characterized the project as the first large-scale cluster purpose-built for inference operations on its cloud platform using this hardware generation. The infrastructure is scheduled to become operational in the first quarter of 2027.

Together AI, a San Francisco-based platform provider, will leverage the cluster to execute inference workloads on open-source AI models. The company currently processes approximately 400 trillion tokens monthly for developers and enterprises building AI applications through its platform.

Why it matters

This deal represents a strategic bet on open-source AI as a viable alternative to proprietary frontier models. As enterprises seek performance comparable to closed models without premium pricing, infrastructure providers are racing to deliver the scale and reliability needed to make open-source inference economically competitive. The $240 million commitment signals confidence that demand for open-weight models will justify massive infrastructure investments.

Infrastructure economics at scale

Together AI CEO Vipul Ved Prakash framed the partnership as addressing a fundamental market need. "Enterprises want the performance of the best frontier models without the closed-model price tag, and that only works if the infrastructure underneath is fast and reliable at scale," he stated. The cluster aims to enable production-grade inference for a broader customer base.

The company recently secured $800 million in Series C funding at an $8.3 billion valuation. According to Together AI, it selected IBM and Nvidia based on their product development trajectories and capacity to deliver GPU resources at the velocity and cost structure required for AI scaling.

Alan Peacock, general manager of IBM Cloud, emphasized the partnership's focus on delivering "scalable, economical, enterprise-grade AI infrastructure" to accelerate Together AI's innovation roadmap.

Broader open-source commitments

The inference cluster agreement aligns with IBM's wider open-source strategy. Earlier this year, the company unveiled Project Lightwell, a $5 billion initiative developed with Red Hat to help enterprises secure open-source software using AI-powered tools and more than 20,000 engineers. That program establishes a trusted clearinghouse that employs AI to validate and verify patches across substantial portions of the open-source ecosystem. Pilot participants include Bank of America, Goldman Sachs, and JPMorgan Chase.

The Together AI partnership also forms part of an expanded collaboration between IBM and Nvidia encompassing GPU-native data analytics, unstructured data extraction, hybrid infrastructure spanning on-premises and cloud environments, and consulting services.

Details of the agreement were first reported by Quartz.