Key Highlights
- Together AI and IBM have entered into a $240 million partnership spanning multiple years to construct a massive AI inference infrastructure on IBM Cloud.
- Nvidia HGX B300 systems featuring Blackwell processors and Spectrum-X Ethernet networking technology will power the infrastructure.
- Together AI plans to leverage this cluster for running open-source AI models, processing approximately 400 trillion tokens each month.
- The infrastructure deployment is scheduled for the first quarter of 2027.
- IBM shares increased 1.51% following the announcement; Nvidia stock rose 0.46%.
A significant $240 million multi-year partnership between IBM and Together AI will bring a cutting-edge AI inference cluster to IBM Cloud, utilizing advanced Nvidia technology. The announcement drove IBM stock up 1.51%, with Nvidia shares climbing 0.46%.
International Business Machines Corporation, IBM
The partnership focuses on AI inference capabilities for open-source models. Inference involves using trained AI systems to produce outputs, and this workload has emerged as a critical driver of computational infrastructure demand.
At the heart of the infrastructure are Nvidia HGX B300 systems, equipped with Nvidia’s advanced Blackwell generation chips. The setup also incorporates Nvidia Spectrum-X Ethernet networking infrastructure for optimal performance.
According to IBM, this represents the inaugural dedicated, enterprise-scale inference cluster on IBM Cloud utilizing these cutting-edge systems. The platform is slated to become operational during the first quarter of 2027.
Together AI provides a comprehensive platform enabling developers and businesses to create and run AI applications using open-source models like DeepSeek, MiniMax, and Kimi. The platform currently processes an impressive 400 trillion tokens on a monthly basis.
Vipul Ved Prakash, CEO of Together AI, explained that the new cluster will enable his company to deliver “production-grade inference to more companies, faster.” He emphasized that businesses are seeking cutting-edge AI performance without the expense associated with proprietary models.
IBM Cloud’s general manager, Alan Peacock, stated that the collaboration between IBM and Nvidia demonstrates their commitment to providing “scalable, economical, enterprise-grade AI infrastructure.”
The momentum behind open-source AI continues to build as organizations seek ways to reduce AI implementation costs. Additionally, recent cybersecurity challenges affecting models from companies like Anthropic, OpenAI, and Meta have accelerated the shift toward open-source solutions.
Nvidia’s Role in the Partnership
The Blackwell processors central to this agreement represent Nvidia’s most recent chip generation, specifically engineered for AI inference applications. Nvidia’s Spectrum-X networking infrastructure is purpose-built to manage the substantial data throughput required by large-scale AI computing environments.
Together AI indicated that their selection of IBM and Nvidia was influenced by their technology development plans and capacity to provide GPU resources efficiently at competitive per-token pricing.
IBM’s Expanding Open-Source Strategy
This partnership aligns with IBM’s strategic direction. The company recently unveiled Project Lightwell, a $5 billion initiative in collaboration with Red Hat aimed at helping businesses strengthen open-source software security through AI-powered tools and a workforce of more than 20,000 engineers.
Early participants in Project Lightwell include major financial institutions such as Bank of America, Goldman Sachs, and JPMorganChase.
The Together AI agreement represents one component of an extensive IBM-Nvidia partnership encompassing GPU-optimized data analytics, unstructured data processing, hybrid cloud and on-premises infrastructure, and professional services.
Together AI most recently completed an $800 million Series C funding round, achieving an $8.3 billion company valuation. This valuation matches the figure established in July of the previous year.
IBM has verified that this cluster will be the first implementation of its kind on IBM Cloud featuring Nvidia HGX B300 systems, with broad availability expected in the first quarter of 2027.
Get 3 Free Stock Ebooks
Discover top-performing stocks in AI, Crypto, and Technology with expert analysis.
- Top 10 AI Stocks - Leading AI companies
- Top 10 Crypto Stocks - Blockchain leaders
- Top 10 Tech Stocks - Tech giants
