IBM and Together AI sign $240 million AI inference deal
IBM and Together AI signed a multi-year $240 million deal to deploy a large-scale AI inference cluster on IBM Cloud using Nvidia HGX B300 systems and Spectrum-X networking, with availability expected in Q1 2027. Together AI will run inference on open-source models via its platform, which it says serves 400 trillion tokens monthly.
How this was made
The 30-second read
Why it matters
A named, multi-year inference infrastructure agreement on IBM Cloud using Nvidia hardware is a tangible step toward production-grade open-source inference at scale, with a stated availability target in Q1 2027.
Market read
Traders can use the deal size and Q1 2027 availability window as a datapoint for AI inference infrastructure demand, particularly GPU plus networking deployments on major cloud platforms.
What to watch
The article does not specify contract duration details, margin profile, or whether IBM Cloud will resell capacity at scale beyond Together AI, which could materially affect financial significance.
Background
IBM is expanding open-source AI infrastructure efforts, including Project Lightwell with Red Hat, while Together AI provides a platform for deploying open-source models.
Ticker impact
IBM signed a multi-year $240M inference-cluster deal on IBM Cloud using Nvidia HGX B300 systems, with availability targeted for Q1 2027.
Moderate positive bias for IBM on deal credibility, but likely limited near-term impact unless IBM discloses financial terms beyond the headline.
The article provides deal size ($240M), scope (inference cluster), and timing (Q1 2027), which are actionable for positioning, but it does not provide incremental financial guidance or immediate revenue recognition details.
The deal specifies Nvidia HGX B300 systems and Spectrum-X Ethernet networking as the core of the $240M inference cluster for Together AI on IBM Cloud.
Slight positive read-through for NVDA, though the article does not quantify Nvidia’s revenue share or incremental capacity beyond the cluster headline.
While Nvidia is explicitly named as the hardware provider, the text does not break out Nvidia’s financial exposure, making the direct earnings impact uncertain.
Market effects
Reinforces the enterprise inference buildout trend, highlighting GPU plus high-speed networking as a packaged requirement for open-source model deployment.
Primarily US enterprise AI infrastructure demand, with Together AI based in San Francisco.
Signals continued global enterprise adoption of Nvidia-based inference infrastructure on major cloud platforms.
Counterpoint
The $240M headline may not translate into near-term earnings impact for IBM or Nvidia if revenue recognition is back-end loaded or if the deal is largely infrastructure provisioning.
Key entities
- companyIBM
Signs a multi-year $240M inference-cluster agreement on IBM Cloud using Nvidia hardware for Together AI.
- companyTogether AI
Uses the cluster to run inference on open-source AI models and reports serving 400 trillion tokens monthly.
- companyNvidia
Provides HGX B300 systems and Spectrum-X Ethernet networking gear for the inference cluster.
