IBM, Together AI team to offer open-source AI inference in the cloud
IBM said it signed a $240 million deal with Together AI to deliver open-source AI inference services on IBM Cloud. The offering will run on a cluster of Nvidia HGX B300 systems with Spectrum-X Ethernet networking. Together AI will use the cluster to serve enterprise customers using open models such as DeepSeek and GLM. Gartner projects AI-optimized IaaS spending to rise 96% through 2026 to $42B.
How this was made
The 30-second read
Why it matters
The disclosed deal adds a concrete enterprise-focused use case for open-model inference on IBM Cloud and strengthens the Nvidia ecosystem narrative around GPU-native networking and inference workloads.
Market read
Traders get a fresh, dollar-denominated cloud infrastructure deal that ties IBM Cloud’s enterprise AI inference push to Nvidia-native hardware and networking.
What to watch
The article does not disclose customer commitments, pricing, or margin economics; traders may need follow-up on contract duration, renewal terms, and whether IBM bears meaningful infrastructure costs.
Background
IBM Cloud will host Together AI’s open-source model inference platform on Nvidia HGX B300 clusters with Spectrum-X networking, targeting enterprise production deployments.
Ticker impact
IBM signed a $240 million deal with Together AI to run open-source AI inference workloads on IBM Cloud using Nvidia HGX B300 systems.
Near-term: modest positive bias for IBM sentiment; follow-through depends on customer adoption and margin profile of the inference service.
This is a disclosed, dollar-denominated cloud infrastructure deal with specific technical details (Nvidia HGX B300, Spectrum-X, IBM Cloud) that can drive incremental enterprise usage, but the article does not provide financial guidance or customer names to size impact.
The service is based on a large cluster of Nvidia HGX B300 systems with Spectrum-X Ethernet networking used for Together AI inference on IBM Cloud.
Near-term: supportive for Nvidia sentiment, but likely limited incremental impact versus Nvidia’s broader AI infrastructure demand.
The article specifies Nvidia hardware and networking in the deployment, but does not quantify unit volumes, revenue recognition, or incremental spend beyond the $240 million deal value.
Market effects
Supports the AI-optimized IaaS thesis that inference workloads are becoming the dominant cloud consumption model, aligning with Gartner’s forecast cited in the article.
No specific regional demand signal beyond enterprise cloud deployments.
Reinforces global enterprise adoption of open-model inference stacks on major cloud platforms using Nvidia-native infrastructure.
Counterpoint
A single $240 million infrastructure deal may be small relative to IBM Cloud’s total revenue and Nvidia’s AI spend, limiting stock impact.
Key entities
- companyIBM
Signed a $240 million deal with Together AI to deliver open-source AI inference services on IBM Cloud.
- companyTogether AI
Provides an inference platform supporting open models and will use the IBM Cloud Nvidia cluster to serve enterprise customers.
- companyNvidia
Supplies HGX B300 compute and Spectrum-X Ethernet networking used in the IBM Cloud inference cluster.
- research_firmGartner
Forecasts rapid growth in AI-optimized IaaS spending and highlights inference as a dominant consumption model.
