Together AI announces $240M IBM Cloud deal for Nvidia B300 inference
Together AI has struck a $240 million deal with IBM Cloud to run its open-weights inference platform on a large cluster of Nvidia B300 GPUs, marking IBM's first large-scale deployment of those GPUs for inference. The capacity is slated to come online in the first quarter of 2027. The arrangement underscores how AI service providers like Together AI remain dependent on third-party infrastructure, even from competitors, due to power, datacenter, and supply chain constraints. Together AI, which also rents compute from other neoclouds and operates its own datacenters, sees the deal as a way to meet growing inference demand while keeping token costs low. A caveat: the B300 is Nvidia's last-generation HGX platform, so the deal signals pragmatic capacity sourcing rather than access to the newest hardware.
Together AI announces $240M IBM Cloud deal for Nvidia B300 inference
Together AI announced a $240M deal to run its open-weights inference platform on a large cluster of Nvidia GPUs housed in IBM Cloud. IBM says it is the first large-scale use of B300s for inference, with capacity expected online in Q1 2027.
Key takeaway
Together AI's $240M IBM Cloud deal for B300 inference underscores the intensifying scramble for GPU capacity, as even leading inference providers turn to competitors to secure compute.
What happened
Together AI announced a $240 million agreement with IBM Cloud to host its open-weights inference platform on a large cluster of Nvidia B300 GPUs. IBM said it is the first large-scale use of the B300 for inference, with the compute capacity expected to be online in Q1 2027.
According to The Register, the deal stems from power, datacenter, and supply chain constraints that make AI infrastructure scarce, prompting service providers to rent compute even from competing clouds. Together AI, which also operates its own datacenters in Maryland, Memphis, and Sweden, is leveraging IBM's capacity to deliver low token costs while scaling inference services.
Evidence
Together AI announced a $240 million deal to run its open-weights inference platform on a large cluster of Nvidia GPUs in IBM Cloud.
The Register · attributed
Together AI announced a $240M deal to run its open-weights inference platform on a large cluster of Nvidia GPUs housed in IBM Cloud.
IBM says it is the first large-scale use of B300s for inference.
The Register · attributed
IBM says it is the first large-scale use of B300s for inference.
The compute capacity is expected to come online in Q1 2027.
The Register · attributed
with capacity expected online in Q1 2027.
AI infrastructure is in short supply due to power, datacenter, and supply chain constraints, prompting providers to rent compute even from competing clouds.
The Register · attributed
With power, datacenter capacity, and other supply chain constraints, AI infrastructure is in short supply, and service providers will rent all the compute they can get, even if it means working with competing cloud providers.
Together AI tapped IBM because Big Blue had the capacity it needed.
The Register · attributed
Translation: Together AI tapped IBM because Big Blue had the capacity it needed when it needed it.
Why it matters
This deal signals that capacity constraints are reshaping the AI infrastructure market, with providers prioritizing availability over hardware freshness, which could influence pricing and deployment strategies across the industry.
Limits and uncertainties
The exact number of GPUs in the 'large cluster' is not disclosed.
The deal is announced but capacity won't be live until Q1 2027, so there is a waiting period.
IBM's claim of 'first large-scale use' for inference is based on the announcement and may be contested.
Practical implications
Enterprises relying on Together AI's inference API should anticipate potential capacity constraints until the B300 deployment comes online in Q1 2027.
Operators building on open-weights models may want to diversify compute sourcing, as the deal illustrates the value of multi-cloud strategies in a constrained market.
What to watch
Monitor when IBM's B300 capacity actually goes live and whether Together AI expands beyond its current datacenters.
Watch for other cloud providers announcing similar large-scale B300 deployments in response to capacity pressure.
Track inference token pricing as Together AI adds capacity through IBM, which could affect competitive dynamics.
