- Together AI taps IBM Cloud in a $240M compute deal to scale open-source AI inferencing on Nvidia GPUs
- IBM gains a neocloud win as Together AI chases enterprise demand for reliable, high-performance AI infrastructure
- The deal highlights rising pressure around AI compute capacity, token economics and enterprise-grade cloud services
Together AI is looking to give open-source AI developers a bit more compute muscle and a bit more enterprise credibility, striking a $240 million deal with IBM to deploy Nvidia GPU clusters on the IBM Cloud.
As part of the deal with IBM, Together AI will be deploying a large-scale cluster of NVIDIA HGX B300 systems on IBM Cloud. The cluster will be used to serve open-source model inferencing. Alan Peacock, GM of IBM Cloud, told Fierce the deal will be key to serving up enterprise-grade AI capabilities as businesses across the spectrum lean into inferencing.
“I see growth in inferencing accelerating enormously over the next few years, and therefore token economics are going to be a huge focus for the industry,” he said.
Vipul Ved Prakash, CEO at Together AI, said in a statement: “Working alongside IBM with NVIDIA accelerates our mission to make advanced AI broadly accessible through open source and to empower builders with the infrastructure and platform capabilities they need to build the future.”
Launched in 2022, Together AI is a neocloud platform for training and running open-source AI models. It also has a firm focus on serving enterprise clients, having launched a dedicated enterprise platform in 2024. The company has already secured $1.15 billion in annual bookings and raised more than $1.2 billion to date. Its latest funding round brought in $800 million, which it said would go toward expanding its capacity and infrastructure footprint and expanding its products and features.
Better together
IBM appears to be a beneficiary of some of that spending, but the relationship is certainly reciprocal. If IBM is benefitting from Together AI’s compute need, Together AI is equally benefitting from IBM’s solid reputation among business customers and its differentiated set of enterprise cloud capabilities, Peacock said.
Peacock said IBM has received strong positive feedback about the stability of its cloud environment versus other cloud providers. Enterprises are equally attracted to its service capabilities and proactive approach to solving problems when things go wrong.
“We see a lot of feedback from neocloud clients that quality of service they’re receiving is not where they need it to be,” Peacock said. “So, we believe there’s an opportunity in that.”
He added while some hyperscalers appear to be focused on securing large deployment deals with large customers, “we see small to mid-tiers struggling to get capacity.” That’s another opportunity for IBM.
IBM’s deal with Together AI currently just involves the latter paying the former for compute capacity. But Peacock said he sees the potential to expand the partnership in the future as both work to serve enterprises better.
IBM’s AI ambitions
As for IBM’s own ambitions, Peacock said it’s not going to go on the same kind of spending spree hyperscalers are engaged in to turn up more capacity.
He noted it recently expanded its compute presence in Toronto and Montreal, Canada, and teamed up with telecom operator Airtel to gain a foothold for its services in India. These moves were sufficient to close the gaps IBM felt it had in its portfolio, which also includes cloud regions in D.C. and Dallas in the U.S., London, Frankfurt and Madrid in Europe and Sydney and Tokyo in the Asia-Pacific region.
“I’m not going to go and sign up for massive infrastructure deals to build a lake and then hope somebody comes to drink from it. That’s just not the commercial proposition I’m targeting,” he said. “But we do see opportunities for the right set of clients that are aligned to our strategic goals.”
Though capacity constraints are rampant throughout the cloud space at the moment, Peacock said he actually feels good about IBM’s current capacity position relative to demand.
“We’re a little bit ahead, but I wouldn’t say massively. It’s not as if I’ve got huge amounts of capacity just waiting to be consumed, but it’s not zero either,” he explained. For large deals, though, it would likely need to build, he conceded.
Asked what to expect from IBM Cloud in the next 12 months, Peacock offered a teaser: “I think you’re going to see our name coming up a lot more in that enterprise grade cloud environment.”
Read more about IBM here:
AI governance gap widens as enterprises race to deploy agentic AI, IBM warns
Enterprises are losing control of AI stacks but telcos can help, says IBM
Red Hat, IBM launch $5B Project Lightwell for open source upkeep
Fragmented data drains the competitive 'moat,' says IBM
IBM drills down to the core of the sovereign AI problem