AI Infrastructure Deals Point to a New Cloud Computing Race

IBM and Together AI Make a Major Cloud Commitment

IBM and Together AI have signed a $240 million multi-year cloud agreement that will place open-source AI inference workloads on IBM Cloud. Together AI will use an IBM Cloud cluster scheduled to become available in the first quarter of 2027.

IBM plans to build the cluster with large numbers of Nvidia HGX B300 systems, giving Together AI dedicated infrastructure for open-source model inference. Nvidia claims this configuration can deliver 30 times more AI factory output than earlier generations.

The agreement shows how cloud infrastructure is becoming a central part of the push to deploy AI models at scale. Together AI’s focus is open-source inference, while IBM and Nvidia are providing the systems needed to run those workloads through IBM Cloud.

“Enterprises are in a race to adopt agentic AI at scale to drive real business outcomes. IBM and Nvidia are delivering scalable, economical, enterprise-grade AI infrastructure that can help Together AI accelerate innovation for the next generation of AI infrastructure,” Alan Peacock said.

Vipul Ved Prakash described the need for dependable infrastructure behind open models. “Enterprises want the performance of the best frontier models without the closed-model price tag, and that only works if the infrastructure underneath is fast and reliable at scale. Working alongside IBM with Nvidia gives us that foundation.”

More Demand for Nvidia’s HGX B300 Systems

Another agreement points to demand for the same class of hardware. AZIO AI Holdings announced an agreement with Power Champion Investment Limited for the purchase of up to 128 NVIDIA HGX B300 GPU systems.

Using an estimated midpoint price of $600,000 per system, the agreement represents an approximate aggregate hardware value of $77 million. The purchase supports the initial phase of AZIO AI’s relationship with Power Champion.

Taken together, the IBM-Together AI agreement and the AZIO AI purchase show two ways organizations are building AI capacity: through a cloud cluster designed for inference and through dedicated GPU systems. Both arrangements center on NVIDIA HGX B300 hardware, though the agreements serve different infrastructure needs.

Nvidia has also announced plans to partner with lenders to finance AI infrastructure through vehicles totaling more than $500 billion. The consortium includes Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs, and KKR.

That financing effort sits alongside the hardware agreements, creating a picture of AI infrastructure that includes cloud deployments, direct system purchases, and large financial vehicles. Each approach addresses the cost of obtaining the computing power needed for AI workloads.

Platforms Target Multimodal AI and Flexible Compute

Atlas Cloud announced a unified AI inference platform for scalable application development. The platform supports multimodal AI workloads, including text, image, audio, and video generation, and lets users integrate models such as Seedance 2.5 and Seedance 2.0 within the same infrastructure.

By bringing those workload types into one platform, Atlas Cloud is positioning its infrastructure around applications that use more than one kind of generated content. Text, images, audio, and video all fall within the platform’s stated support.

The broader infrastructure market also includes efforts to give organizations more control over how they obtain and use GPU capacity. B3IQ allows organizations to purchase dedicated NVIDIA GPU systems through an incremental payment plan, while its opt-in network puts unused GPU capacity to productive use.

“Organizations want more control over where their AI runs, how their data is handled, and what they pay for compute. B3IQ brings those decisions into one system: dedicated hardware for private workloads and an opt-in network that puts unused GPU capacity to productive use,” said Sean Geng, CTO of B3 Labs.

B3IQ plans to scale its inventory of U.S.-assembled NVIDIA GPU systems to meet demand. Its early users include faculty members, AI researchers, and student-led teams at NYU, Dartmouth, UH Mānoa, and Stanford University.

Private Infrastructure Reaches Research Teams

Research teams at NYU, Stanford, Dartmouth, and UH Mānoa are using B3IQ to power their AI work. That use gives the product a place in academic research, where access to predictable computing costs can shape how teams plan projects.

Pavel Bushuyeu, an AI researcher at the University of Hawaiʻi, described the difference between renting capacity and owning dedicated systems: “B3IQ’s owner-controlled model is a promising path between renting and buying — and that is why we decided to partner with B3IQ. Grants are fixed and awarded upfront; cloud costs are variable and can quietly consume a line item mid-project. Turning compute into a known, budgetable cost makes it easier to plan — and to answer to a PI or a grants office.”

A 2026 Broadcom survey found that 56% of enterprises now run or plan to run production inference on private cloud infrastructure. That figure helps explain why dedicated systems and private infrastructure appear across these announcements, alongside public cloud agreements and unified inference platforms.

B3 Labs, the company behind B3IQ, was founded in 2024 by a team of Coinbase alumni and has raised over $21 million from investors including Pantera Capital and Coinbase Ventures. B3 Labs also operates a 27,000-square-foot Oregon facility.

Across the announcements dated August 6, August 7, August 10, August 11, and August 12, 2026, the same theme keeps returning: AI deployment depends on access to specialized computing. Cloud clusters, dedicated GPU systems, private infrastructure, financing vehicles, and multimodal platforms are all becoming parts of that larger buildout.

Stay connected via Google News

Follow us for the latest AI News