Nvidia-backed Upscale AI has launched a networking platform designed to allow data centers to use artificial intelligence chips from competing suppliers, addressing a growing challenge as cloud operators seek greater flexibility in their infrastructure. The move could have strategic significance for the AI infrastructure market by reducing the complexity of connecting different processors and potentially improving the utilization of increasingly expensive computing capacity.

Token Fabric Targets a Fragmented AI Chip Market

Called Token Fabric, the platform combines hardware and software to connect AI processors across a data center without requiring customers to separately integrate multiple networking systems. Upscale is targeting customers such as neocloud providers and major cloud companies that are expanding data-center capacity while seeking to deploy chips from multiple suppliers.

The approach addresses a practical constraint in the rapidly expanding AI infrastructure market. Data centers increasingly require large numbers of processors to operate together efficiently, but different chip architectures and networking technologies can create integration challenges. By providing a common networking layer, Upscale aims to make heterogeneous AI infrastructure easier to deploy and manage.

Keeping Expensive AI Processors Working

Upscale Chief Executive Barun Kar said the company expects Token Fabric to generate tens of millions of dollars in revenue next year, with the potential to reach the low hundreds of millions of dollars. The company’s strategy is built around improving the efficiency of AI infrastructure by reducing periods in which processors are waiting for data rather than performing computational work.

Token Fabric combines Upscale’s networking chips with equipment using Nvidia’s Spectrum-X technology and software designed to identify bottlenecks and flag equipment that may require replacement. For data-center operators, the potential economic benefit lies in extracting greater utilization from expensive AI processors rather than allowing costly computing capacity to remain underused because of networking constraints.

Funding and Nvidia Support Strengthen Upscale’s Position

Upscale’s strategy is supported by substantial private-market financing. The startup was valued at $2 billion in June, when it raised a $190 million extension to an early-stage funding round. The financing brought its total funding to $500 million and was led by Premji Invest, the investment arm of Indian billionaire Azim Premji.

The company is also backed by Nvidia, Salesforce and Singapore state investor Temasek. Nvidia’s involvement is particularly relevant as the chipmaker continues to build a broader AI infrastructure ecosystem. However, Upscale’s proposition is not limited to Nvidia hardware; its core pitch is enabling customers to connect processors from rival suppliers, potentially supporting a more diversified approach to AI computing.

AI Infrastructure Moves Toward Greater Flexibility

Upscale plans to release the first part of Token Fabric in the fourth quarter, initially connecting groups of AI chips across a data center. The full platform is expected to roll out in stages through 2027, giving the company time to expand its networking capabilities as customer deployments develop.

The success of the platform will depend on whether large cloud operators and neoclouds view multi-chip infrastructure as sufficiently valuable to justify adopting another networking layer. If demand continues to favor diversified AI infrastructure, technologies that improve interoperability and utilization could become increasingly important. Investors will therefore be watching Token Fabric’s commercial adoption, revenue trajectory and ability to support multiple chip ecosystems as Upscale moves from private funding growth toward broader market deployment.