What happened
Upscale AI launched Fabric, a hardware and software platform designed to connect AI processors from various suppliers in data centers. The system aims to reduce bottlenecks and keep expensive AI chips active rather than idle. The first phase, connecting groups of chips, is scheduled for release in the fourth quarter of 2026, with the full platform rolling out in stages through 2027.
Upscale AI, a startup valued at $2 billion as of June 2026, officially launched its networking platform called Fabric on Thursday, October 8, 2026. The platform is designed to allow customers to build data centers using AI chips from different suppliers without needing to integrate separate, proprietary networking systems for each vendor.
The system combines Upscale’s own networking chips, equipment utilizing Nvidia’s Spectrum-X technology, and software that identifies bottlenecks and flags equipment that may need replacing. The primary goal is to keep expensive AI chips working rather than waiting for data, thereby helping customers get more value from their hardware investments.
Upscale AI is targeting growing demand from neoclouds and large cloud providers that are expanding their data center footprints and seeking to utilize chips from multiple suppliers. CEO Barun Kar stated that the company expects Fabric to generate revenue in the tens of millions of dollars next year, with the potential to reach the low hundreds of millions.
The company plans to release the first part of Fabric, which connects groups of AI chips across a data center, in the fourth quarter of 2026. The full platform is scheduled to roll out in stages through 2027. Upscale AI has raised a total of $500 million in funding, including a $190 million extension to its early-stage round in June 2026 led by Premji Invest, the investment unit of Indian billionaire Azim Premji. Other backers include Nvidia, Salesforce, and Singapore state investor Temasek.
Why it matters
This launch addresses a critical inefficiency in AI infrastructure where data centers often struggle to integrate chips from different vendors, leading to underutilized hardware. By providing a unified networking layer, Upscale AI allows customers like neoclouds and large providers to maximize the return on their significant capital expenditures in AI hardware. This move also signals a shift toward more flexible, multi-vendor data center architectures, potentially reducing dependency on single-supplier ecosystems and lowering the barrier to entry for organizations building diverse AI infrastructure.
The launch of Fabric is significant because it tackles a major operational challenge in AI infrastructure: the complexity and cost of integrating heterogeneous hardware. By abstracting the networking layer, Upscale AI enables data centers to mix and match chips from different vendors, which can lead to better hardware utilization and lower total cost of ownership.
This development is particularly relevant for neoclouds and large cloud providers that are rapidly scaling their AI capabilities. These organizations often face pressure to optimize their capital expenditures, and a platform that ensures AI chips are not idle due to networking bottlenecks directly impacts their bottom line.
The involvement of Nvidia as a backer and the use of its Spectrum-X technology in the platform suggest a strategic alignment that may help Upscale AI gain credibility and adoption in a market dominated by Nvidia's ecosystem. However, the platform's ability to connect rival suppliers could also position it as a neutral arbiter in a multi-vendor environment.
The projected revenue of tens to low hundreds of millions of dollars for next year indicates that Upscale AI is aiming for rapid commercial traction. If successful, this could establish a new standard for data center networking in the AI era, potentially influencing how future data centers are designed and operated.
Interactive Mechanism: How It Actually Works
Explore the underlying technology behind this development interactively.
Which component of an AI application is the machine-learning model itself?
What to watch next
Monitor the actual performance metrics and adoption rates of Fabric as the first phase rolls out in Q4 2026. Watch for announcements regarding which specific chip suppliers are initially supported and whether major cloud providers publicly confirm their use of the platform. Additionally, track Upscale AI's revenue milestones against CEO Barun Kar's projections of tens to low hundreds of millions in revenue for next year.
The first phase of Fabric, scheduled for release in Q4 2026, will be a critical test of the platform's ability to connect groups of AI chips effectively. Early customer feedback and performance data will be key indicators of its success.
Watch for announcements from major cloud providers and neoclouds regarding their adoption of Fabric. Public case studies or partnerships will provide concrete evidence of the platform's real-world impact.
The rollout of the full platform through 2027 will likely involve support for a wider range of chip suppliers. The specific vendors supported and the technical details of the integration will be important for understanding the platform's scope and capabilities.
Upscale AI's financial performance will be closely watched, particularly as it moves from a startup phase to generating significant revenue. The company's ability to meet or exceed its revenue projections will be a key metric for its long-term viability and market position.