Zero Latency launches closed beta of Zerogrid AI grid

Zero Latency has launched the closed beta of Zerogrid, a distributed AI inference grid. The programme includes a select group of Fortune 1000 companies, telecommunications operators and DevOps platforms.

Zerogrid is designed to route AI inference workloads in real time to edge computing capacity based on operational constraints such as latency, data locality and burst demand. It is aimed at organisations that need to run inference closer to where data is generated or where regulatory and geographic limits affect where workloads can be processed.

Zero Latency, formerly known as Hyphastructure, said beta users will get access to a workload and image management dashboard. A command-line interface will be added during the beta and refined in response to user feedback.

Grid model

The system is based on a model more commonly associated with energy networks. Its founders drew on experience in distributed virtual power plants, where many smaller assets are coordinated as a single pool and dispatched according to changing conditions.

In Zerogrid, the same principle is applied to computing infrastructure. Zero Latency owns and operates a network of edge computing clusters across the United States and manages them as a shared pool of capacity rather than as fixed regional deployments.

That means workloads are not assigned to a static location in advance. Instead, they are matched on a day-ahead, real-time or longer-term basis to available compute resources that meet the requirements of each inference request.

The approach reflects a broader shift in the AI market as companies look beyond the large centralised systems built for model training and focus more closely on inference, where speed, location and policy constraints can directly affect performance and compliance.

Zero Latency does not seek to compete with hyperscale cloud providers or newer cloud infrastructure groups in AI training. Instead, it argues that inference presents a different infrastructure problem, especially for customers that cannot rely on regional cloud routing alone or find on-premises deployments too inflexible.

Inference focus

Cloud providers typically offer customers a choice of region, availability zone or local deployment, but they do not usually route individual inference decisions based on a broader mix of latency, data residency and operational limits. On-premises systems can address some of those concerns, but they often require capacity to be provisioned in fixed locations and can be difficult to scale for variable demand.

Zerogrid was built to address that gap by treating inference tasks as routing decisions in their own right. In practice, that means sending a specific workload to the compute environment that best fits its constraints at that moment rather than to a preselected site.

Zero Latency also linked the model to what it described as an AI grid concept, in which compute is treated as a dispatchable service rather than static infrastructure. Zerogrid adds a layer of constraint-aware routing, meaning available capacity alone is not enough and the system must also account for the conditions attached to each workload.

This matters as AI deployment becomes more fragmented across industries and jurisdictions. Companies operating across borders are increasingly dealing with rules on data sovereignty, sector-specific compliance demands and internal governance policies that can limit where AI systems are run and where data can move.

At the same time, demand for inference is expected to rise as more applications use AI models continuously in production rather than in occasional experiments. That shift has increased attention on the economics and logistics of running inference at scale, particularly in sectors where response times and local processing matter.

Zero Latency said its team had previously worked on decentralised infrastructure in energy, including battery, solar, demand response, electric bus, vehicle and distributed natural gas systems. That background informed the company's view that compute, like electricity, can be orchestrated across dispersed assets rather than concentrated in a small number of large sites.

Co-founder Michael Huerta framed the launch in those terms.

"Innovation through decentralization is not a thesis we arrived at recently. It is the lens through which we have built, financed and operated infrastructure for decades. We have applied the successes and hard lessons from deploying decentralized power infrastructure to unlock architectural and routing innovations for AI workloads. Zerogrid is the result: infrastructure designed for an inference world that the cloud was never built to serve," Huerta said.