Startup intros new platform for AI inferencing at the edge

AI inferencing at the edge

Vapor IO's Zero Gap AI platform comes as many enterprises continue to show an interest in running AI inferencing at the edge, said Craig Matsumoto, an analyst at S&P Global Market Intelligence's 451 Research group.

"The core idea of people wanting to run inference at the edge -- that's solid. We know that's true," Matsumoto said, adding that enterprises want to run the inferencing in both centralized public clouds and small data centers.

Zero Gap AI delivers 5G and GPU-based micro clouds to specific locations that are near where capabilities are needed, according to Vapor IO.

Examples include retail storefronts, factories and city intersections.

The AI platform consists of fiber optic networks or AI access points and is optimized for running high-performance AI workloads.

It lets users deliver hyperlocal AI services over private wired and wireless networks without running the servers.

"Vapor takes on the responsibility of delivering this capability to you," said Cole Crawford, the vendor's CEO and founder. "You now no longer have to own, depreciate and amortize physical assets. We get to help you with that part of it and you get all of the benefits of having that gear."

Benefit for enterprises

The benefit of Vapor IO's platform is that the Texas-based startup has micro data centers networked together, Matsumoto added.

This lets customers run AI inferencing on a centralized network.

For example, if a retailer wants to run AI inferencing in each of its stores, it won't need a contingent of servers and equipment at each store. It can just put the micro network into the Zero Gap AI environment.

Moreover, since Vapor IO's network is connected to peering points or internet exchanges, if a customer needs to feed its data into a public cloud, it can.

The partnership with Supermicro with the 5G component provides privacy for customers using the platform, Matsumoto said.

"Having this mesh of connections in any given metro -- none of that is exposed to the public internet," he said. "You can exchange data over that network without having to go out to the public internet. There's extra comfort in that extra safety."

The core idea of people wanting to run inference at the edge -- that's solid. We know that's true.

The startup's Zero Gap AI platform is also helpful for enterprises facing budget and talent challenges when running AI inferencing at the edge.

"They have gone through the trouble of setting up the servers, backing up the servers, connecting them into a network [and] supplying the power," Matsumoto said. "All of those things are taken care of for you,"

The cost of using the Zero Gap AI platform depends on the customer's architecture and usage rate, Crawford said.

"We certainly think if you need access to GPUs, the barrier of entry for both access and cost is … better in Zero Gap than it would be if you had to wait for that product," he said.

Challenges for Vapor IO

While Vapor IO has seen initial success and has raised $90 million of venture funding since its founding in 2015, the vendor's challenge is that enterprises have other options to run AI inferencing at the edge. The startup competes against established vendors that do the same thing.

For example, networking vendor Cloudflare has a different architecture than Vapor IO, but provides a similar service for enterprises.

Also, smaller public clouds such as Vultr and OVHcloud also compete with Vapor IO.

"Vapor IO does not have the edge all to themselves," Matsumoto said. "They have a specialty, but they do have to stand out against a lot of other edge competitors. For some customers, that difference will mean a lot. For some customers, they might not. That's going to be part of the challenge for marketing inference at the edge."

Esther Ajao is a TechTarget Editorial news writer covering artificial intelligence software and systems.