TL;DR · 30-second read
The Short Version
Businesses that want to use artificial intelligence increasingly rent the powerful chips it runs on from specialist providers. A California software company, insidepacket, argues the hard part is often not getting the chips but building a private, secure connection to wherever those chips sit.
With European network operator Zayo Europe, it plans to sell that connection as something switched on in minutes, more like adding a phone line than running a months-long construction project.
The bet: whoever owns those connections, not just the chips, earns a real share of a market forecast to reach about $400 billion a year by 2031.
insidepacket, a Palo Alto-based network software company, announced on October 12, 2026 its AI Access Platform, a white-label service that connects enterprises to participating GPU and neocloud providers through private network paths the company says can be activated in minutes. The platform is planned for introduction in Europe, first with Zayo Europe, whose DynamicLink network-as-a-service (NaaS) offering already runs on insidepacket’s software.
It comes in two models: a dedicated on-ramp that a single GPU cloud sells under its own brand, or an operator layer that presents many GPU clouds to an operator’s enterprise customers as a marketplace. insidepacket says its underlying NaaS platform has been in production in the US since September 2025 and in Europe since mid-2026, and that a defined initial-site deployment can be completed in as little as 90 days, subject to scope and integration requirements.
Executive Summary
The AI Access Platform is a connectivity product, not a compute product. insidepacket sells no GPUs; instead it supplies the software that lets a network operator or GPU provider offer each enterprise its own private network reaching GPU clouds, data centers, offices and the internet, with built-in security, traffic segmentation and bandwidth that can be scaled up or down on each link. The operator or provider keeps the brand, the customer relationship and the margin.
The announcement matters because it stakes out a specific thesis about where enterprise AI adoption gets stuck. Citing Synergy Research Group figures showing neocloud revenue above $25 billion in 2025 and forecast to approach $400 billion by 2031, insidepacket argues that GPU capacity is landing as isolated “islands” and that the connectivity around it is still treated as a separate, slow engineering project. If that is right, the companies that control the network path to compute gain leverage over a fast-growing market.
Europe is the proving ground, and data sovereignty is the pitch: the platform supports in-country, auditable network paths so an enterprise’s AI traffic can stay inside national borders and on the operator’s own network.
Why the On-Ramp, Not the GPU, Is the Bet
The headline number frames the stakes. Synergy Research Group’s estimate puts neocloud revenue (the GPU-specialist cloud providers that rent AI compute outside the big three public clouds) at more than $25 billion in 2025, with a forecast approaching $400 billion by 2031, roughly a sixteenfold increase. insidepacket’s argument is that this growth is being built as scattered pockets of capacity, often sited wherever power could be found, and that enterprises reaching that capacity must still commission a private connection as a standalone project. AvidThink’s Roy Chua made the power link explicit in the announcement: constrained power availability is fragmenting compute, which in turn makes flexible connectivity more valuable.
The mechanism insidepacket proposes is familiar from the last cloud cycle. “Public clouds did not win enterprise workloads on processors. They won on reach,” Karpilovski said. Translated into operational terms: an enterprise choosing a GPU provider weighs not only price per hour but how quickly, securely and compliantly it can move data there. If connecting to a new neocloud means weeks of circuit ordering, firewall design and contract negotiation, the enterprise either stays put or does not start. A pre-integrated, on-demand path removes that friction, and the party offering it sits between demand and supply.
Who is affected follows directly. Neoclouds with underused capacity stand to gain if easier access improves utilization, a benefit insidepacket explicitly claims. Enterprises gain optionality across providers. And network operators, long cast as suppliers of a commodity pipe, get a way to sell a higher-value service. That is the bet; whether connectivity rather than GPU price or availability is the binding constraint for most enterprise AI buyers remains an assertion the companies have not backed with adoption data.
The Real Contest Is Over Who Owns the Customer
insidepacket notes that in the past six months platform players and Tier-1 operators (the largest carriers with their own global backbones) have moved to close the connectivity gap, “with implications for brand, margin and customer ownership across the value chain.” That is the commercial core of the announcement. Whoever presents the menu of GPU options to an enterprise is positioned to shape which providers win and to take a cut of each transaction.
insidepacket’s answer is neutrality: it supplies the platform, while the operator or GPU provider sells under its own name. The on-net, white-label model lets an operator bill both sides, charging enterprises for access and providers for distribution, while managed service providers and integrators can package it into broader offerings. “For an operator, that means taking a position in AI instead of supplying the pipe,” Karpilovski said. It is a coherent pitch to carriers worried about being disintermediated, though it also depends on operators wanting to run a marketplace and on GPU providers accepting an intermediary between themselves and their customers.
Sovereignty Gives Europe a Reason to Go First
Launching in Europe is not incidental. insidepacket describes data sovereignty, the requirement that sensitive data stay within a given country’s legal jurisdiction, as a leading criterion in European AI buying. An on-ramp that can keep traffic in-country and on a single operator’s own fibre, with auditable paths, addresses a compliance question that a generic internet connection to a GPU cloud cannot answer as cleanly.
Zayo Europe’s role strengthens the case. Its Chief Product Officer, Michael Katz, pointed to the company’s “fully owned European infrastructure” and to DynamicLink’s track record of scaling high-capacity connectivity “in minutes, not months.” Owning the network end to end is what makes an in-country, on-net guarantee credible; an operator that leased segments from others would find that promise harder to keep.
Minutes for Paths, 90 Days for Sites
Two timeframes appear in the announcement and they describe different things. “Minutes” refers to activating a private path to a participating GPU provider once an enterprise is already connected to the operator’s network. The “as little as 90 days” figure covers a defined initial-site deployment and is explicitly conditional on scope, network integration and deployment requirements. Buyers should read the speed claim as applying after onboarding, not from a standing start.
What is substantiated is the foundation: insidepacket’s NaaS software already powers Zayo’s DynamicLink, has been in US production since September 2025, and earned a 2024 MEF NaaS Excellence award for service automation. What is not yet demonstrated is the AI-specific layer. The platform is described as planned rather than live, and no participating GPU or neocloud provider is named.
Background
insidepacket is a Palo Alto-headquartered company whose cloud-native NaaS software lets network operators, managed service providers, integrators and GPU providers launch connectivity services on infrastructure they already own. Its best-known deployment is Zayo’s DynamicLink, a self-service, high-capacity connectivity offering. The company won Most Innovative Service Automation Vendor at the 2024 MEF NaaS Excellence Awards and is a finalist in two categories at the 2026 Global Connectivity Awards.
The broader market context is the rapid rise of neoclouds, GPU-focused cloud providers that rent AI compute to enterprises and model developers. As this capacity has spread across many providers and locations, often shaped by where power is available, the question of how enterprises connect to it privately, securely and in compliance with data-residency rules has become a contested layer of the AI infrastructure stack, with carriers, platform companies and software vendors all vying to own it. Source: insidepacket Launches the AI Access Platform: One On-Ramp Between Enterprises and GPU Clouds, insidepacket’s announcement of a white-label GPU cloud on-ramp launching first with Zayo Europe.Sources