Turn your fiber into the on-ramp every GPU cloud is reached through.
Every GPU and neocloud provider your enterprises want, in your market and beyond it, listed on an on-ramp you own and live in 90 days. Both sides billable, at 50 to 70 percent operator margin.
ONE ON-RAMPEVERY PROVIDER ON ITBOTH SIDES BILLABLE
On your fiber. Under your name. Sovereign at home, reaching every GPU site beyond it.
THEIR COMPUTE, THEIR PRICE
THEIR COMPUTE, THEIR PRICE
REACHED FROM THE SAME NODE
The compute is landing.
The road to it is not built.
Capital is going into AI compute faster than any operator has seen it move, and every build lands as its own island. Nobody can reach it privately, so every request becomes a one-off networking project quoted per account, and the capacity sits there waiting.
The on-ramp has to come from the network side, and you are the network side ,and the layer above the fiber already earns $0.70 of advanced services per $1 of connectivity at a major US operator, up from $0.20.
Every build is an island. The on-ramp is what turns them into a market.
Reached over your on-ramp: a GPU cloud in your market, a GPU site in another country, an enterprise data center, a colo cage, a branch site, a public cloud region, whatever lands next.
- 5% Average GPU utilization today
- 70% What a 1,024 GPU H100 cluster needs to break even
- $330K Lost every month it runs 15 points below that
- Weeks Of lead time per private path. On an on-ramp, minutes.
- 111 to 114% Net revenue retention at the platform layer above your pipe
WHAT MAKES THIS DIFFERENT FOR AN OPERATOR
One on-ramp you own. Every provider in your market on it, live in a quarter.
You become the on-ramp every GPU and neocloud provider in your market is reached through, and the marketplace where enterprises find them. It runs on your fiber and under your name. A workload that has to stay in-country stays in-country, and the telemetry proves it. A workload that wants compute in another country reaches it from the same node. Both sides are billable. Three things make that a quarter of work rather than a roadmap item.
One on-ramp. Every provider in your market listed on it.
You stand the on-ramp up once and run it as your marketplace. Every GPU and neocloud provider in your market is listed on it under its own logo, with its own capacity and its own price, and so is the one your enterprises want in the next country over. Any enterprise that lands on your on-ramp reaches any of them in minutes, from your on-ramp, under your name.
Your name is on the access contract with the enterprise and on the listing agreement with the provider. Both sides are billable. Adding the second provider is a listing on a platform that is already up.
Sovereign where it has to be. Reaching wherever the compute is.
Your home market is yours to own. Every path that has to stay in-country runs on fiber you own, hands off inside your borders, answers to the license you already hold, and the telemetry proves it flow by flow. The node is not tied to a GPU site: from London it reaches a cluster in Frankfurt, across your network and the carrier handoffs you already hold, still under your name and still on your bill. A marketplace sitting above your pipe carries the same enterprise off-net, in-country or not, and can only promise best effort.
Sovereignty is the one thing a player above your pipe cannot rent from you. On your own fiber it is yours by construction, and it never caps your reach: the deployment is per market, not per GPU site.
Start with a few sites. Grow when a provider asks.
A high availability on-ramp edge node pair, in facilities you already run, handed off over a standard NNI, and live within 90 days. The deployment is per market, not per GPU site: one pair lists and reaches every provider on your marketplace, in your metro or in another country.
The footprint follows provider demand, so every site you add is one a provider has already asked for. No catalog, no capex program, no rip and replace. What is deliberately out of scope is what keeps the number small.
SCOPE
The whole product, in one table.
Most platform conversations with an operator start with a catalog and end with a program. This one is one service. Here is everything you stand up, and everything that stays exactly as it is.
The right column is the reason this is a quarter. Every line on it is either a product you already sell under your own name, or a program this keeps off your roadmap. The access layer is the one piece the market is short of, so it is the one piece here.
THE TWO SIDES OF YOUR MARKETPLACE
One side brings the compute. The other side brings the workloads.
A marketplace only works if both sides turn up, and both sides have to be better off on yours than anywhere else. Here is what each one gets, and how fast it gets it.
One on-ramp · two sides · both of them billable
SUPPLYGPU and neocloud providers, listed under their own logosONBOARDED IN DAYS
DEMANDEnterprise customers, each one in its own gVPCTURNED UP IN MINUTES
Every GPU and neocloud provider, in your market and beyond it, sits above one on-ramp, and every enterprise customer sits below it. The on-ramp is yours: your marketplace, your name, and a bill on each side, the listing and access on the provider side and the access and capacity on the enterprise side.
Reach that arrives already built
A provider has a few campuses. You have metros, buildings and meet-me rooms. Listing on your on-ramp is the only way they reach an enterprise where it already is, and it costs them no capital to do it. A provider whose campus sits in another country lists the same way and reaches your enterprises the same way.
Enterprise access, fully operated
Their engineers stay on GPUs, power and contracts, which is where they want to be. The private path, the policy, the telemetry and the routing are yours, and they never hire a single person for any of it.
Turned up in minutes
Today every enterprise that wants a private path into them is an LOA, a cross-connect order and weeks of lead time. On your on-ramp it is a turn-up in minutes, at any hour, with no order form.
Neutral by construction
You never rent compute to the enterprises they are chasing, and you never will. For a provider deciding where its private on-ramp lives, that neutrality is the whole decision, and it gets more valuable every quarter.
What the market already prices
Compute discounts. The on-ramp does not ,and on your marketplace, the on-ramp is yours.
Published neocloud price lists, September 2026.
A gVPC per enterprise customer
Where a cloud VPC is private to one region of one cloud, a gVPC is private to one enterprise customer and reaches everywhere that customer works: any cloud, any data center, any site and the internet.
Elastic capacity, billed by the window
Provisioned in minutes, gone at teardown. Hundreds of TB at 100G and 40G bursts, with cheap persistent links in between. The customer pays for the window, not the year.
NEO, so an agent can order it
An enterprise describes two endpoints in plain language and NEO discovers, prices and turns up the path. Everything the portal does is also an API, aligned to MEF LSO Sonata and Cantata and TM Forum TMF622 and TMF641.
See all enterprise features including Security Groups, telemetry and unified routing.
Neither side waits on a build. A listing is configuration rather than construction, and for the enterprise the path is already there.
Every screen, every API response and every invoice on the on-ramp carries your name. The provider's brand stays on its compute, where it belongs. Yours is the only name on the on-ramp.
HOW IT RUNS
An installation, a listing and a license.
Three questions decide whether this is a quarter or a program: what goes in the ground, who runs it, and how it is paid for.
An edge node in a facility you already run
A high availability on-ramp edge node pair, handed off over a standard NNI or a cross-connect. White-box service nodes, commodity hardware, lightweight demarcation at the customer, and no forklift anywhere in your network.
It sits on your network, not over the top of it. The deployment is per market, not per GPU site: one node pair carries every provider on your marketplace, wherever their compute sits, so a provider in another country is a listing, not a new node. Start with the sites your first provider needs, and add sites as providers ask for them.
Yours to run, or ours on your behalf
Run it with your own NOC and your own processes, the way you run the rest of your wholesale book. Or InsidePacket operates it for you end to end: routing, security, observability, provider onboarding, billing records and day-2 support.
Either way, your engineers stay on the network you already run and no new operating model lands on them.
It lands inside the systems you have. The portal, the APIs and the billing records speak MEF LSO Sonata and Cantata and TM Forum TMF622 and TMF641, so your OSS and BSS integrate against standards rather than against us, and nothing here is a proprietary hook you have to unpick later.
Usage-aligned, with the margin on your side
A per site or unlimited platform license on commodity hardware, with operator margin in the range of 50 to 70 percent. Costs scale with the access revenue you bill on both sides, so the on-ramp pays for itself as the marketplace fills.
It is a license, so you keep every dollar you bill on both sides of your marketplace. It runs as operating spend and stays off your capital plan.
Both sides sell through motions you already run. The provider side is wholesale access sold to a new class of customer. On the enterprise side the marketplace does the finding and NEO does the ordering, so demand arrives without a rep in front of it. Scoped to your network in a working session.
You already sell access to competitors, price it, provision it and support it. A wholesale book is already a two-sided one. A marketplace is that same muscle, pointed at both sides of a new market.
THE CASE FOR NOW
Five years, three years, or this quarter.
Building a comparable access layer internally is a standing engineering commitment that does not finish inside the window. Assembling one from vendors trades the build for an integration program. Neither gets you there in time.
TIME TO A COMPARABLE SERVICE IN MARKET
THE WINDOW · THE NEXT 4 TO 8 QUARTERS · AI ON-RAMPS BEING CLAIMED
| THE ACCESS LAYER | AN OVER-THE-TOP PLAYER | YOU, ON THE ON-RAMP |
|---|---|---|
| Whose fiber the path runs on | Yours, rented from you | Yours |
| Whose brand the enterprise sees | Theirs | Yours, on the on-ramp |
| Where the access margin lands | With them | With you, on both sides |
| Sovereign, in-country routing when the workload needs it | Best effort, off-net | On-net, and provable |
| Compute outside your market | Reached through them, at their price | Reached from your node, on your bill |
| Who owns the customer relationship | They do, and they keep it | You own the access, they own the compute |
| Time to a service in market | Theirs is already live | 90 days, on your network |
THE MARKET YOUR ON-RAMP WOULD SIT IN FRONT OF
NEOCLOUD REVENUE, 2025 TO 2031
1 Synergy Research, April 2026: neocloud revenue above $25B in 2025, heading toward roughly $400B by 2031 at a 58% compound annual growth rate. The curve is that published rate plotted between those two published endpoints, not a forecast of ours. The window is a judgement about how these providers buy, not a projection.
2 Gartner, June 2026. 3 Gartner, enterprises on a multicloud networking platform by 2027, up from 10% in early 2024. 4 The Business Research Company, July 2026. Three separate populations, shown together only to size the same opportunity from three directions.
Bar length is time to a comparable service in market for those building it. Your deployment is an edge node in a facility you already run, not a build. Build and assemble durations are planning assumptions, not measured outcomes.
THE STRATEGIC FRAME
BLOGWho owns the on-ramp to intelligence?
Eli Karpilovski, Founder and CEO
Public clouds did not win enterprise workloads on processors alone. They won on reach and operability: private connectivity in major metros, a consistent interface and a common way to consume networking, storage and identity. When AWS launched Direct Connect, it looked like a useful feature. It became part of the moat.
The same shift is beginning around AI infrastructure. Capacity now lives across hyperscale regions, GPU clouds, colocation halls and private clusters, and demand arrives in bursts.
The question behind this product is the strategic one: who owns the on-ramp? Providers that cannot expose their own and their partners' assets through a common, programmable model risk becoming interchangeable suppliers beneath someone else's customer relationship.