macrostack

aggregator · container · Layer 2.5

Northflank

Teams who want to ship containers, jobs and databases to GPUs without running Kubernetes themselves, on their own cloud account or on Northflank's.

medium portability Movable with effort. Expect to redo images, storage wiring and networking.

What you are buying

Business model
Sells access across other people's clouds through one contract and one API.
How you reach it
You push a container image and it runs. No cluster to operate.
Accelerators
NVIDIA L4, A100 (40GB and 80GB), H100, H200, RTX PRO 6000 and B200 on Northflank's own cloud. Its GPU page lists about 25 models across connected clouds, AMD MI300X and MI355X among them.
Regions
17 managed regions on its own cloud (its docs, 17 Nov 2026). Bring your own account and it runs wherever AWS, GCP, Azure, Oracle, CoreWeave, Nebius or Civo do.
Pricing model
Billed by the second. A free Sandbox; pay as you go per vCPU-hour and per GB-hour of memory; per-hour GPUs on its own cloud; and, on your own cloud, a fee per GB of GPU memory per month on top of that cloud's GPU bill. Enterprise by contract.
Getting started
Self-serve. The Sandbox needs no card: 2 services, 2 jobs, 1 database and one bring-your-own-cloud cluster of up to 12 vCPU and 48GB.
Capacity
Its own cloud does not name the companies whose machines it runs on. H100, H200 and B200 come only as 8-GPU nodes.
Ownership
Private, London. $22.3M raised: a $16M Series A led by Bain Capital Ventures and a $6.3M seed led by Vertex Ventures US (Business Wire, November 2024).

Verified 2026-11-17. Fields reading “not published” are exactly that — we do not estimate a figure a vendor withholds.

The counterparty

A small, venture-backed platform company between you and the hardware. Bring-your-own-cloud is the hedge: the GPUs then sit in your own AWS, GCP, Azure, Oracle, CoreWeave or Nebius account, so the hardware relationship outlives the platform.

A multi-year GPU commitment is a credit decision wearing a cloud contract. This is the section no benchmark covers and the one that decides what happens to your workload in 2028.

The catch

On its own cloud the biggest chips come only as 8-GPU nodes, and it does not say whose datacentres they sit in. On your own cloud the platform fee is added to that cloud's GPU bill.

The economics

The business is the platform layer, not the silicon. Its own GPU rates and its bring-your-own-cloud fee sit on one page, so both routes can be priced before you choose.

No rate is quoted on this page on purpose. Published list prices at this layer move weekly and essentially nobody signing a real contract pays them. Dated, sourced figures live in the GPU cost calculator, where the cheapest and dearest published on-demand price for the identical H100 sit 5.6× apart.

The layers underneath this one

Whatever you rent here is a chip in a building that needs power. In 2026 megawatts, not silicon, are the binding constraint on the whole industry — a frontier rack draws 120–200 kW against a 2026 average near 27 kW, and the US interconnection queue exceeds 2,600 GW.

Layer 2 — Silicon · Layer 1 — Energy · The interconnection queue · Tokens per watt

Related providers

Sources

Macrostack does not benchmark clusters and does not pretend to. For hands-on performance ratings of GPU clouds, SemiAnalysis’s ClusterMAX tests the hardware directly. This page covers what their scorecard does not: who owns the company and how you get out.

The Macrostack brief

New swaps, worth your inbox.

A short, occasional email when we add a high-intent alternative or ship a new head-to-head. No spam, no selling your address — unsubscribe in one click.