macrostack

neocloud · vm · Layer 2.5

Lightning AI

Teams who want a notebook-style development environment, training jobs and multi-node clusters on one bill, from one GPU to a reserved cluster.

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

What you are buying

Business model
Purpose-built GPU cloud. Owns or leases its own datacentre capacity and sells it directly.
How you reach it
You get virtual machines. The familiar cloud model.
Accelerators
NVIDIA T4, L4, L40S, RTX PRO 6000, A100, H100, H200 and B200 on its price page, plus TPUs and CPUs. The merger brought Voltage Park's owned H100, B200 and GB300 fleet.
Regions
No region list. Its FAQ says it runs across more than six clouds, and it can also run inside your own AWS or GCP account.
Pricing model
Credits, billed by the second, at published per-GPU hourly rates. A Free plan, Pro and Teams plans on monthly or annual billing, and Enterprise by contract; interruptible capacity at up to 80% off.
Getting started
Self-serve. Sign-up gives 5 free credits and adding a card gives 25 more; a phone number is required to prevent abuse, and purchased credits expire after 12 months (its FAQ, 17 Nov 2026).
Capacity
More than 36,000 owned and operated H100, B200 and GB300 GPUs from Voltage Park, by the January 2026 merger announcement, which also counts 400,000+ users.
Ownership
Private. Lightning AI and Voltage Park completed their merger in January 2026 and the combined company is named Lightning AI; chief executive William Falcon, president Ozan Kaya, formerly Voltage Park's chief executive (Business Wire, 21 Jan 2026).

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

The counterparty

A developer platform that now owns a datacentre fleet. You sign with one company for both the tools and the hardware; the merger announcement says contracts signed with Voltage Park carry on unchanged.

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

The Studio environment, the plans and the credit balance are Lightning's own; the PyTorch inside them moves, the setup around it does not. Interruptible capacity is the cheap way in, and it can be taken back mid-run.

The economics

Plans bundle a monthly credit allowance with seats and features, and GPU time is drawn from credits at published hourly rates. Since the merger it sells capacity it owns beside capacity from other clouds.

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

Read next

  • Voltage Park — merged into Lightning AI in January 2026, and brought the GPU fleet with it

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.