Questions? Open a ticket from your console — an engineer replies within one business day.
GPU Cloud HQDedicated, billed monthly
vs
Thunder ComputeLow-cost GPU virtualization
A low-cost GPU cloud built on its own GPU virtualization stack, with H100, A100 and L40 instances in North America billed by the minute. Below: their published rates read on 15 October 2026, what they add up to over a full month, and the cases where Thunder Compute is simply the better choice.
Every model Thunder Compute publishes that we also rent. Their rate is the standard on-demand tier — spot and preemptible tiers are cheaper and noted where they exist.
| GPU | Thunder Compute published rate | One month, 24/7 | Our monthly | Difference |
|---|---|---|---|---|
| NVIDIA H100 SXM80 GB HBM3 | $3.20/h (SXM), 1–8 GPUs | $2,336 | $1,249 | −47% |
| NVIDIA A10080 GB HBM2e | $1.09/h at 1–2 GPUs · $1.49 per GPU at 4–8 | $796 | $649 | −18% |
Our price includes the dedicated CPU, RAM and NVMe listed on the pricing table, 20 TB of outbound transfer and a 99.9% SLA. Read on 15 October 2026 from thundercompute.com.
Both of these lists are real. Pick the column that describes your workload.
Choose Thunder Compute if you need cheap A100 or H100 hours in North America for development and experiments, by the minute.
Choose us if the GPUs stay busy, you want the bill to be predictable, and you would rather not hand over an identity document to rent a computer.
| What to check | Thunder Compute | GPU Cloud HQ |
|---|---|---|
| How you are billed | Per minute while running | Flat monthly price, prepaid balance, invoices settle themselves |
| Commitment | Reserved capacity for 8 to 512+ GPUs through sales | Rolling month, or 3/6/12 months for 5–15% off |
| Signing up & identity | Email, then a payment method | Self-serve, email address only — no KYC |
| How you pay | Saved card or preloaded, non-refundable credit | Bitcoin, Ether, USDT (ERC-20 / TRC-20) — no KYC, no card |
| What you actually get | Virtual machines on its own GPU virtualization stack, never preempted | Single-tenant bare metal, root access |
| Storage | First 100 GB per GPU included, then $0.219/GB/month | Local NVMe included; block volumes at $25/TB/mo |
| Network | No egress fees | 10–25 Gbps dedicated, 20 TB egress included |
| Where it runs | North America | 5 Tier III sites in North America and Europe |
For a machine that stays busy, yes: Thunder Compute publishes $3.20/h (SXM), 1–8 GPUs, which is $2,336 for a 730-hour month, against $1,249 flat here for a dedicated H100 — about 47% less. Below roughly 390 hours of use a month, their hourly rate works out cheaper.
You need cheap A100 or H100 hours in North America for development and experiments, by the minute. If that is you, stay with them — we would rather say so than sell you the wrong thing.
No. Every price on this page is public on both sides. When you do order here, an email address is all we ask for: no ID document, no company registration, no card — the balance is funded in Bitcoin, Ether or USDT.
They were read on 15 October 2026 from Thunder Compute's own pricing page (https://www.thundercompute.com/pricing). Providers change prices without notice — the link is there so you can check today's figure yourself.
Every GPU is in stock, the price is public, and the first invoice settles from your balance in minutes.
