GPU EconomyCloud GPU prices, read every hour
Read every hour — latest read 85 providers tracked24,210 prices read in 24 h
H100 index$3.5229 providersWidestV100 32GB19.2×Mover, 7dGeForce RTX 3090+18.3%

Analysis · · figures frozen on that day

B200 rates ran $3.69 to $16.11 a GPU-hour on October 3, 2026

Every figure below is a reading this site published - a rate read on a provider's own page, or a value of the index computed from such rates - listed with its source under the text and frozen as of that day: prices move, and a dated piece says what was true on its date. The live tables are linked beside each figure.

In the read of 05:40 UTC on October 3, 2026, the on-demand B200 rate cards on our pages ran from $3.69 per GPU-hour at Deep Infra to $16.11 at Google Cloud — 4.4× apart across 31 providers, with the median rate card at $6.79. Two of those 31 clouds entered the table the day before: Hostinger and Fireworks AI, both read for the first time on October 2, 2026. Hostinger’s B200 is the seventh lowest of the 31 at $4.50; Fireworks AI’s is the fourth highest at $13.00.

Two first readings on one card, at $4.50 and at $13.00

The gap between the two is $13.00 ÷ $4.50, or 2.9×. Sized as one B200 left running for a month, $4.50 × 730 h is $3,285 and $13.00 × 730 h is $9,490.

The two figures are labelled differently on our side, because the two pages read differently. Hostinger prints all seven of its cards as “From $ x /hour”, so each one carries our advertised from tag: B200 from $4.50, B200 (Dedicated) from $7.08, down to RTX 4090 from $0.42. Of its two B200 products, only the dearer names its sharing — “Dedicated Blackwell B200, non-shared access” — and that word is the whole of what we record in its region field; the page does not say what the cheaper one shares, so neither do we. Fireworks AI prints a flat figure with no “from” at all, in a table its own page heads “On-Demand Pricing”, with a per-minute column beside the per-hour one we read.

The same $8.00 for 80 GB and 141 GB

Fireworks AI’s table prints five cards: H100 80 GB GPU $8.00, H200 141 GB GPU $8.00, B200 180 GB GPU $13.00, B300 288 GB GPU $15.00 and GB300 288 GB GPU $20.00. The first two are the same figure for different memory. At this seller the 80 GB card and the 141 GB card carry one list price; across the rate cards we read on the same day they do not, the median H100 80GB rate card standing at $3.09 over 20 providers and the median H200 rate card at $4.38 over 42. Its GB300 figure is the dearest of the three rate cards that card has: $10.11 at Verda to $20.00 at Fireworks AI, 2.0× apart.

One card where a first reading changed the cheap end

On the same read, Hostinger’s RTX PRO 6000 (Server) at from $0.65 stands as the lowest on-demand rate card our RTX PRO 6000 Blackwell page carries: that page’s spread reads $0.65 (Hostinger) to $5.50 (Microsoft Azure), 8.5× apart across 32 providers. It is a from rate, and the page’s rate table names every seller above it with its own source link and read time.

How to check these figures, and what they are not

Every figure above is listed under this piece with the page it was read from and the time of the read, and each links to a card page that keeps the same figure current: /gpu/b200, /gpu/rtx-pro-6000, /gpu/gb300. The two provider pages print their rates in plain HTML, so the numbers can be found by reading them.

One thing these readings are not: a price change by either seller. October 2 is the day we first read these two pages, not a day either page was rewritten — we hold no earlier reading of them to compare against. Nothing here is forecast and nothing is a recommendation; every figure is a list price read at a stated time, and list prices are not availability.

Every figure in this piece, and where it was read

As published on October 3, 2026. Each price links to the page it was read from; the last column links the page that keeps it current.
FigureAs publishedRead from, whenLive page
Median rate card, B200, on-demand$6.79 per GPU-hour across 31 providerssource
/gpu/b200
Spread between B200 rate cards, on-demand$3.69 (Deep Infra) to $16.11 (Google Cloud), 4.4× apart across 31 providerssource
/gpu/b200
Cheapest B200 rate card, Deep Infra$3.69 per GPU-hour, B200 180GBsource
/providers/deepinfra
Dearest B200 rate card, Google Cloud (us-west1, GPU only)$16.11 per GPU-hour, A4 Nvidia B200 (1 gpu slice)source
/providers/gcp
B200, Hostingerfrom $4.50 per GPU-hour, an advertised from ratesource
/providers/hostinger
B200 (Dedicated), Hostingerfrom $7.08 per GPU-hour, an advertised from ratesource
/providers/hostinger
RTX PRO 6000 (Server), Hostingerfrom $0.65 per GPU-hour, an advertised from ratesource
/providers/hostinger
RTX 4090, Hostingerfrom $0.42 per GPU-hour, an advertised from ratesource
/providers/hostinger
Spread between RTX PRO 6000 Blackwell rate cards, on-demand$0.65 (Hostinger) to $5.50 (Microsoft Azure), 8.5× apart across 32 providerssource
/gpu/rtx-pro-6000
B200 180 GB GPU, Fireworks AI$13.00 per GPU-hoursource
/providers/fireworks
H100 80 GB GPU, Fireworks AI$8.00 per GPU-hoursource
/providers/fireworks
H200 141 GB GPU, Fireworks AI$8.00 per GPU-hoursource
/providers/fireworks
B300 288 GB GPU, Fireworks AI$15.00 per GPU-hoursource
/providers/fireworks
GB300 288 GB GPU, Fireworks AI$20.00 per GPU-hoursource
/providers/fireworks
Spread between GB300 NVL72 rate cards, on-demand$10.11 (Verda) to $20.00 (Fireworks AI), 2.0× apart across 3 providerssource
/gpu/gb300
Median rate card, H100 80GB (form factor not stated), on-demand$3.09 per GPU-hour across 20 providerssource
/gpu/h100
Median rate card, H200, on-demand$4.38 per GPU-hour across 42 providerssource
/gpu/h200

Answers the reader question: shortlist GPU clouds by price before picking one. Written because Hostinger and Fireworks AI were read for the first time on 2026-10-02; their B200 rates entered our table as its seventh lowest and fourth highest. All of it is open data under CC BY 4.0; the numbers here are a snapshot, the index and the card pages are current.

Earlier pieces: — Nebius H100 SXM went $3.85 to $4.50 on October 1, 2026 · — H100 SXM averaged $3.88 an hour in September 2026 · — Four rates for one RTX 5090, $0.50 down to $0.25, 30 September 2026. All of them: Analysis.

Data updated: