We first read Yotta Labs’ GPU price table at 16:39 UTC on October 28, 2026. In our read of 05:39 UTC on October 29, its A100 80GB stood at $0.92 per GPU-hour on demand, the lowest of the 66 A100 80GB rate cards on the site. Until that first read the lowest was CloudRift at $1.05, which is now second. The A100 80GB page printed a median rate card of $1.92 over those 66 providers.
What was read
Yotta Labs publishes one table on its pricing page: a row per card with memory, RAM, vCPU, an on-demand column and a spot column. We read ten cards from it on demand, each priced per GPU, with no region named on the page. The A100 row is labelled “A100 PCIe 80G”, while CloudRift’s row is labelled “A100 SXM4”: both are 80 GB A100 cards, and the site counts them as one card, but the two pages name different form factors. The H100 row names no form factor, so it is counted on the H100 page that holds such rows, not on the H100 SXM page.
This is a first read, not a price change. Yotta Labs did not cut anything on October 28; that is the day our parser began reading its page, and every figure here is from reads since then.
Where its other rates stood
On the H200, Yotta Labs read $2.50 per GPU-hour, third of 58 rate cards. Beam’s advertised-from rate of $2.09 was the lowest, and the H200 page printed a median rate card of $4.52. On the same page Yotta Labs’ H200 row was $0.06 below its own H100 row: $2.56 - $2.50 = $0.06. That H100 rate was fifth of 29 rate cards, against a median of $3.80.
On the Blackwell cards it sat nearer the middle. The B200 read $5.37, seventh of 40, against a median rate card of $6.94. The B300 read $7.64, eleventh of 27, against a median of $8.96. Over a 730-hour month, one A100 80GB at Yotta Labs comes to $0.92 × 730 = $671.60, against $1.05 × 730 = $766.50 at CloudRift.
The spot column is read too, but it is printed without a dollar sign and in uneven figures, and on the A100 PCIe row the spot cell was above the on-demand one; our parser does not read a spot cell that is not below its row’s on-demand rate. This piece names only on-demand rates. Nothing here is forecast: the figures are what the page printed on the day we first read it.
How to check these numbers yourself
Each Yotta Labs figure links its pricing page at yottalabs.ai/pricing, where the table prints the on-demand rate per card. CloudRift’s rate links its own pricing page, and Beam’s links beam.cloud/pricing. The medians and ranks are the ones /gpu/a100-80, /gpu/h200, /gpu/h100, /gpu/b200 and /gpu/b300 print over each card’s rate cards, with community capacity and marketplace floors left out. Every figure above is listed under this piece with its source and read time, frozen on October 29, 2026; the Yotta Labs provider page keeps the current reading.