# Compute Cost Index ## The same chip rents for 5.4 times as much from one provider as from another Published on-demand rate, $ per petaFLOP-day (log scale) →Price verified2026-09-042026-04-192026-03-29$100$150$200$300$500$100$150$200$300$500B2003 offerings · 3 providersB200 · median published on-demand rate $91.73 per petaFLOP-day across 3 offeringsLambda · B200 · US (Texas) · $74.56 per petaFLOP-day ($6.99/hour, single-GPU instance) · verified 2026-04-19CoreWeave · B200 · US East (N. Virginia) · $91.73 per petaFLOP-day ($68.80/hour, 8-GPU node) · verified 2026-04-19Oracle Cloud · B200 · Global · $149.33 per petaFLOP-day ($112.00/hour, 8-GPU node) · verified 2026-09-042.0× spreadL40S1 offering · 1 providerL40S · median published on-demand rate $116.01 per petaFLOP-day across 1 offeringOracle Cloud · L40S · Global · $116.01 per petaFLOP-day ($14.00/hour, 8-GPU node) · verified 2026-09-04MI300X17 offerings · 1 providerMI300X · median published on-demand rate $147.81 per petaFLOP-day across 17 offeringsAzure · MI300X · East US 2 · $110.14 per petaFLOP-day ($48.00/hour, 8-GPU node) · verified 2026-09-04Azure · MI300X · West US 3 · $110.14 per petaFLOP-day ($48.00/hour, 8-GPU node) · verified 2026-09-04Azure · MI300X · Canada Central · $132.17 per petaFLOP-day ($57.60/hour, 8-GPU node) · verified 2026-09-04Azure · MI300X · North Europe · $132.17 per petaFLOP-day ($57.60/hour, 8-GPU node) · verified 2026-09-04Azure · MI300X · Central US · $135.48 per petaFLOP-day ($59.04/hour, 8-GPU node) · verified 2026-09-04Azure · MI300X · Korea South · $137.68 per petaFLOP-day ($60.00/hour, 8-GPU node) · verified 2026-09-04Azure · MI300X · North Central US · $142.30 per petaFLOP-day ($62.02/hour, 8-GPU node) · verified 2026-09-04Azure · MI300X · West Central US · $142.30 per petaFLOP-day ($62.02/hour, 8-GPU node) · verified 2026-09-04Azure · MI300X · France Central · $147.81 per petaFLOP-day ($64.42/hour, 8-GPU node) · verified 2026-09-04Azure · MI300X · UK South · $147.81 per petaFLOP-day ($64.42/hour, 8-GPU node) · verified 2026-09-04Azure · MI300X · Germany West Central · $153.32 per petaFLOP-day ($66.82/hour, 8-GPU node) · verified 2026-09-04Azure · MI300X · Sweden Central · $154.20 per petaFLOP-day ($67.20/hour, 8-GPU node) · verified 2026-09-04Azure · MI300X · West US · $154.36 per petaFLOP-day ($67.27/hour, 8-GPU node) · verified 2026-09-04Azure · MI300X · UAE North · $157.50 per petaFLOP-day ($68.64/hour, 8-GPU node) · verified 2026-09-04Azure · MI300X · Australia Southeast · $164.33 per petaFLOP-day ($71.62/hour, 8-GPU node) · verified 2026-09-04Azure · MI300X · Poland Central · $169.62 per petaFLOP-day ($73.92/hour, 8-GPU node) · verified 2026-09-04Azure · MI300X · Brazil South · $220.28 per petaFLOP-day ($96.00/hour, 8-GPU node) · verified 2026-09-042.0× spreadGB2001 offering · 1 providerGB200 · median published on-demand rate $153.60 per petaFLOP-day across 1 offeringOracle Cloud · GB200 · Global · $153.60 per petaFLOP-day ($128.00/hour, 8-GPU node) · verified 2026-09-04H200 SXM36 offerings · 4 providersH200 SXM · median published on-demand rate $321.91 per petaFLOP-day across 36 offeringsCoreWeave · H200 SXM · US East (N. Virginia) · $152.93 per petaFLOP-day ($50.44/hour, 8-GPU node) · verified 2026-04-19AWS · H200 SXM · US East (N. Virginia) · $191.90 per petaFLOP-day ($63.30/hour, 8-GPU node) · verified 2026-09-04AWS · H200 SXM · US West (Oregon) · $191.90 per petaFLOP-day ($63.30/hour, 8-GPU node) · verified 2026-09-04AWS · H200 SXM · Asia Pacific (Tokyo) · $239.88 per petaFLOP-day ($79.12/hour, 8-GPU node) · verified 2026-09-04Oracle Cloud · H200 SXM · Global · $242.55 per petaFLOP-day ($80.00/hour, 8-GPU node) · verified 2026-09-04Azure · H200 SXM · East US 2 · $257.10 per petaFLOP-day ($84.80/hour, 8-GPU node) · verified 2026-09-04Azure · H200 SXM · West US 3 · $257.10 per petaFLOP-day ($84.80/hour, 8-GPU node) · verified 2026-09-04Azure · H200 SXM · West US 2 · $270.49 per petaFLOP-day ($89.22/hour, 8-GPU node) · verified 2026-09-04Azure · H200 SXM · Mexico Central · $282.81 per petaFLOP-day ($93.28/hour, 8-GPU node) · verified 2026-09-04Azure · H200 SXM · South Central US 2 · $282.81 per petaFLOP-day ($93.28/hour, 8-GPU node) · verified 2026-09-04Azure · H200 SXM · Central US · $290.52 per petaFLOP-day ($95.82/hour, 8-GPU node) · verified 2026-09-04Azure · H200 SXM · Indonesia Central · $300.81 per petaFLOP-day ($99.22/hour, 8-GPU node) · verified 2026-09-04Azure · H200 SXM · Malaysia West · $300.81 per petaFLOP-day ($99.22/hour, 8-GPU node) · verified 2026-09-04Azure · H200 SXM · France Central · $321.37 per petaFLOP-day ($106.00/hour, 8-GPU node) · verified 2026-09-04Azure · H200 SXM · Japan East · $321.37 per petaFLOP-day ($106.00/hour, 8-GPU node) · verified 2026-09-04Azure · H200 SXM · UK South · $321.37 per petaFLOP-day ($106.00/hour, 8-GPU node) · verified 2026-09-04Azure · H200 SXM · US Gov Arizona · $321.37 per petaFLOP-day ($106.00/hour, 8-GPU node) · verified 2026-09-04Azure · H200 SXM · North Central US · $321.91 per petaFLOP-day ($106.18/hour, 8-GPU node) · verified 2026-09-04Azure · H200 SXM · North Europe · $321.91 per petaFLOP-day ($106.18/hour, 8-GPU node) · verified 2026-09-04Azure · H200 SXM · Australia Southeast · $334.23 per petaFLOP-day ($110.24/hour, 8-GPU node) · verified 2026-09-04Azure · H200 SXM · Korea South · $334.76 per petaFLOP-day ($110.42/hour, 8-GPU node) · verified 2026-09-04Azure · H200 SXM · US Gov Virginia · $334.76 per petaFLOP-day ($110.42/hour, 8-GPU node) · verified 2026-09-04Azure · H200 SXM · Germany West Central · $347.62 per petaFLOP-day ($114.66/hour, 8-GPU node) · verified 2026-09-04Azure · H200 SXM · Italy North · $347.62 per petaFLOP-day ($114.66/hour, 8-GPU node) · verified 2026-09-04Azure · H200 SXM · Southeast Asia · $347.62 per petaFLOP-day ($114.66/hour, 8-GPU node) · verified 2026-09-04Azure · H200 SXM · Spain Central · $347.62 per petaFLOP-day ($114.66/hour, 8-GPU node) · verified 2026-09-04Azure · H200 SXM · West US · $347.62 per petaFLOP-day ($114.66/hour, 8-GPU node) · verified 2026-09-04Azure · H200 SXM · Sweden Central · $347.72 per petaFLOP-day ($114.69/hour, 8-GPU node) · verified 2026-09-04Azure · H200 SXM · UK West · $350.19 per petaFLOP-day ($115.50/hour, 8-GPU node) · verified 2026-09-04Azure · H200 SXM · UAE North · $367.65 per petaFLOP-day ($121.26/hour, 8-GPU node) · verified 2026-09-04Azure · H200 SXM · Norway East · $381.04 per petaFLOP-day ($125.68/hour, 8-GPU node) · verified 2026-09-04Azure · H200 SXM · Australia East · $386.18 per petaFLOP-day ($127.38/hour, 8-GPU node) · verified 2026-09-04Azure · H200 SXM · South Africa North · $388.75 per petaFLOP-day ($128.22/hour, 8-GPU node) · verified 2026-09-04Azure · H200 SXM · Poland Central · $395.93 per petaFLOP-day ($130.59/hour, 8-GPU node) · verified 2026-09-04Azure · H200 SXM · Japan West · $411.89 per petaFLOP-day ($135.86/hour, 8-GPU node) · verified 2026-09-04Azure · H200 SXM · Brazil South · $527.59 per petaFLOP-day ($174.02/hour, 8-GPU node) · verified 2026-09-043.4× spreadH100 SXM31 offerings · 6 providersH100 SXM · median published on-demand rate $371.10 per petaFLOP-day across 31 offeringsLambda · H100 SXM · US (Texas) · $104.05 per petaFLOP-day ($4.29/hour, single-GPU instance) · verified 2026-04-19CoreWeave · H100 SXM · US East (N. Virginia) · $149.29 per petaFLOP-day ($49.24/hour, 8-GPU node) · verified 2026-03-29AWS · H100 SXM · US East (N. Virginia) · $166.87 per petaFLOP-day ($55.04/hour, 8-GPU node) · verified 2026-09-04AWS · H100 SXM · US West (Oregon) · $166.87 per petaFLOP-day ($55.04/hour, 8-GPU node) · verified 2026-09-04AWS · H100 SXM · Asia Pacific (Tokyo) · $208.59 per petaFLOP-day ($68.80/hour, 8-GPU node) · verified 2026-09-04Oracle Cloud · H100 SXM · Global · $242.55 per petaFLOP-day ($80.00/hour, 8-GPU node) · verified 2026-09-04GCP · H100 SXM · US Central (Iowa) · $268.29 per petaFLOP-day ($88.49/hour, 8-GPU node) · verified 2026-04-19Azure · H100 SXM · East US · $311.48 per petaFLOP-day ($102.74/hour, 8-GPU node) · verified 2026-09-04Azure · H100 SXM · East US 2 · $311.48 per petaFLOP-day ($102.74/hour, 8-GPU node) · verified 2026-09-04Azure · H100 SXM · West US 2 · $311.48 per petaFLOP-day ($102.74/hour, 8-GPU node) · verified 2026-09-04Azure · H100 SXM · West US 3 · $311.48 per petaFLOP-day ($102.74/hour, 8-GPU node) · verified 2026-09-04Azure · H100 SXM · South Central US · $357.66 per petaFLOP-day ($117.97/hour, 8-GPU node) · verified 2026-09-04Azure · H100 SXM · West Central US · $357.71 per petaFLOP-day ($117.98/hour, 8-GPU node) · verified 2026-09-04Azure · H100 SXM · Central US · $366.65 per petaFLOP-day ($120.93/hour, 8-GPU node) · verified 2026-09-04Azure · H100 SXM · Canada Central · $371.10 per petaFLOP-day ($122.40/hour, 8-GPU node) · verified 2026-09-04Azure · H100 SXM · North Central US · $371.10 per petaFLOP-day ($122.40/hour, 8-GPU node) · verified 2026-09-04Azure · H100 SXM · UK South · $372.61 per petaFLOP-day ($122.90/hour, 8-GPU node) · verified 2026-09-04Azure · H100 SXM · US Gov Arizona · $372.61 per petaFLOP-day ($122.90/hour, 8-GPU node) · verified 2026-09-04Azure · H100 SXM · Poland Central · $387.52 per petaFLOP-day ($127.82/hour, 8-GPU node) · verified 2026-09-04Azure · H100 SXM · Spain Central · $387.52 per petaFLOP-day ($127.82/hour, 8-GPU node) · verified 2026-09-04Azure · H100 SXM · Sweden Central · $387.52 per petaFLOP-day ($127.82/hour, 8-GPU node) · verified 2026-09-04Azure · H100 SXM · West Europe · $387.52 per petaFLOP-day ($127.82/hour, 8-GPU node) · verified 2026-09-04Azure · H100 SXM · West US · $387.52 per petaFLOP-day ($127.82/hour, 8-GPU node) · verified 2026-09-04Azure · H100 SXM · Korea Central · $415.81 per petaFLOP-day ($137.15/hour, 8-GPU node) · verified 2026-09-04Azure · H100 SXM · UAE North · $426.27 per petaFLOP-day ($140.60/hour, 8-GPU node) · verified 2026-09-04Azure · H100 SXM · Japan East · $432.23 per petaFLOP-day ($142.56/hour, 8-GPU node) · verified 2026-09-04Azure · H100 SXM · Australia East · $432.34 per petaFLOP-day ($142.60/hour, 8-GPU node) · verified 2026-09-04Azure · H100 SXM · Switzerland North · $439.66 per petaFLOP-day ($145.01/hour, 8-GPU node) · verified 2026-09-04Azure · H100 SXM · Japan West · $475.43 per petaFLOP-day ($156.81/hour, 8-GPU node) · verified 2026-09-04Azure · H100 SXM · South Africa North · $524.77 per petaFLOP-day ($173.09/hour, 8-GPU node) · verified 2026-09-04Azure · H100 SXM · South Africa West · $565.76 per petaFLOP-day ($186.61/hour, 8-GPU node) · verified 2026-09-045.4× spreadA100 SXM37 offerings · 6 providersA100 SXM · median published on-demand rate $409.63 per petaFLOP-day across 37 offeringsCoreWeave · A100 SXM · US East (N. Virginia) · $207.69 per petaFLOP-day ($21.60/hour, 8-GPU node) · verified 2026-03-29Lambda · A100 SXM · US (Texas) · $214.62 per petaFLOP-day ($2.79/hour, single-GPU instance) · verified 2026-04-19AWS · A100 SXM · US East (N. Virginia) · $263.91 per petaFLOP-day ($27.45/hour, 8-GPU node) · verified 2026-09-04AWS · A100 SXM · US West (Oregon) · $263.91 per petaFLOP-day ($27.45/hour, 8-GPU node) · verified 2026-09-04GCP · A100 SXM · US Central (Iowa) · $282.60 per petaFLOP-day ($29.39/hour, 8-GPU node) · verified 2026-03-29Oracle Cloud · A100 SXM · Global · $307.69 per petaFLOP-day ($32.00/hour, 8-GPU node) · verified 2026-09-04Azure · A100 SXM · East US · $315.10 per petaFLOP-day ($32.77/hour, 8-GPU node) · verified 2026-09-04Azure · A100 SXM · West US 2 · $315.10 per petaFLOP-day ($32.77/hour, 8-GPU node) · verified 2026-09-04AWS · A100 SXM · EU West (Ireland) · $340.35 per petaFLOP-day ($35.40/hour, 8-GPU node) · verified 2026-04-19Azure · A100 SXM · East US 2 · $357.56 per petaFLOP-day ($37.19/hour, 8-GPU node) · verified 2026-09-04Azure · A100 SXM · West US 3 · $357.56 per petaFLOP-day ($37.19/hour, 8-GPU node) · verified 2026-09-04AWS · A100 SXM · Asia Pacific (Tokyo) · $361.75 per petaFLOP-day ($37.62/hour, 8-GPU node) · verified 2026-09-04Azure · A100 SXM · Canada East · $378.12 per petaFLOP-day ($39.32/hour, 8-GPU node) · verified 2026-09-04Azure · A100 SXM · South Central US · $378.12 per petaFLOP-day ($39.32/hour, 8-GPU node) · verified 2026-09-04Azure · A100 SXM · France Central · $393.87 per petaFLOP-day ($40.96/hour, 8-GPU node) · verified 2026-09-04Azure · A100 SXM · Spain Central · $393.87 per petaFLOP-day ($40.96/hour, 8-GPU node) · verified 2026-09-04Azure · A100 SXM · UK South · $393.87 per petaFLOP-day ($40.96/hour, 8-GPU node) · verified 2026-09-04Azure · A100 SXM · Central US · $398.52 per petaFLOP-day ($41.45/hour, 8-GPU node) · verified 2026-09-04Azure · A100 SXM · West US · $409.63 per petaFLOP-day ($42.60/hour, 8-GPU node) · verified 2026-09-04Azure · A100 SXM · UK West · $412.78 per petaFLOP-day ($42.93/hour, 8-GPU node) · verified 2026-09-04Azure · A100 SXM · Canada Central · $420.58 per petaFLOP-day ($43.74/hour, 8-GPU node) · verified 2026-09-04Azure · A100 SXM · Italy North · $436.33 per petaFLOP-day ($45.38/hour, 8-GPU node) · verified 2026-09-04Azure · A100 SXM · Sweden Central · $436.33 per petaFLOP-day ($45.38/hour, 8-GPU node) · verified 2026-09-04Azure · A100 SXM · West Europe · $436.33 per petaFLOP-day ($45.38/hour, 8-GPU node) · verified 2026-09-04Azure · A100 SXM · UAE North · $450.59 per petaFLOP-day ($46.86/hour, 8-GPU node) · verified 2026-09-04Azure · A100 SXM · Poland Central · $450.87 per petaFLOP-day ($46.89/hour, 8-GPU node) · verified 2026-09-04Azure · A100 SXM · South India · $475.80 per petaFLOP-day ($49.48/hour, 8-GPU node) · verified 2026-09-04Azure · A100 SXM · Japan West · $488.40 per petaFLOP-day ($50.79/hour, 8-GPU node) · verified 2026-09-04Azure · A100 SXM · Norway East · $493.05 per petaFLOP-day ($51.28/hour, 8-GPU node) · verified 2026-09-04Azure · A100 SXM · Switzerland North · $493.05 per petaFLOP-day ($51.28/hour, 8-GPU node) · verified 2026-09-04Azure · A100 SXM · Australia East · $499.35 per petaFLOP-day ($51.93/hour, 8-GPU node) · verified 2026-09-04Azure · A100 SXM · Japan East · $499.36 per petaFLOP-day ($51.93/hour, 8-GPU node) · verified 2026-09-04Azure · A100 SXM · South Africa North · $502.50 per petaFLOP-day ($52.26/hour, 8-GPU node) · verified 2026-09-04Azure · A100 SXM · Germany North · $574.97 per petaFLOP-day ($59.80/hour, 8-GPU node) · verified 2026-09-04Azure · A100 SXM · Switzerland West · $585.76 per petaFLOP-day ($60.92/hour, 8-GPU node) · verified 2026-09-04Azure · A100 SXM · UAE Central · $585.76 per petaFLOP-day ($60.92/hour, 8-GPU node) · verified 2026-09-04Azure · A100 SXM · Brazil South · $630.19 per petaFLOP-day ($65.54/hour, 8-GPU node) · verified 2026-09-043.0× spreadHourly instance rate ÷ the instance’s dense BF16 throughput, over 24 hours. Cheaper is left. 126 Published on-demand offerings drawn, each a distinct provider, instance and region. **1** further offering has no placeable rate and is named in the caption. 7 GPU classes, one row each, ordered by median rate. The ordering is itself a reading: the cheapest compute per unit of throughput is the newest silicon’s. 6 Providers, across **51** priced regions. **101** of the drawn offerings come from Azure alone, so much of the width below is one seller’s own regional spread. 5.4× The widest spread inside a single GPU class: H100 SXM, from Lambda’s single-GPU rate in US (Texas) at **$104.05** (verified 2026-04-19) to Azure’s 8-GPU rate in South Africa West at **$565.76** (verified 2026-09-04). 3.4× The same measurement restricted to the newest refresh (2026-09-04), across the **28** of 126 drawn offerings verified that day. The headline pair above was verified 138 days apart, so the difference between these two figures measures the share of the published dispersion that is refresh staleness. _The price dispersion plate._ One mark per published on-demand offering, positioned on a shared logarithmic axis of dollars per petaFLOP-day (the instance’s hourly rate divided by its dense BF16 throughput, with no sparsity uplift). The axis is logarithmic because the drawn range spans 8.5 times end to end and the reading this plate offers is a ratio; equal ratios are equal distances here. Mark ink encodes the date the price was verified: full ink for the 2026-09-04 refresh, half-tone for older offerings, and a hairline outline with no fill for the oldest at 2026-03-29. The short vertical rule in each row is that row’s median; marks are nudged off the centre line only so a cluster stays countable. One offering is not drawn: GCP’s H200 SXM in US Central (Iowa) (no published rate). 3 of the drawn offerings are priced as single-GPU instances and the rest as multi-GPU nodes; the denominator divides node size out, so they share an axis, but they are different products to buy. These are list prices. A large buyer signs a committed-capacity contract at rates no rate card publishes, so the plate measures the dispersion of published on-demand rates. It does not measure what compute actually costs the firms buying most of it. A rate card quotes dollars per hour for a machine, which leaves any two of them incomparable until the hardware inside is divided out. Normalizing to $/petaFLOP-day (the hourly instance rate over its dense BF16 throughput, with no sparsity uplift) puts 126 published on-demand offerings from 6 providers across 51 regions on one axis. Inside H100 SXM alone the cheapest and most expensive published rates are 5.4 times apart, and the pair was verified 138 days apart; restricted to the newest refresh alone, the spread is 3.4 times. Much of that width is one seller’s geography: 101 of the 126 offerings are Azure’s own regional price list. None of it is what a hyperscale buyer pays: committed-capacity contracts are negotiated and no rate card publishes them. The index below tracks the cheapest of these published rates from the day it began recording. ### The index over time Cite CSVJSONJSON+Prov PrintScreenshot [Download dataset (CSV + Croissant + BibTeX)](/datasets/compute-cost-index) Price index over time ## The $/petaFLOP-day index Index base date 2026-07-20·47 daily points, 2026-07-20 → 2026-09-04 Cheapest on-demand $/petaFLOP-day for H100 SXM (8-GPU node), across the daily-refreshed providers (AWS, Azure, Oracle). GCP, CoreWeave, and Lambda publish no fetchable pricing API and are captured by manual snapshot, so this line excludes them; they appear, with their verification dates, in the vintage view below. **Pool composition.** This is a cheapest-across-the-pool statistic, so its level moves when the pool changes even if no provider changed price. 3 stretches were drawn from fewer than the full 3: **2026-07-21 → 2026-08-08**, Azure only; **2026-08-13 → 2026-08-13**, AWS + Oracle (OCI) only; **2026-08-20 → 2026-08-20**, AWS + Oracle (OCI) only. The capture cron’s AWS and Oracle branches were failing silently over that window (fixed 2026-08-09); the step up and back down at its edges is the pool changing width, and no reading should be taken across it. The shaded band on the plot marks it. The line is computed over the daily-refresh provider pool, and the vertical rules mark the dates that pool changed width. A level shift at a rule reflects that change in composition; it is not a price movement. ### Current quotes by vintage The age of each live on-demand quote for H100 SXM (8-GPU node) right now; it is not a historical series. Oldest first, and a manual-snapshot quote can be materially older than a daily-refreshed one. A provider whose regions share one verification date folds to one line; every region’s own figure remains in the rate table below. Provider Region $/pFLOP-day Verified Cadence CoreWeave US East (N. Virginia) $149.29 2026-03-29 manual snapshot GCP US Central (Iowa) $268.29 2026-04-19 manual snapshot Lambda US (Texas) $104.05 2026-04-19 manual snapshot AWS 3 regions $166.87–$208.59 2026-09-04 daily Azure 24 regions $311.48–$565.76 2026-09-04 daily Oracle (OCI) Global $242.55 2026-09-04 daily ### Every rate, and what building instead would cost The plate above draws the on-demand tier because that is the tier every provider publishes. The table below adds the rest (spot, one-year and three-year reserved, each cell dated with its verification) alongside an on-prem total-cost buildup at three- and five-year horizons. Both columns are stated in the same $/petaFLOP-day, so the gap between the cheapest cloud cell and the on-prem figure is the build-versus-buy boundary. [How the index is derived↗](/methodology#cost-index "How this is derived: the Compute Cost Index ($/petaFLOP-day). Published cloud rate cards normalized to one petaFLOP-day of BF16 dense compute (the Scrutica Compute Unit).") ~$104.05per SCU CHEAPEST ON-DEMAND Lambda • US (Texas) • 1×-GPU node ratemanual snapshot · 2026-04-19 ~$55.09per SCU CHEAPEST SPOT Azure • East US 2 • priced 2026-09-04 ~$91.10per SCU ON-PREM (3YR AMORT.) 88% of on-demand [How $/SCU is derived↗](/methodology#cost-index "How this is derived: the Compute Cost Index ($/petaFLOP-day). Published cloud rate cards normalized to one petaFLOP-day of BF16 dense compute (the Scrutica Compute Unit).") List ratesCloud PricingBuild vs buyOn-PremiseSide by sideComparisonBy regionGeographicOptimizerProcurement Mix GPUA100 SXMH100 SXMH200 SXMMI300XB200L40SGB200 Filtered pricesCSVJSONJSON+Prov Pricing tier On-DemandSpotReserved 1yrReserved 3yr GPU cloud rental cost index across hyperscaler and neocloud providers: hourly published rates normalized to $/SCU ($/petaFLOP-day, Scrutica metric; BF16 dense throughput per H100/A100/B200 sourced from Epoch AI ML Hardware). Each row is source-attributed. Provider Instance Region $/hr $/SCU Priced Source AWS p5.48xlarge US East (N. Virginia) ↓ $43.28 since 2026-04-19$55.04 $166.87 2026-09-04 AWS Pricing Bulk API AWS p5.48xlarge Asia Pacific (Tokyo) ↓ $54.39 since 2026-04-19$68.80 $208.59 2026-09-04 AWS Pricing Bulk API AWS p5.48xlarge US West (Oregon) ↓ $43.28 since 2026-04-19$55.04 $166.87 2026-09-04 AWS Pricing Bulk API Azure Standard\_ND96isr\_H100\_v5 Central US $120.93 $366.65 2026-09-04 Azure Retail Prices API Azure Standard\_ND96isr\_H100\_v5 Japan West $156.81 $475.43 2026-09-04 Azure Retail Prices API Azure Standard\_ND96isr\_H100\_v5 Spain Central $127.82 $387.52 2026-09-04 Azure Retail Prices API Azure Standard\_ND96isr\_H100\_v5 Poland Central $127.82 $387.52 2026-09-04 Azure Retail Prices API Azure Standard\_ND96isr\_H100\_v5 West Europe $127.82 $387.52 2026-09-04 Azure Retail Prices API Azure Standard\_ND96isr\_H100\_v5 US Gov Arizona $122.90 $372.61 2026-09-04 Azure Retail Prices API Azure Standard\_ND96isr\_H100\_v5 East US ↑ $4.42 since 2026-04-19$102.74 $311.48 2026-09-04 Azure Retail Prices API Azure Standard\_ND96isr\_H100\_v5 Korea Central $137.15 $415.81 2026-09-04 Azure Retail Prices API Azure Standard\_ND96isr\_H100\_v5 Japan East $142.56 $432.23 2026-09-04 Azure Retail Prices API Azure Standard\_ND96isr\_H100\_v5 West Central US $117.98 $357.71 2026-09-04 Azure Retail Prices API Azure Standard\_ND96isr\_H100\_v5 West US 3 $102.74 $311.48 2026-09-04 Azure Retail Prices API Azure Standard\_ND96isr\_H100\_v5 Switzerland North $145.01 $439.66 2026-09-04 Azure Retail Prices API Azure Standard\_ND96isr\_H100\_v5 Sweden Central $127.82 $387.52 2026-09-04 Azure Retail Prices API Azure Standard\_ND96isr\_H100\_v5 South Africa North $173.09 $524.77 2026-09-04 Azure Retail Prices API Azure Standard\_ND96isr\_H100\_v5 West US $127.82 $387.52 2026-09-04 Azure Retail Prices API Azure Standard\_ND96isr\_H100\_v5 North Central US $122.40 $371.10 2026-09-04 Azure Retail Prices API Azure Standard\_ND96isr\_H100\_v5 UK South $122.90 $372.61 2026-09-04 Azure Retail Prices API Azure Standard\_ND96isr\_H100\_v5 East US 2 ↑ $4.42 since 2026-04-19$102.74 $311.48 2026-09-04 Azure Retail Prices API Azure Standard\_ND96isr\_H100\_v5 South Africa West $186.61 $565.76 2026-09-04 Azure Retail Prices API Azure Standard\_ND96isr\_H100\_v5 Canada Central $122.40 $371.10 2026-09-04 Azure Retail Prices API Azure Standard\_ND96isr\_H100\_v5 West US 2 ↑ $4.42 since 2026-04-19$102.74 $311.48 2026-09-04 Azure Retail Prices API Azure Standard\_ND96isr\_H100\_v5 UAE North $140.60 $426.27 2026-09-04 Azure Retail Prices API Azure Standard\_ND96isr\_H100\_v5 South Central US $117.97 $357.66 2026-09-04 Azure Retail Prices API Azure Standard\_ND96isr\_H100\_v5 Australia East $142.60 $432.34 2026-09-04 Azure Retail Prices API CoreWeave HGX-H100\_SXM5\_80GB US East (N. Virginia) $49.24 $149.29 2026-03-29 CoreWeave published rate card GCP a3-highgpu-8g US Central (Iowa) ↓ $9.66 since 2026-03-29$88.49 $268.29 2026-04-19 gcloud-compute.com (Google Cloud rate mirror) Lambda H100\_SXM5\_80GB-1gpu US (Texas) $4.29 $104.05 2026-04-19 Lambda published rate card Oracle Cloud OCI-H100\_SXM5\_80GB Global $80.00 $242.55 2026-09-04 Oracle Cloud price list **Unit.** $/SCU values use BF16 **dense** throughput; the divisor for each class is tabled below. NVIDIA marketing quotes sparsity-inclusive rates (1,979 / 624 / 4,500 for H100 / A100 / B200); a reader deriving $/SCU from provider $/hour against those sparsity-inclusive numbers will read values ~2× lower than the index here. See the Methodology section below for the unit choice. Dense BF16 throughput per GPU for each class priced on this page — the denominator every $/SCU figure here is divided by. Sourced from the Epoch AI ML Hardware dataset. GPU class Dense BF16 TFLOP/s Offerings A100 SXM 312 37 H100 SXM 989.5 31 H200 SXMCompute-identical to the H100 SXM; the H200’s advantage is HBM3e memory bandwidth, which this denominator does not measure. 989.5 37 MI300XAMD datasheet dense FP16/BF16 (2,614.9 with sparsity). AMD’s sparsity support differs from NVIDIA’s 2:4, so the dense figure is the only like-for-like basis across the two vendors. 1,307.4 17 B200 2,250 3 L40SNVIDIA quotes 362.05 dense and 733 with sparsity for this part; the dense figure is used, as everywhere else on this page. 362.05 1 GB200 2,500 1 Dense BF16 TFLOP/s per GPU, from the Epoch AI ML Hardware dataset; no 2:4 structured sparsity applied. **Refresh cadence.** The cadence is split; the Priced column gives each row's own verification date. AWS on-demand refreshes daily from the AWS Pricing Bulk API (GitHub Actions worker); Azure (all four tiers), AWS spot / reserved (ec2.shop), and OCI on-demand refresh daily via cron. GCP, CoreWeave, and Lambda publish no fetchable pricing API and refresh only when the manual snapshot pipeline is run, with each run verified against the provider's public pricing page. **Coverage.** 7 GPU classes are priced here: A100 SXM (37 offerings from 6 providers), H200 SXM (37 offerings from 5 providers, 1 without a published rate), H100 SXM (31 offerings from 6 providers), MI300X (17 offerings from 1 provider), B200 (3 offerings from 3 providers), GB200 (1 offering from 1 provider) and L40S (1 offering from 1 provider). Lambda rows price single-GPU (1×) instances and are labeled as such. The per-GPU rate on Lambda's 8-GPU training nodes is a separate listing and arrives with the next manual snapshot refresh (these rows verified 2026-04-19). Methodology **Scrutica Compute Unit (SCU)** \= 1 petaFLOP-day of BF16 training compute = 8.64 × 1019 FLOP. At 40% MFU against BF16 dense throughput, one H100 SXM produces ~0.40 SCU per day (989.5 TFLOP/s × 0.40 = 395.8 TFLOP/s = 0.396 PFLOP/s). **Cloud pricing**: $/SCU = hourly instance price × 24 / (num\_gpus × bf16\_tflops / 1000). BF16 TFLOP/s dense (no 2:4 structured sparsity) from the Epoch AI ML Hardware dataset; the divisor for each of the 7 classes priced here is tabled beneath the rate table. NVIDIA marketing sheets quote sparsity-inclusive figures; Scrutica uses dense for consistency with the FLOP Capacity Engine and real-world training throughput. **On-premise**: $/SCU = (hardware amortization + lifetime power cost) / (daily PFLOP output × useful life in days). Includes GPU, networking ($5K/GPU), and facility share ($3K/GPU) amortized over the useful life set on the On-Premise tab — 3 years at present. Cloud prices verified 2026-03-29 – 2026-09-04 (dot tracks the oldest row)Estimation bounds = parameter sensitivity ranges ### Where the throughput figures come from Every $/petaFLOP-day on this page divides a published price by a published throughput figure, and the throughput figure comes from Epoch AI’s ML Hardware database: dense BF16 TFLOP/s per accelerator, without 2:4 structured sparsity, which is the baseline production training runs are calibrated against. [The full derivation, with its assumptions](/methodology#cost-index). Cited[ML Hardware database](https://epoch.ai/data/machine-learning-hardware)  Epoch AI · Epoch AI · published May 8, 2026 · Scrutica accessed May 19, 2026 · CC-BY 4.0 Per-accelerator BF16 / FP16 throughput specs (e.g., 989.5 dense BF16 TFLOP/s on H100 SXM; no 2:4 structured sparsity factor) feed the denominator in the $/petaFLOP-day computation across the AWS / Azure / GCP / OCI / CoreWeave / Lambda hourly rate cards. The cost index inherits Epoch's published spec verbatim; the hourly pricing layer is Scrutica's contribution.