Weekly Pricing Pulse

About Pulse

The Weekly Pricing Pulse presents what changed in published GPU cloud pricing across 9 providers in the last 7 days. Every week, AIForge Works runs a fresh harvest across AWS, Azure, GCP, OCI, CoreWeave, Lambda, Vultr, Nebius, and Crusoe, diffs the new data against the prior week, and surfaces the moves worth knowing about: price changes greater than 1% on existing comparable entries, New Silicon, Deprecated SKUs, and likely price/SKU rolls. The Pulse is not a live quote, not workload-specific advice, and not personalized to your account. The numbers reflect public catalog prices at the moment of harvest. For a live quote, check the provider. For workload-specific recommendations, use Cloud Advisor. To see neighboring SKUs, regions, or commitment terms for any mover that catches your eye, open the Multi-Cloud Pricing Analyzer. For an audit of your own GPU spend against a comparable floor, use GPU Audit.

Editor's note

This was a catalog-availability week, not a repricing week. All 9 providers reported clean harvests, and no comparable on-demand per-GPU price observation moved more than 1% week over week. The visible activity came from Lambda and Vultr showing New Silicon and Deprecated SKU entries, with Lambda accounting for most of the churn.

For buyers, the practical read is simple: do not rewrite budget assumptions from this week's public catalog alone, but do refresh regional availability before carrying forward last week's shortlist. Lambda's movement touches A100, H100, and GH200 entries across multiple markets, and Vultr changed a smaller set of GH200, L40S, A16, and A40 entries. The A100 entries should be read as newly visible public-catalog entries, not proof of newly installed hardware; older silicon can appear because of region availability, feed normalization, retagging, or actual capacity changes. This is enough to affect what appears selectable in a procurement screen, even though it is not evidence of broad market repricing.

OCPI vs Global Public Catalog floor

This week's Pulse includes the full OCPI comparison inside the report. It compares Ornn x ICE OCPI readings with AIForge Works indexed Public Catalog floors for H100 SXM, H200, B200, and A100 SXM4.

Read the comparison as directional market context, not a live quote or a geography-matched procurement claim.

What moved this week

No provider had comparable on-demand offers over the 1% movement threshold. Lambda and Vultr changed catalog visibility through New Silicon and Deprecated SKU entries, shown below.

• On-demand pricing was quiet this week. We did not find any public on-demand GPU price changes above the 1% reporting threshold; the visible movement is catalog availability.

Top on-demand movers by provider

No comparable offers met the 1% movement threshold for this section.

Up to 5 price observations per provider, ranked by absolute per-GPU rate change. See Methodology for how per-GPU rates are computed.

Provider roll-up

Existing price changes track comparable provider + region + instance/SKU observations over the 1% movement threshold. New Silicon excludes same-provider, same-geo, same-instance likely SKU rolls. Deprecated SKUs remain exact catalog SKUs no longer visible as purchasable rows.

New Silicon this week

The full New Silicon list is 14 provider-region-instance entries: 12 Lambda and 2 Vultr. The same SKU appearing in multiple regions counts once per region. New Silicon means newly visible in this week's public catalog snapshot, not necessarily newly installed physical capacity.

Deprecated SKUs this week

The full Deprecated SKU list is 56 provider-region-instance entries: 49 Lambda and 7 Vultr. These entries were present in last week's public catalog snapshot and absent in this week's snapshot; that does not by itself prove the provider retired the hardware or permanently withdrew capacity.

Catalog continuity checks

These checks do not collapse the New Silicon / Deprecated SKU counts. They flag public-catalog patterns that may reflect SKU/package relabeling, region-label normalization, repricing, or true availability movement. They are broader continuity candidates, not strict same-provider, same-geo, same-instance likely price/SKU rolls, so the likely-roll count remains 0.

Methodology

What we compared. This Pulse compares last week's harvest (May 18, 2026) against this week's (May 25, 2026). Both harvests run automatically against the same 9 cloud providers.

How we compare prices fairly. Different providers package GPUs differently: some sell single-GPU instances, others sell 8-GPU instances. To compare apples-to-apples, we divide every instance price by the number of GPUs on that instance. All movement percentages in this Pulse are based on that per-GPU hourly rate, not the raw instance price.

What counts as a real move. We only flag price changes greater than 1%. Anything smaller is treated as rounding noise and excluded.

Reserved/committed terms. This Pulse reports on-demand movement from the weekly diff. Reserved and committed pricing require separate review before publication.

Currency. All prices are in US dollars. Catalog entries missing a clean price or verified GPU specs are dropped before comparison.

How we choose summary callouts. The short summary only highlights price moves that are broad enough to read as a market signal. Narrow one-provider movements may still appear in detail tables, but we do not treat them as market-wide repricing.

This week's coverage. All 9 providers reported clean harvests. After normalizing 5.45M raw catalog price observations against verified hardware specs, the broader MPA catalog includes all deployable instance offers. This Pulse narrows that catalog to roughly 13,000 comparable on-demand GPU observations for week-over-week movement analysis.

This Pulse compiles publicly published prices. It is not a live quote and is not workload-specific advice. See the About section at the top for links to AIForge Works pricing and availability tools.

Next Pulse: Week of 2026-06-01.