The 12 GPU rows
Leafcloud publishes on its own public pricing page, mirrored once a day and
shown exactly as published. Below the table is something no price-comparison site
normally shows: the recorded legal basis on which GPUDeck republishes these
figures, including the condition that would void it.
Mirrored 2026-08-26T20:33:06Z ·
read from https://www.leaf.cloud/pricing ·
document digest f41582fad9de5095 · EUR per month, full VM
Flavor
GPU
Cards
TDP
Price / month
eg1.a100x1.V12-84
NVIDIA A100
1
300 W
€1,174.48
eg1.a100x2.V25-164
NVIDIA A100
2
300 W
€2,338.34
eg1.a100x4.V50-324
NVIDIA A100
4
300 W
€4,692.62
eg1.a100x8.V100-680
NVIDIA A100
8
300 W
€9,385.23
eg1.a30x1.V8-32
NVIDIA A30
1
165 W
€441.10
eg1.a30x2.V16-64
NVIDIA A30
2
165 W
€887.50
eg1.a30x4.V32-128
NVIDIA A30
4
165 W
€1,775.01
eg1.a30x8.V64-256
NVIDIA A30
8
165 W
€3,544.70
eg1.h100x1.V24_96
NVIDIA H100
1
700 W
€3,006.43
eg2.rtxp6000x1.V32-256-860
NVIDIA RTX 6000 Blackwell
1
600 W
€2,014.80
eg2.rtxp6000x2.V64-512-1720
NVIDIA RTX 6000 Blackwell
2
600 W
€4,029.60
eg2_rtxp6000x4_V128_1024_3440
NVIDIA RTX 6000 Blackwell
4
600 W
€8,059.20
Why GPUDeck is allowed to publish this
Every source on this site carries a written admission record before a
single row ships, under methodology §8. Silence
from a provider is not a basis; the basis has to be either a licence or a
recorded, reasoned conclusion that no permission is required. This is that record,
verbatim from the source registry.
Which law decides this
governing_lawNETHERLANDS_EU - Dutch law. Leafcloud's June 2026 terms state that its website pricing IS the advertised price source, and contain no scraping, copying or republication clause.
database_right_screenPASS. The decisive object is NOT 'Leafcloud's maintained cloud catalogue' - it is the 12 first-party GPU SKU rows embedded in the public /pricing HTML, with no competitor rows, calculator output, availability, regions or backend access. CJEU authority carries both halves of the rule: C-203/02 excludes investment in CREATING prices, SKUs and configurations, while C-46/02 paras 39-40 allow that a database of self-created data may STILL qualify where substantial resources independent of creation went into collecting, systematically arranging, individually exposing or continuously verifying it. The admitted object shows none of those operational features: 12 predetermined rows, five functional fields, no third-party offers, no regional or availability matrix, no authenticated catalogue or calculator, no searchable interface, no separately monetized data product, and routine table presentation through a generic component. That is not merely 'smaller than Nebius'.
Article 7 places the qualifying burden on the MAKER. The recorded conclusion is therefore 'substantial independent investment is not evidenced for the defined 12-row GPU tariff' - NOT a claim that no internal Leafcloud catalogue exists. GPUDeck does not have to invent hidden investment because it might exist.
LARGER-CATALOGUE FALLBACK: even if Leafcloud's whole public catalogue qualified, GPUDeck takes only the 12 GPU rows and publishes them as dated attributed observations with a source link. It exposes no query interface over the catalogue and does not replace ordering, which materially distinguishes C-762/19 CV-Online Latvia, where the service effectively exposed the protected database's searchable contents.
retrieval_scopeAt most one daily request to the public /pricing page. Leafcloud's OWN GPU tariff rows only. EXCLUDED: its AWS/Azure/RunPod/Lambda comparison table, its derived percent-cheaper claims, dynamic-price forecasts, calculator-generated prices, and pricing.leaf.cloud backend data.
What is published
publication_scopeDated rows naming Leafcloud as provider, carrying VAT basis, billing unit and observation time, each an exact first-party GPU price or spec fact, linking back to Leafcloud.
Could this substitute for the provider
substitution_riskLOW - GPUDeck is not reproducing the ordering service. Actual purchasing, configuration, current terms and current availability remain with Leafcloud.
The provider's own terms
terms_statusREAD_NO_REUSE_RESTRICTION_FOUND
What would void this conclusion
disqualifying_conditionEvidence of substantial independent investment in obtaining, verifying or presenting the tariff would engage Article 7 and end this. So would any reuse clause appearing in Leafcloud's terms, or any expansion into the comparison table, the calculator, or the pricing backend.
This is a point-in-time reading, and it is watched.
Leafcloud’s governing terms were read at
https://www.leaf.cloud/terms/ and hashed
(4a2ec897fd1429e769187bea1582df7c). A daily check re-fetches that
document and compares the digest; if the terms change to restrict reuse, the basis above
stops being true and these rows are withdrawn.
What is deliberately NOT here
At most one daily request to the public /pricing page. Leafcloud's OWN GPU tariff rows only. EXCLUDED: its AWS/Azure/RunPod/Lambda comparison table, its derived percent-cheaper claims, dynamic-price forecasts, calculator-generated prices, and pricing.leaf.cloud backend data.
No figure on this page is combined with any other source. GPUDeck never
computes a median, ranking or “cheapest” across providers, because a number
pooled from sources with different segments, regions and bundles describes nothing anyone
can actually buy. Compare the records yourself on prices.
Provenance
Analysis
Sol (ChatGPT) round-4 ruling; recorded by the gpudeck session · recorded 2026-08-26T16:25:00Z
Relationship
GPUDeck has no commercial relationship with
Leafcloud. No referral link, no sponsorship, no payment. The
independence disclosure lists every relationship that does
exist.
Corrections
Errors are corrected through the public
corrections register. Frozen editions are never edited.