Lowest observed rate
Nebius AI Cloud provides the lowest row in this exact evidence set at $0.7900 per GPU-hour. That is a list-price observation, not proof that capacity is rentable now.
Exact GPU provider comparison
Compare Microsoft Azure vs Nebius AI Cloud H100 GPU prices: 185 official-source rates across 32 regions, $0.790–$23.326/GPU-hour. Estimate monthly cost.
Decision-page price snapshot:Checking the declared provider-SLA boundary.Valid through ; historical after that instant.
Decision readout
Lowest observed rate
Nebius AI Cloud provides the lowest row in this exact evidence set at $0.7900 per GPU-hour. That is a list-price observation, not proof that capacity is rentable now.
Price dispersion
The highest row is 2852.7% above the lowest. At 160 GPU-hours, selecting opposite ends of the observed range changes the compute line item by $3605.76.
Capacity boundary
2 of 185 rates carry separate typed availability evidence. Check each row's evidence kind and status; coverage does not guarantee current availability, and unknown never inherits confidence from a fresh price.
on demand floor gap
Nebius AI Cloud's lowest observed on demand H100 rate is 35.5% below Microsoft Azure's floor ($4.5000 vs $6.9800/GPU-hour). At 160 GPU-hours, that floor-to-floor difference is $396.80. Same GPU model and purchase model; variant, region, node configuration, contract terms and evidenced availability can still differ.
Variant evidence: Nebius AI Cloud: NVLink 80GB · Microsoft Azure: NVL 94GB, SXM 80GB.
Variant mismatch: both providers expose explicit H100 variants in this matched market slice, but none overlap. Treat this floor gap as model-level rather than variant-equivalent until an exact shared variant is selected.
Availability evidence coverage: Nebius AI Cloud 1/1 rows (100%) vs Microsoft Azure 0/94 (0%). Coverage can include deployment eligibility, stock status or allocatable capacity observations; inspect each evidence kind and status before treating an offer as available.
spot floor gap
Nebius AI Cloud's lowest observed spot H100 rate is 44.3% below Microsoft Azure's floor ($0.7900 vs $1.4183/GPU-hour). At 160 GPU-hours, that floor-to-floor difference is $100.53. Same GPU model and purchase model; variant, region, node configuration, contract terms and evidenced availability can still differ.
Variant evidence: Nebius AI Cloud: NVLink 80GB · Microsoft Azure: NVL 94GB, SXM 80GB.
Variant mismatch: both providers expose explicit H100 variants in this matched market slice, but none overlap. Treat this floor gap as model-level rather than variant-equivalent until an exact shared variant is selected.
Availability evidence coverage: Nebius AI Cloud 1/1 rows (100%) vs Microsoft Azure 0/89 (0%). Coverage can include deployment eligibility, stock status or allocatable capacity observations; inspect each evidence kind and status before treating an offer as available.
This comparison holds the accelerator model constant at H100 and changes the provider. Nebius AI Cloud has the lowest observed row; Microsoft Azure contains the highest observed row. Before choosing, compare accelerator variant, region, GPU count per node, market type and separately evidenced capacity—not only the headline rate.
Side-specific evidence profile
The profile below prevents a large catalog from silently overwhelming a smaller one. Counts, regions, access models and observed price bands are stated independently for every provider.
Official configuration set
Nebius AI Cloud contributes 2 qualifying configurations and 2 distinct source records. Nebius AI Cloud's observed price band runs from $0.7900 to $4.5000 per GPU-hour, equivalent to $126.40–$720.00 for 160 GPU-hours.
Nebius AI Cloud covers 1 named region: eu-north1. The included node configurations range from 1 to 1 GPUs, with 1 on-demand and 1 spot record.
Nebius AI Cloud has independent typed availability evidence for 2 of 2 rows. Check the evidence kind and status before treating those rows as available. Its latest included price observation is 2026-11-01. GPU models in this exact comparison are H100.
Official configuration set
Microsoft Azure contributes 183 qualifying configurations and 183 distinct source records. Microsoft Azure's observed price band runs from $1.4183 to $23.3260 per GPU-hour, equivalent to $226.93–$3732.16 for 160 GPU-hours.
Microsoft Azure covers 31 named regions: australiaeast, canadacentral, centralindia, centralus, eastus, eastus2, germanywestcentral, indonesiacentral and 23 more. The included node configurations range from 1 to 8 GPUs, with 94 on-demand and 89 spot records.
Microsoft Azure has independent typed availability evidence for 0 of 183 rows. Check the evidence kind and status before treating those rows as available. Its latest included price observation is 2026-11-01. GPU models in this exact comparison are H100.
Transparent scenario calculator
The calculation multiplies GPU count by hours per GPU and the selected normalized rate. It excludes storage, egress, support, taxes, reservations and idle orchestration overhead.
Configuration-level evidence
Every row preserves the source configuration and observation date. Node rates are divided by GPU count to produce the normalized comparison. Equal normalized prices can still represent different regions, node sizes or access conditions.
| Provider / configuration | GPU | Region | Market | Node price | GPU-hour | 160h cost | Snapshot availability | Evidence |
|---|---|---|---|---|---|---|---|---|
| Nebius AI Cloudnebius:h100-nvlink:eu-north1:spot | H100NVLink 80GB · 1 GPU/node | eu-north1 | spot | $0.7900 | $0.7900 | $126.40 | availableEvidence: deployment eligibilityScope: exact configurationObserved Valid through Availability source ↗ | Official source ↗Checked 2026-11-01 |
| Nebius AI Cloudnebius:h100-nvlink:eu-north1:on_demand | H100NVLink 80GB · 1 GPU/node | eu-north1 | on demand | $4.5000 | $4.5000 | $720.00 | availableEvidence: deployment eligibilityScope: exact configurationObserved Valid through Availability source ↗ | Official source ↗Checked 2026-11-01 |
| Microsoft Azureazure:DZH318Z0F8W2/000J:eastus:spot | H100NVL 94GB · 2 GPU/node | eastus | spot | $2.8367 | $1.4183 | $226.93 | Not evidencedEvidence: unknownScope: unknown | Official source ↗Checked 2026-11-01 |
| Microsoft Azureazure:DZH318Z0F8W2/000R:eastus:spot | H100NVL 94GB · 1 GPU/node | eastus | spot | $1.4183 | $1.4183 | $226.93 | Not evidencedEvidence: unknownScope: unknown | Official source ↗Checked 2026-11-01 |
| Microsoft Azureazure:DZH318Z0F8W2/0099:centralus:spot | H100NVL 94GB · 2 GPU/node | centralus | spot | $3.1730 | $1.5865 | $253.84 | Not evidencedEvidence: unknownScope: unknown | Official source ↗Checked 2026-11-01 |
| Microsoft Azureazure:DZH318Z0F8W2/009B:centralus:spot | H100NVL 94GB · 1 GPU/node | centralus | spot | $1.5865 | $1.5865 | $253.84 | Not evidencedEvidence: unknownScope: unknown | Official source ↗Checked 2026-11-01 |
| Microsoft Azureazure:DZH318Z0F8W2/0079:koreasouth:spot | H100NVL 94GB · 2 GPU/node | koreasouth | spot | $3.2248 | $1.6124 | $257.98 | Not evidencedEvidence: unknownScope: unknown | Official source ↗Checked 2026-11-01 |
| Microsoft Azureazure:DZH318Z0F8W2/007B:koreasouth:spot | H100NVL 94GB · 1 GPU/node | koreasouth | spot | $1.6124 | $1.6124 | $257.98 | Not evidencedEvidence: unknownScope: unknown | Official source ↗Checked 2026-11-01 |
| Microsoft Azureazure:DZH318Z0F8W2/0018:germanywestcentral:spot | H100NVL 94GB · 2 GPU/node | germanywestcentral | spot | $3.3537 | $1.6769 | $268.30 | Not evidencedEvidence: unknownScope: unknown | Official source ↗Checked 2026-11-01 |
| Microsoft Azureazure:DZH318Z0F8W2/002G:germanywestcentral:spot | H100NVL 94GB · 1 GPU/node | germanywestcentral | spot | $1.6769 | $1.6769 | $268.30 | Not evidencedEvidence: unknownScope: unknown | Official source ↗Checked 2026-11-01 |
Showing up to eight lowest normalized rows per provider (10 of 185) so one large catalog cannot crowd out the other side. Open the global interactive database to filter the complete catalog.
Comparison discipline
All included rows passed the provider-SLA and official-source gates when this snapshot was generated. The page is regenerated when source evidence changes and becomes historical after its displayed deadline.
A fresh price is not inventory. Availability is shown only when a separate exact-configuration source supports it, with its own observation clock and limitations.
Networking, storage, support, taxes, commitments, setup work and interruption risk can change total cost. Confirm those terms before purchasing.
This page exists only while automated evidence and uniqueness thresholds pass. One failed cycle removes it from the index; three consecutive failed cycles retire the route.
Adjacent decisions
These links are selected by shared provider or GPU model, so the next page changes one decision variable instead of sending you to a generic article.
provider model comparison snapshot
184 published official-source rows across 32 regions. Open the page to inspect its own price range and validity deadline.
Open comparisonprovider model comparison snapshot
184 published official-source rows across 31 regions. Open the page to inspect its own price range and validity deadline.
Open comparisonprovider model comparison snapshot
184 published official-source rows across 31 regions. Open the page to inspect its own price range and validity deadline.
Open comparisonprovider model comparison snapshot
184 published official-source rows across 31 regions. Open the page to inspect its own price range and validity deadline.
Open comparisonprovider model comparison snapshot
184 published official-source rows across 31 regions. Open the page to inspect its own price range and validity deadline.
Open comparisonprovider model comparison snapshot
184 published official-source rows across 31 regions. Open the page to inspect its own price range and validity deadline.
Open comparison