Hyperscaler · 11 GPU models tracked
Microsoft Azure
$0.07–$27.04per GPU-hour, the cheapest row to the dearest
Linux virtual machines: pay-as-you-go, spot, and 1- and 3-year reserved rates. For each VM size the cheapest region is shown and named in the row. Multi-GPU VMs are priced per VM; the rate here is divided by the GPU count.
As of , Microsoft Azure lists 11 GPU models we track, from $0.07 per GPU-hour (NVIDIA T4, spot) to $27.04 (NVIDIA GB200 NVL72, on-demand).
- Cheapest row$0.07NVIDIA T4 · spot
- H200$10.60on-demand · westus2, 8-GPU node · 133% above the median $4.54 (51)
- H100 SXM$11.06on-demand · eastus, 8-GPU node · 214% above the median $3.52 (38) · spot from $2.04
- A100 80GB$3.67on-demand · eastus2, 2-GPU node · 84% above the median $2.00 (58) · spot from $0.68
| Website | azure.microsoft.com/en-us |
|---|---|
| Pricing page | azure.microsoft.com/en-us/pricing/details/virtual-machines/linux/ |
| Regions and tiers | australiacentral2, centralus, eastus, eastus2, koreasouth, northcentralus, northeurope, uaenorth, westus2, westus3 — as labelled in the rates below |
| Models tracked | 11 |
Price by GPU
As of , per GPU-hour: Microsoft Azure’s on-demand rate for each card it lists, a cheaper market it also sells, and where the on-demand rate stands among the providers we track.
| GPU | Microsoft Azure on-demand, per GPU-hour | Against the market (median rate card) |
|---|---|---|
| NVIDIA GB200 NVL72 | $27.04 westus3, 4-GPU node captured spot $6.22 Deploy ↗ | 104% above the median $13.25 · 6 providers 5 of 6 list lower · lowest Bitdeer AI $6.94 from · compare |
| NVIDIA H100 SXM | $11.06 eastus, 8-GPU node captured spot $2.04 Deploy ↗ | 214% above the median $3.52 · 38 providers 37 of 38 list lower · lowest Cudo Compute $1.79 GPU only · compare |
| NVIDIA H200 | $10.60 westus2, 8-GPU node captured Deploy ↗ | 133% above the median $4.54 · 51 providers 49 of 51 list lower · lowest Beam $2.09 from · compare |
| NVIDIA H100 NVL | $6.98 eastus captured spot $1.29 Deploy ↗ | 65% above the median $4.23 · 8 providers 6 of 8 list lower · lowest GPU.ai $2.06 from · compare |
| AMD Instinct MI300X | $6.00 eastus2, 8-GPU node captured spot $1.11 Deploy ↗ | 114% above the median $2.81 · 8 providers 6 of 8 list lower · lowest TensorWave $1.71 from |
| NVIDIA RTX PRO 6000 Blackwell | $5.50 eastus2 captured spot $1.02 Deploy ↗ | 148% above the median $2.21 · 38 providers 37 of 38 list lower · lowest Hostinger $0.65 from · compare |
| NVIDIA A100 80GB | $3.67 eastus2, 2-GPU node captured spot $0.68 Deploy ↗ | 84% above the median $2.00 · 58 providers 48 of 58 list lower · lowest CloudRift $1.05 · compare |
| NVIDIA A100 40GB | $3.40 westus2, 8-GPU node captured spot $0.75 Deploy ↗ | 71% above the median $1.98 · 20 providers 19 of 20 list lower · lowest Trooper.AI $0.68 from · compare |
| NVIDIA A10 | $3.20 westus3 captured spot $0.59 Deploy ↗ | 162% above the median $1.22 · 9 providers 8 of 9 list lower · lowest Intelion Cloud $0.34 GPU only · compare |
| NVIDIA V100 | $3.06 eastus, 2-GPU node captured spot $0.57 Deploy ↗ | 23% above the median $2.48 · 11 providers 9 of 11 list lower · lowest Verda $0.23 · compare |
| NVIDIA T4 | $0.53 northcentralus captured spot $0.07 Deploy ↗ | 4% below the median $0.55 · 15 providers 5 of 15 list lower · lowest immers.cloud $0.24 · compare |
The same, in words
- Microsoft Azure lists the NVIDIA GB200 NVL72 from $27.04 per GPU-hour on-demand (westus3, 4-GPU node, cheapest of 19 regions read; dearest $54.08), $6.22 spot (koreasouth, 4-GPU node), $11.90 reserved (3yr, cheapest of 19 regions read; dearest $23.80). The on-demand rate is unchanged since first read, Sep 7, 2026 (7-day change: none). Against the market: 104% above the median rate card ($13.25 across 6 providers); 5 of the 6 providers we track list a lower on-demand rate, the lowest Bitdeer AI at $6.94 (Bitdeer AI’s rates, Microsoft Azure vs Bitdeer AI).
- Microsoft Azure lists the NVIDIA H100 SXM from $11.06 per GPU-hour on-demand (eastus, 8-GPU node, cheapest of 24 regions read; dearest $23.33), $2.04 spot (eastus, 8-GPU node), $4.86 reserved (3yr, cheapest of 24 regions read; dearest $10.24). The on-demand rate is unchanged since first read, Sep 7, 2026 (7-day change: none). Against the market: 214% above the median rate card ($3.52 across 38 providers); 37 of the 38 providers we track list a lower on-demand rate, the lowest Cudo Compute at $1.79 (Cudo Compute’s rates, Microsoft Azure vs Cudo Compute).
- Microsoft Azure lists the NVIDIA H200 from $10.60 per GPU-hour on-demand (westus2, 8-GPU node, cheapest of 29 regions read; dearest $21.20), $5.28 reserved (3yr, cheapest of 29 regions read; dearest $10.56). The on-demand rate is unchanged since first read, Sep 7, 2026 (7-day change: none). Against the market: 133% above the median rate card ($4.54 across 51 providers); 49 of the 51 providers we track list a lower on-demand rate, the lowest Beam at $2.09 (Beam’s rates, Microsoft Azure vs Beam).
- Microsoft Azure lists the NVIDIA H100 NVL from $6.98 per GPU-hour on-demand (eastus, cheapest of 28 regions read; dearest $10.12), $1.29 spot (eastus), $3.84 reserved (3yr, cheapest of 26 regions read; dearest $5.57). The on-demand rate is unchanged since first read, Sep 16, 2026 (7-day change: none). Against the market: 65% above the median rate card ($4.23 across 8 providers); 6 of the 8 providers we track list a lower on-demand rate, the lowest GPU.ai at $2.06 (GPU.ai’s rates, Microsoft Azure vs GPU.ai).
- Microsoft Azure lists the AMD Instinct MI300X from $6.00 per GPU-hour on-demand (eastus2, 8-GPU node, cheapest of 17 regions read; dearest $12.00), $1.11 spot (eastus2, 8-GPU node), $2.63 reserved (3yr, cheapest of 17 regions read; dearest $5.27). The on-demand rate is unchanged since first read, Sep 19, 2026 (7-day change: none). Against the market: 114% above the median rate card ($2.81 across 8 providers); 6 of the 8 providers we track list a lower on-demand rate, the lowest TensorWave at $1.71 (TensorWave’s rates).
- Microsoft Azure lists the NVIDIA RTX PRO 6000 Blackwell from $5.50 per GPU-hour on-demand (eastus2, cheapest of 23 regions read; dearest $7.98), $1.02 spot (eastus), $2.42 reserved (3yr, cheapest of 23 regions read; dearest $3.51). The on-demand rate is unchanged since first read, Sep 26, 2026 (7-day change: none). Against the market: 148% above the median rate card ($2.21 across 38 providers); 37 of the 38 providers we track list a lower on-demand rate, the lowest Hostinger at $0.65 (Hostinger’s rates, Microsoft Azure vs Hostinger).
- Microsoft Azure lists the NVIDIA A100 80GB from $3.67 per GPU-hour on-demand (eastus2, 2-GPU node, cheapest of 38 regions read; dearest $7.61), $0.68 spot (eastus2, 2-GPU node), $1.36 reserved (3yr, cheapest of 36 regions read; dearest $3.35). The on-demand rate is unchanged since first read, Sep 7, 2026 (7-day change: none). Against the market: 84% above the median rate card ($2.00 across 58 providers); 48 of the 58 providers we track list a lower on-demand rate, the lowest CloudRift at $1.05 (CloudRift’s rates, Microsoft Azure vs CloudRift).
- Microsoft Azure lists the NVIDIA A100 40GB from $3.40 per GPU-hour on-demand (westus2, 8-GPU node, cheapest of 6 regions read; dearest $4.42), $0.75 spot (westus2, 8-GPU node), $1.36 reserved (3yr, cheapest of 6 regions read; dearest $1.77). The on-demand rate is unchanged since first read, Sep 7, 2026 (7-day change: none). Against the market: 71% above the median rate card ($1.98 across 20 providers); 19 of the 20 providers we track list a lower on-demand rate, the lowest Trooper.AI at $0.68 (Trooper.AI’s rates, Microsoft Azure vs Trooper.AI).
- Microsoft Azure lists the NVIDIA A10 from $3.20 per GPU-hour on-demand (westus3, cheapest of 37 regions read; dearest $6.40), $0.59 spot (centralus), $1.41 reserved (3yr, cheapest of 35 regions read; dearest $2.82). The on-demand rate is unchanged since first read, Sep 23, 2026 (7-day change: none). Against the market: 162% above the median rate card ($1.22 across 9 providers); 8 of the 9 providers we track list a lower on-demand rate, the lowest Intelion Cloud at $0.34 (Intelion Cloud’s rates, Microsoft Azure vs Intelion Cloud).
- Microsoft Azure lists the NVIDIA V100 from $3.06 per GPU-hour on-demand (eastus, 2-GPU node, cheapest of 23 regions read; dearest $6.12), $0.57 spot (eastus, 4-GPU node). The on-demand rate is unchanged since first read, Sep 16, 2026 (7-day change: none). Against the market: 23% above the median rate card ($2.48 across 11 providers); 9 of the 11 providers we track list a lower on-demand rate, the lowest Verda at $0.23 (Verda’s rates, Microsoft Azure vs Verda).
- Microsoft Azure lists the NVIDIA T4 from $0.53 per GPU-hour on-demand (northcentralus, cheapest of 31 regions read; dearest $1.29), $0.07 spot (australiacentral2), $0.20 reserved (3yr, cheapest of 30 regions read; dearest $0.34). The on-demand rate is unchanged since first read, Sep 23, 2026 (7-day change: none). Against the market: 4% below the median rate card ($0.55 across 15 providers); 5 of the 15 providers we track list a lower on-demand rate, the lowest immers.cloud at $0.24 (immers.cloud’s rates, Microsoft Azure vs immers.cloud).
All listed prices
One table per price type, never mixed. Every rate normalized to one GPU for one hour; a price opens the page it was read from, and 7d marks a rate that moved in the past seven days. The day each figure was first listed is in this page’s data, below.
On-demand
| GPU · what is sold | $ / GPU-hour | Read | Source |
|---|---|---|---|
| NVIDIA T4APInorthcentralus | $0.53 | captured | Deploy ↗ |
| NVIDIA V100API$6.12/node ÷ 2 · eastus | $3.06 | captured | Deploy ↗ |
| NVIDIA A10APIwestus3 | $3.20 | captured | Deploy ↗ |
| NVIDIA A100 40GBAPI$27.20/node ÷ 8 · westus2 | $3.40 | captured | Deploy ↗ |
| NVIDIA A100 80GBAPI$7.35/node ÷ 2 · eastus2 | $3.67 | captured | Deploy ↗ |
| NVIDIA RTX PRO 6000 BlackwellAPIeastus2 | $5.50 | captured | Deploy ↗ |
| AMD Instinct MI300XAPI$48.00/node ÷ 8 · eastus2 | $6.00 | captured | Deploy ↗ |
| NVIDIA H100 NVLAPIeastus | $6.98 | captured | Deploy ↗ |
| NVIDIA H200API$84.80/node ÷ 8 · westus2 | $10.60 | captured | Deploy ↗ |
| NVIDIA H100 SXMAPI$88.49/node ÷ 8 · eastus | $11.06 | captured | Deploy ↗ |
| NVIDIA GB200 NVL72API$108.16/node ÷ 4 · westus3 | $27.04 | captured | Deploy ↗ |
Spot / interruptible
| GPU · what is sold | $ / GPU-hour | Read | Source |
|---|---|---|---|
| NVIDIA T4APIaustraliacentral2 | $0.07 | captured | Deploy ↗ |
| NVIDIA V100API$1.13/node ÷ 2 · eastus2 | $0.57 | stoppedlast confirmed | |
| NVIDIA V100API$2.26/node ÷ 4 · eastus · −$2.49 (−82%) vs its on-demand $3.06 | $0.57 | captured | Deploy ↗ |
| NVIDIA A10APIeastus2 | $0.59 | stoppedlast confirmed | |
| NVIDIA A10APIcentralus | $0.59 | captured | Deploy ↗ |
| NVIDIA A100 80GBAPI$1.36/node ÷ 2 · westus3 | $0.68 | stoppedlast confirmed | |
| NVIDIA A100 80GBAPI$1.36/node ÷ 2 · eastus2 · −$2.99 (−82%) vs its on-demand $3.67 | $0.68 | captured | Deploy ↗ |
| NVIDIA A100 40GBAPI$5.98/node ÷ 8 · westus2 · −$2.65 (−78%) vs its on-demand $3.40 | $0.75 | captured | Deploy ↗ |
| NVIDIA RTX PRO 6000 BlackwellAPIeastus | $1.02 | captured | Deploy ↗ |
| AMD Instinct MI300XAPI$8.87/node ÷ 8 · eastus2 · −$4.89 (−82%) vs its on-demand $6.00 | $1.11 | captured | Deploy ↗ |
| NVIDIA H100 NVLAPIeastus · −$5.69 (−82%) vs its on-demand $6.98 | $1.29 | captured | Deploy ↗ |
| NVIDIA H100 SXMAPI$16.35/node ÷ 8 · eastus · −$9.02 (−82%) vs its on-demand $11.06 | $2.04 | captured | Deploy ↗ |
| NVIDIA GB200 NVL72API$24.88/node ÷ 4 · koreasouth | $6.22 | captured | Deploy ↗ |
| NVIDIA H200API$101.76/node ÷ 8 · northeurope | $12.72 | stoppedlast confirmed | |
| NVIDIA H200API$121.26/node ÷ 8 · uaenorth | $15.16 | stoppedlast confirmed |
Reserved (committed use)
| GPU · what is sold | $ / GPU-hour | Read | Source |
|---|---|---|---|
| NVIDIA T4API3-year · eastus · min. commitment 3yr | $0.20 | captured | Reserve ↗ |
| NVIDIA T4API1-year · northcentralus · min. commitment 1yr · −$0.22 (−41%) vs its on-demand $0.53 | $0.31 | captured | Reserve ↗ |
| NVIDIA A100 40GBAPI3-year · $10.88/node ÷ 8 · westus2 · min. commitment 3yr · −$2.04 (−60%) vs its on-demand $3.40 | $1.36 | captured | Reserve ↗ |
| NVIDIA A100 80GBAPI3-year · $2.73/node ÷ 2 · eastus2 · min. commitment 3yr · −$2.31 (−63%) vs its on-demand $3.67 | $1.36 | captured | Reserve ↗ |
| NVIDIA A10API3-year · centralus · min. commitment 3yr | $1.41 | captured | Reserve ↗ |
| NVIDIA A10API1-year · westus3 · min. commitment 1yr · −$1.15 (−36%) vs its on-demand $3.20 | $2.05 | captured | Reserve ↗ |
| NVIDIA A100 40GBAPI1-year · $18.83/node ÷ 8 · westus2 · min. commitment 1yr · −$1.05 (−31%) vs its on-demand $3.40 | $2.35 | captured | Reserve ↗ |
| NVIDIA A100 80GBAPI1-year · westus3 · min. commitment 1yr | $2.40 | captured | Reserve ↗ |
| NVIDIA RTX PRO 6000 BlackwellAPI3-year · eastus2 · min. commitment 3yr · −$3.08 (−56%) vs its on-demand $5.50 | $2.42 | captured | Reserve ↗ |
| AMD Instinct MI300XAPI3-year · $21.07/node ÷ 8 · westus3 · min. commitment 3yr | $2.63 | captured | Reserve ↗ |
| NVIDIA RTX PRO 6000 BlackwellAPI1-year · westus2 · min. commitment 1yr | $3.52 | stoppedlast confirmed | |
| NVIDIA RTX PRO 6000 BlackwellAPI1-year · westus3 · min. commitment 1yr | $3.52 | captured | Reserve ↗ |
| NVIDIA H100 NVLAPI3-year · eastus · min. commitment 3yr · −$3.14 (−45%) vs its on-demand $6.98 | $3.84 | captured | Reserve ↗ |
| AMD Instinct MI300XAPI1-year · $30.72/node ÷ 8 · westus3 · min. commitment 1yr | $3.84 | captured | Reserve ↗ |
| NVIDIA H100 SXMAPI3-year · $38.85/node ÷ 8 · eastus · min. commitment 3yr · −$6.21 (−56%) vs its on-demand $11.06 | $4.86 | captured | Reserve ↗ |
| NVIDIA H100 NVLAPI1-year · $10.47/node ÷ 2 · eastus2 · min. commitment 1yr | $5.24 | captured | Reserve ↗ |
| NVIDIA H200API3-year · $42.22/node ÷ 8 · westus2 · min. commitment 3yr · −$5.32 (−50%) vs its on-demand $10.60 | $5.28 | captured | Reserve ↗ |
| NVIDIA H200API1-year · $46.52/node ÷ 8 · westus2 · min. commitment 1yr · −$4.78 (−45%) vs its on-demand $10.60 | $5.82 | captured | Reserve ↗ |
| NVIDIA H100 SXMAPI1-year · $56.63/node ÷ 8 · eastus · min. commitment 1yr · −$3.98 (−36%) vs its on-demand $11.06 | $7.08 | captured | Reserve ↗ |
| NVIDIA GB200 NVL72API3-year · $47.59/node ÷ 4 · westus3 · min. commitment 3yr · −$15.14 (−56%) vs its on-demand $27.04 | $11.90 | captured | Reserve ↗ |
| NVIDIA GB200 NVL72API1-year · $69.22/node ÷ 4 · eastus · min. commitment 1yr | $17.31 | captured | Reserve ↗ |
What else Microsoft Azure bills
The GPU-hour is not the whole bill. What Microsoft Azure's own pages say about the rest, read by hand and checked on ; each line links the page it was read from. "Not published" means we looked and the pages do not say. Every provider side by side.
| What | What Microsoft Azure states | Source |
|---|---|---|
| In the rate | VM price includes vCPU, RAM and local temp disk only; OS and data managed disks billed separately | source checked |
| Data out | First 100 GB/month free, then $0.087/GB for next 10 TB, from North America (Microsoft premium network routing) | source checked |
| Storage | Premium SSD LRS P30 (1 TiB) $135.17/month in East US, about $0.132/GiB-month; Premium SSD v2 $0.00011/GiB-hour our reading | source checked |
| While stopped | Stopped (allocated) keeps billing compute; Stopped (deallocated) stops compute billing, disks and networking still charged | source checked |
| Billing step | Per full minute; partial minute not charged; no minimum published | source checked |
Every rate on this page as data: /providers/azure/prices.json — the same fields as prices.json, narrowed to this provider. CC BY 4.0, no key, no rate limit. On your own page: embed this table (/embed/providers/azure, an iframe, no script).
Compare Microsoft Azure with
Every provider that lists at least two of the same GPUs on-demand, most shared GPUs first; each opens the side-by-side page.
- Oracle Cloud (OCI) 9 shared GPUs
- RunPod 8 shared GPUs
- Amazon Web Services 7 shared GPUs
- Modal 7 shared GPUs
- Vast.ai 7 shared GPUs
- Cerebrium 6 shared GPUs
- Google Cloud 6 shared GPUs
- Sesterce 6 shared GPUs
- Verda 6 shared GPUs
- CoreWeave 5 shared GPUs
- E2E Networks 5 shared GPUs
- EmpirioLabs 5 shared GPUs
- GPU.ai 5 shared GPUs
- immers.cloud 5 shared GPUs
- Jarvislabs 5 shared GPUs
- Lambda 5 shared GPUs
- Paperspace 5 shared GPUs
- Seeweb 5 shared GPUs
- Civo 4 shared GPUs
- Cloud.ru Evolution 4 shared GPUs
- Crusoe 4 shared GPUs
- Geodd 4 shared GPUs
- HexGrid 4 shared GPUs
- Hugging Face 4 shared GPUs
- Hyperstack 4 shared GPUs
- Lium 4 shared GPUs
- Massed Compute 4 shared GPUs
- Northflank 4 shared GPUs
- Barrack AI 3 shared GPUs
- Beam 3 shared GPUs
- Bitdeer AI 3 shared GPUs
- Denvr Dataworks 3 shared GPUs
- FPT AI Factory 3 shared GPUs
- Gcore 3 shared GPUs
- Hikube 3 shared GPUs
- Inferless 3 shared GPUs
- Koyeb 3 shared GPUs
- Nebius 3 shared GPUs
- Servercore 3 shared GPUs
- SynpixCloud 3 shared GPUs
- Trooper.AI 3 shared GPUs
- Atlas Cloud 2 shared GPUs
- Bamboozle Web Services 2 shared GPUs
- Baseten 2 shared GPUs
- Canopy Wave 2 shared GPUs
- centron 2 shared GPUs
- Clore.ai 2 shared GPUs
- CloudRift 2 shared GPUs
- Cudo Compute 2 shared GPUs
- Dataoorts 2 shared GPUs
- Daytona 2 shared GPUs
- Deep Infra 2 shared GPUs
- Elice Cloud 2 shared GPUs
- fal 2 shared GPUs
- GMI Cloud 2 shared GPUs
- Hostinger 2 shared GPUs
- HPC-AI.COM 2 shared GPUs
- Hyperbolic 2 shared GPUs
- Intelion Cloud 2 shared GPUs
- Leafcloud 2 shared GPUs
- Lyceum 2 shared GPUs
- Neysa 2 shared GPUs
- OVHcloud 2 shared GPUs
- packet.ai 2 shared GPUs
- QuantaCloud 2 shared GPUs
- Replicate 2 shared GPUs
- RunComfy 2 shared GPUs
- Selectel 2 shared GPUs
- STACKIT 2 shared GPUs
- Taiga Cloud (Northern Data Group) 2 shared GPUs
- Thunder Compute 2 shared GPUs
- Together AI 2 shared GPUs
- VESSL AI 2 shared GPUs
- Yandex Cloud 2 shared GPUs
Data updated: