Why choose the GH200?

Grace Hopper superchip combines ARM CPU with H100 GPU and 96GB or 144GB HBM3e. Unified memory architecture with 900 GB/s NVLink. Efficient for AI workloads that benefit from tight CPU-GPU integration.

When is the GH200 not a good fit?

ARM CPU ecosystem may have compatibility issues with some software stacks. For standard GPU cloud workloads, a regular H100 is simpler to deploy and more widely available.

Are GH200 prices going up or down?

The median on-demand price across providers has fallen about 35% since August 2025, from $3.90 to $2.52/hr per GPU. This reflects the market-wide median, which moves both when providers change prices and when lower or higher priced offerings enter the market.

What size AI models can the GH200 run?

With 96GB of VRAM, the GH200 is well suited to 30B-class models in FP16, and 70B-class models in 4-bit or 8-bit quantized form.

How much VRAM does the GH200 have?

The GH200 has 96GB of VRAM. Multi-GPU setups increase total memory, but that memory is not automatically pooled across GPUs.

What is the GH200's memory bandwidth?

The GH200 has 4,000 GB/s of memory bandwidth. Higher bandwidth helps with faster data transfer between GPU memory and compute cores.

What data types does the GH200 support?

The GH200 supports 7 precision formats. Training: BF16, FP16, TF32, FP32. Inference: FP8, INT8. Scientific: FP64.

Does the GH200 support NVLink?

Yes. The GH200 supports NVLink with 900 GB/s of bidirectional bandwidth. This helps accelerate multi-GPU communication.

How much does the GH200 cost per hour?

GH200 pricing currently ranges from $1.99/hr to $6.50/hr per GPU, depending on the provider, instance type, and billing model.

How much does the GH200 cost per month?

At 720 hours per month, one GH200 can cost between $1,432.80 to $4,680.00 per month, depending on the provider. Reserved and spot pricing can lower that further.

Which cloud providers offer the GH200?

The GH200 is available from 3 cloud providers: CoreWeave, Vultr, Lambda Labs.

Can I rent the GH200 in the cloud?

Yes. We currently track 3 GH200 listings across 3 cloud providers: Billing type Listings Avg $/GPU/hr On-demand 2 $4.39/hr Spot 1 $1.99/hr

Nvidia GH200

Grace Hopper superchip combining ARM CPU with H100 GPU.

Last update 8 minutes ago

Compare vs other GPUs →

Aggregating historical prices...

Key Specifications

Architecture

Grace Hopper

Memory

96GB HBM3 + 480GB LPDDR5X

Memory Bandwidth

4,000 GB/s + 512 GB/s

Release date

Q2 2023

GH200 Pricing and Availability

The GH200 is listed by 3 cloud providers. Available capacity currently sits around $6.50/hr per GPU (on-demand). Spot instances start lower, at $1.99/hr per GPU.

3 providers

How this list works

The two views

By provider shows one card per company, placed where its best offer ranked. By configuration lists every offer.

Order

What you can rent comes first: in stock, then waitlist, then not reported, then out of stock. Priced offers rank ahead of quote-only, and on-demand ahead of other billing types.

Within a group, five factors set the order, largest first:

Location: datacenter proximity, blended with provider HQ.
Price: hourly price, per GPU and in total.
Billing type: reserved ahead of spot, spot ahead of quote-only.
Specs: more VRAM, vCPUs and RAM.
Provider diversity: in By configuration, a provider's repeat rows rank slightly lower, so one company can't take all the top spots.

Search

We match the provider, GPU model, form factor, billing type, availability and instance name. Matching is partial and case-insensitive. Everyday words work too, so "interruptible" finds spot and "sold out" finds out of stock.

Transparency and funding

Ads and sponsors: Paid placements sit at the top of the list and are always labeled as sponsored content. Sponsorship never influences the organic ranking itself.
Affiliates: Affiliate links are marked. We may earn a commission if you click them, but commissions never affect the order.
Prices: Shown in USD, converted at daily reference rates where a provider publishes in another currency. A month means 720 hours. How we estimate costs.

Vultr In stock

USA 1 config 1x

Spot from $1.99

From $1.99 / GPU / hr Spot Visit website


1x GH200 vbm-72c-480gb-gh200-gpu	96GB	72	480GB	Spot	$1.99	$1.99	In stock

CoreWeave

USA 1 config 1x

On-Demand from $6.50

From $6.50 / GPU / hr On-Demand Visit website


1x GH200	96GB	72	480GB	On-Demand	$6.50	$6.50	Not reported

Lambda Labs Out of stock

USA 1 config 1x

On-Demand from $2.29

From $2.29 / GPU / hr On-Demand Visit website


1x GH200 gpu_1x_gh200	96GB	64	432GB	On-Demand	$2.29	$2.29	Out of stock

Heads up: We do our best to keep these specs & prices accurate. However, cloud costs may fluctuate based on region, usage, and other factors not listed here. These are estimates based on common setups and are for informational purposes only. Always verify current rates & exact specs with the provider before provisioning.

Frequently Asked Questions

Why choose the GH200?: Grace Hopper superchip combines ARM CPU with H100 GPU and 96GB or 144GB HBM3e. Unified memory architecture with 900 GB/s NVLink. Efficient for AI workloads that benefit from tight CPU-GPU integration.
When is the GH200 not a good fit?: ARM CPU ecosystem may have compatibility issues with some software stacks. For standard GPU cloud workloads, a regular H100 is simpler to deploy and more widely available.
Are GH200 prices going up or down?: The median on-demand price across providers has fallen about 35% since August 2025, from $3.90 to $2.52/hr per GPU. This reflects the market-wide median, which moves both when providers change prices and when lower or higher priced offerings enter the market.
What size AI models can the GH200 run?: With 96GB of VRAM, the GH200 is well suited to 30B-class models in FP16, and 70B-class models in 4-bit or 8-bit quantized form.
How much VRAM does the GH200 have?: The GH200 has 96GB of VRAM. Multi-GPU setups increase total memory, but that memory is not automatically pooled across GPUs.
What is the GH200's memory bandwidth?: The GH200 has 4,000 GB/s of memory bandwidth. Higher bandwidth helps with faster data transfer between GPU memory and compute cores.
What data types does the GH200 support?: The GH200 supports 7 precision formats. Training: BF16, FP16, TF32, FP32. Inference: FP8, INT8. Scientific: FP64.
Does the GH200 support NVLink?: Yes. The GH200 supports NVLink with 900 GB/s of bidirectional bandwidth. This helps accelerate multi-GPU communication.
How much does the GH200 cost per hour?: GH200 pricing currently ranges from $1.99/hr to $6.50/hr per GPU, depending on the provider, instance type, and billing model.
How much does the GH200 cost per month?: At 720 hours per month, one GH200 can cost between $1,432.80 to $4,680.00 per month, depending on the provider. Reserved and spot pricing can lower that further.
Which cloud providers offer the GH200?: The GH200 is available from 3 cloud providers: CoreWeave, Vultr, Lambda Labs.
Can I rent the GH200 in the cloud?: Yes. We currently track 3 GH200 listings across 3 cloud providers:

Billing type Listings Avg $/GPU/hr

On-demand 2 $4.39/hr

Spot 1 $1.99/hr

Billing type	Listings	Avg $/GPU/hr
On-demand	2	$4.39/hr
Spot	1	$1.99/hr

Technical Specifications

Feature	GH200	GH200 NVL2
CPU core count	72 Arm Neoverse V2 cores	144 Arm Neoverse V2 cores
L1 cache	64KB i-cache + 64KB d-cache	64KB i-cache + 64KB d-cache
L2 cache	1MB per core	1MB per core
L3 cache	114MB	228MB
Base frequency \| all-core single instruction, multiple data (SIMD) frequency	3.1GHz	3.0GHz
LPDDR5X size	480GB, 120GB, 240GB	960GB, 240GB, 480GB
Memory bandwidth	Up to 384GB/s, Up to 512GB/s	Up to 768GB/s, Up to 1024GB/s
PCIe links	Up to 4x PCIe x16 (Gen5)	Up to 8x PCIe x16 (Gen5)
FP64	34 teraFLOPS	68 teraFLOPS
FP64 Tensor Core	67 teraFLOPS	134 teraFLOPS
FP32	67 teraFLOPS	134 teraFLOPS
TF32 Tensor Core	989 teraFLOPS*, 494 teraFLOPS	1,979 teraFLOPS*, 990 teraFLOPS
BFLOAT16 Tensor Core	1,979 teraFLOPS*, 990 teraFLOPS	3,958 teraFLOPS*, 1,979 teraFLOPS
FP16 Tensor Core	1,979 teraFLOPS*, 990 teraFLOPS	3,958 teraFLOPS*, 1,979 teraFLOPS
FP8 Tensor Core	3,958 teraFLOPS*, 1,979 teraFLOPS	7,916 teraFLOPS*, 3,958 teraFLOPS
INT8 Tensor Core	3,958 TOPS*, 1,979 TOPS	7,916 TOPS*, 3,958 TOPS
High-bandwidth memory (HBM) size	96GB HBM3	144GB HBM3e, Up to 288GB HBM3e
Memory bandwidth	Up to 4TB/s, Up to 4.9TB/s	Up to 9.8TB/s
NVIDIA NVLink-C2C CPU-to-GPU bandwidth	900 GB/s	900 GB/s
Power	Configurable 450 to 1000W (Memory + CPU + GPU)	Configurable 900W to 2000W (Memory + CPU + GPU)
Thermal solution	Air cooled or liquid cooled	Air cooled or liquid cooled