Kovara GPU Intelligence · GPU0008
NVIDIA logo

NVIDIA H200 SXM

AI training / inference / HPC

Technical Profile

Standardized accelerator master data recorded by Kovara.

GPU ID
GPU0008
Manufacturer
NVIDIA
Architecture
Hopper
Generation
Hopper
VRAM
141 GB
Memory Type
HBM3E
Memory Bandwidth
4.8 TB/s
Form Factor
SXM
TDP
700 W
Launch Year
2023
Generation Status
Previous
Last Verified
2026-09-03

Current Market Coverage

Current public NVIDIA H200 SXM Offering records tracked by Kovara.

Current Offerings
4
Current public records
Providers
4
Distinct providers represented
Regions
4
Distinct regions represented
Available
3
Offerings recorded as Available
Observed Comparable Range
$3.99–$10.60 USD / GPU-hour
Normalized tracked values only
Freshest Market Check
Checked 1 day ago
Offering last_checked signal

Current Tracked Offerings

Providers, regions, comparable prices, availability and evidence quality.

Explore on Compute →
Availability
Available
Confirmed capacity: Not confirmed
Configuration: 1 GPU configuration · 1 GPU per node
Price
$3.99 USD / GPU-hour
Original: USD 3.99 instance-hour
Commercial
On-Demand
GPU VM · Per-minute
Trust
ConfidenceA
Checked 1 day ago
Confirmed 03 Sept 2026
Source: SRC000002
Availability
Capacity Block
Confirmed capacity: Not confirmed
Configuration: 8 GPU configuration · 8 GPUs per node
Price
$5.97 USD / GPU-hour
Original: USD 47.76 instance-hour
Commercial
Capacity Block
EC2 p5e.48xlarge · Hourly
Trust
ConfidenceA
Checked 1 day ago
Confirmed 03 Sept 2026
Source: SRC000011
Availability
Available
Confirmed capacity: Not confirmed
Configuration: 4 GPU configuration · 4 GPUs per node
Price
$6.72 USD / GPU-hour
Original: EUR 23.12 instance-hour
Commercial
On-Demand
h200-4-gpu · Per-minute
Trust
ConfidenceA
Checked 1 day ago
Confirmed 03 Sept 2026
Source: SRC000020
Availability
Available
Confirmed capacity: Not confirmed
Configuration: 8 GPU configuration · 8 GPUs per node
Price
$10.60 USD / GPU-hour
Original: USD 84.81 instance-hour
Commercial
On-Demand
a3-ultragpu-8g · Hourly
Trust
ConfidenceA
Checked 1 day ago
Confirmed 03 Sept 2026
Source: SRC000016

Platform Capabilities

Hardware capabilities recorded in GPU master data.

NVLink Supported
Yes
MIG Supported
Yes
Primary Use
AI training / inference / HPC