Datasets:

alpha-one-index
/

ai-infra-index

Error code:   StreamingRowsError
Exception:    CastError
Message:      Couldn't cast
metadata: struct<updated: timestamp[s], source: string, url: string, providers_count: int64, total_skus: int64 (... 279 chars omitted)
  child 0, updated: timestamp[s]
  child 1, source: string
  child 2, url: string
  child 3, providers_count: int64
  child 4, total_skus: int64
  child 5, methodology: string
  child 6, last_curated: string
  child 7, pricing_sources: struct<Azure: string, RunPod: string, Lambda: string, CoreWeave: string, Together AI: string, Vast.a (... 116 chars omitted)
      child 0, Azure: string
      child 1, RunPod: string
      child 2, Lambda: string
      child 3, CoreWeave: string
      child 4, Together AI: string
      child 5, Vast.ai: string
      child 6, Vultr: string
      child 7, Nebius: string
      child 8, OCI: string
      child 9, Cudo Compute: string
      child 10, Fluidstack: string
      child 11, Paperspace: string
providers: struct<RunPod: list<item: struct<gpu: string, cnt: int64, mem: int64, on_demand: double, spot: doubl (... 877 chars omitted)
  child 0, RunPod: list<item: struct<gpu: string, cnt: int64, mem: int64, on_demand: double, spot: double>>
      child 0, item: struct<gpu: string, cnt: int64, mem: int64, on_demand: double, spot: double>
          child 0, gpu: string
          child 1, cnt: int64
          child 2, mem: int64
          child 3, on_demand: double
          child 4, spot: double
  child 1, Lambda: list<item: struct<gpu: string, cnt: int64, mem: int64, on_demand: double>>
      child 0, item: struct<
...
>
          child 0, gpu: string
          child 1, cnt: int64
          child 2, mem: int64
          child 3, on_demand: double
  child 7, OCI: list<item: struct<gpu: string, cnt: int64, mem: int64, on_demand: double>>
      child 0, item: struct<gpu: string, cnt: int64, mem: int64, on_demand: double>
          child 0, gpu: string
          child 1, cnt: int64
          child 2, mem: int64
          child 3, on_demand: double
  child 8, Cudo Compute: list<item: struct<gpu: string, cnt: int64, mem: int64, on_demand: double>>
      child 0, item: struct<gpu: string, cnt: int64, mem: int64, on_demand: double>
          child 0, gpu: string
          child 1, cnt: int64
          child 2, mem: int64
          child 3, on_demand: double
  child 9, Fluidstack: list<item: struct<gpu: string, cnt: int64, mem: int64, on_demand: double>>
      child 0, item: struct<gpu: string, cnt: int64, mem: int64, on_demand: double>
          child 0, gpu: string
          child 1, cnt: int64
          child 2, mem: int64
          child 3, on_demand: double
  child 10, Paperspace: list<item: struct<gpu: string, cnt: int64, mem: int64, on_demand: double>>
      child 0, item: struct<gpu: string, cnt: int64, mem: int64, on_demand: double>
          child 0, gpu: string
          child 1, cnt: int64
          child 2, mem: int64
          child 3, on_demand: double
provider: null
gpu_name: null
gpu_memory_gb: null
price_per_hour_usd: null
gpu_arch: null
vram_gb: null
fp16_tflops: null
source: null
to
{'provider': Value('string'), 'gpu_name': Value('string'), 'gpu_memory_gb': Value('string'), 'price_per_hour_usd': Value('float64'), 'gpu_arch': Value('string'), 'vram_gb': Value('int64'), 'fp16_tflops': Value('float64'), 'source': Value('string')}
because column names don't match
Traceback:    Traceback (most recent call last):
                File "/src/services/worker/src/worker/utils.py", line 99, in get_rows_or_raise
                  return get_rows(
                         ^^^^^^^^^
                File "/src/libs/libcommon/src/libcommon/utils.py", line 272, in decorator
                  return func(*args, **kwargs)
                         ^^^^^^^^^^^^^^^^^^^^^
                File "/src/services/worker/src/worker/utils.py", line 77, in get_rows
                  rows_plus_one = list(itertools.islice(ds, rows_max_number + 1))
                                  ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.12/site-packages/datasets/iterable_dataset.py", line 2543, in __iter__
                  for key, example in ex_iterable:
                                      ^^^^^^^^^^^
                File "/usr/local/lib/python3.12/site-packages/datasets/iterable_dataset.py", line 2060, in __iter__
                  for key, pa_table in self._iter_arrow():
                                       ^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.12/site-packages/datasets/iterable_dataset.py", line 2092, in _iter_arrow
                  pa_table = cast_table_to_features(pa_table, self.features)
                             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.12/site-packages/datasets/table.py", line 2192, in cast_table_to_features
                  raise CastError(
              datasets.table.CastError: Couldn't cast
              metadata: struct<updated: timestamp[s], source: string, url: string, providers_count: int64, total_skus: int64 (... 279 chars omitted)
                child 0, updated: timestamp[s]
                child 1, source: string
                child 2, url: string
                child 3, providers_count: int64
                child 4, total_skus: int64
                child 5, methodology: string
                child 6, last_curated: string
                child 7, pricing_sources: struct<Azure: string, RunPod: string, Lambda: string, CoreWeave: string, Together AI: string, Vast.a (... 116 chars omitted)
                    child 0, Azure: string
                    child 1, RunPod: string
                    child 2, Lambda: string
                    child 3, CoreWeave: string
                    child 4, Together AI: string
                    child 5, Vast.ai: string
                    child 6, Vultr: string
                    child 7, Nebius: string
                    child 8, OCI: string
                    child 9, Cudo Compute: string
                    child 10, Fluidstack: string
                    child 11, Paperspace: string
              providers: struct<RunPod: list<item: struct<gpu: string, cnt: int64, mem: int64, on_demand: double, spot: doubl (... 877 chars omitted)
                child 0, RunPod: list<item: struct<gpu: string, cnt: int64, mem: int64, on_demand: double, spot: double>>
                    child 0, item: struct<gpu: string, cnt: int64, mem: int64, on_demand: double, spot: double>
                        child 0, gpu: string
                        child 1, cnt: int64
                        child 2, mem: int64
                        child 3, on_demand: double
                        child 4, spot: double
                child 1, Lambda: list<item: struct<gpu: string, cnt: int64, mem: int64, on_demand: double>>
                    child 0, item: struct<
              ...
              >
                        child 0, gpu: string
                        child 1, cnt: int64
                        child 2, mem: int64
                        child 3, on_demand: double
                child 7, OCI: list<item: struct<gpu: string, cnt: int64, mem: int64, on_demand: double>>
                    child 0, item: struct<gpu: string, cnt: int64, mem: int64, on_demand: double>
                        child 0, gpu: string
                        child 1, cnt: int64
                        child 2, mem: int64
                        child 3, on_demand: double
                child 8, Cudo Compute: list<item: struct<gpu: string, cnt: int64, mem: int64, on_demand: double>>
                    child 0, item: struct<gpu: string, cnt: int64, mem: int64, on_demand: double>
                        child 0, gpu: string
                        child 1, cnt: int64
                        child 2, mem: int64
                        child 3, on_demand: double
                child 9, Fluidstack: list<item: struct<gpu: string, cnt: int64, mem: int64, on_demand: double>>
                    child 0, item: struct<gpu: string, cnt: int64, mem: int64, on_demand: double>
                        child 0, gpu: string
                        child 1, cnt: int64
                        child 2, mem: int64
                        child 3, on_demand: double
                child 10, Paperspace: list<item: struct<gpu: string, cnt: int64, mem: int64, on_demand: double>>
                    child 0, item: struct<gpu: string, cnt: int64, mem: int64, on_demand: double>
                        child 0, gpu: string
                        child 1, cnt: int64
                        child 2, mem: int64
                        child 3, on_demand: double
              provider: null
              gpu_name: null
              gpu_memory_gb: null
              price_per_hour_usd: null
              gpu_arch: null
              vram_gb: null
              fp16_tflops: null
              source: null
              to
              {'provider': Value('string'), 'gpu_name': Value('string'), 'gpu_memory_gb': Value('string'), 'price_per_hour_usd': Value('float64'), 'gpu_arch': Value('string'), 'vram_gb': Value('int64'), 'fp16_tflops': Value('float64'), 'source': Value('string')}
              because column names don't match

Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.

AI Infrastructure Index

Dataset Description

The AI Infrastructure Index is a comprehensive open-source reference for AI hardware specifications, cloud GPU pricing, and infrastructure intelligence. It catalogs major AI hardware platforms currently in production, covering data center GPUs, custom AI accelerators (TPUs, LPUs, IPUs, WSEs), cloud pricing, benchmarks, and cost optimization data.

Homepage: https://github.com/alpha-one-index/ai-infra-index
Repository: https://github.com/alpha-one-index/ai-infra-index
Live Data: https://alpha-one-index.github.io/ai-infra-index/
API Access: https://alphaoneindex.com/api-access/
Research: https://alphaoneindex.com/research/
Point of Contact: Alpha One Index

Dataset Summary

This dataset provides structured, machine-readable data on:

Cloud GPU Pricing — Real-time pricing from 12 cloud providers (Azure, RunPod, Lambda Labs, CoreWeave, Together AI, Vast.ai, etc.)
GPU Specifications — Detailed specs for NVIDIA (H100, H200, B200, GB200), AMD (MI300X, MI325X), and Intel (Gaudi 3) data center accelerators
Performance Benchmarks — FP16/FP32/INT8 throughput, memory bandwidth, and interconnect specs
Cost Optimization — Price-per-TFLOP calculations and cost efficiency rankings across providers

API Access

The AI Infrastructure Index offers a REST API for programmatic access to cloud GPU pricing data.

Base URL: https://gpu-pricing-api.alphaoneindex.workers.dev

Endpoints

Endpoint	Description
`GET /api/v1/pricing`	All cloud GPU pricing data
`GET /api/v1/pricing?provider=runpod`	Filter by provider
`GET /api/v1/pricing?gpu=H100`	Filter by GPU model
`GET /api/v1/gpu-specs`	GPU specifications
`GET /api/v1/gpu-specs?vendor=nvidia`	Filter specs by vendor

Quick Start

import requests

# Get all cloud GPU pricing
response = requests.get("https://gpu-pricing-api.alphaoneindex.workers.dev/api/v1/pricing")
data = response.json()

# Filter by provider
runpod = requests.get("https://gpu-pricing-api.alphaoneindex.workers.dev/api/v1/pricing?provider=runpod")
print(runpod.json())

Python (Hugging Face Datasets)

from datasets import load_dataset

# Load cloud pricing data
ds = load_dataset("alpha-one-index/ai-infra-index", split="cloud_pricing")
print(ds[0])

# Load GPU specifications
gpu_specs = load_dataset("alpha-one-index/ai-infra-index", split="gpu_specs")
print(gpu_specs[0])

Data Fields

Cloud Pricing Split

Field	Type	Description
`provider`	string	Cloud provider name
`gpu_name`	string	GPU model name
`gpu_memory_gb`	string	GPU VRAM
`price_per_hour_usd`	float	Hourly price in USD

GPU Specs Split

Field	Type	Description
`gpu_name`	string	GPU model name
`gpu_arch`	string	Architecture (Hopper, Ada, CDNA3, etc.)
`vram_gb`	int	Video memory in GB
`fp16_tflops`	float	FP16 performance in TFLOPS
`source`	string	Data source

Supported Providers

Provider	GPUs Tracked
RunPod	H100, A100, A6000, RTX 4090
Lambda Labs	H100, A100, A10
CoreWeave	H100, H200, A100
Together AI	H100, A100
Vast.ai	H100, A100, RTX 4090, RTX 3090
Azure	H100, A100, T4
AWS	H100, A100, T4, Inferentia
GCP	H100, A100, T4, TPU v5
Paperspace	H100, A100, A6000
Fluidstack	H100, A100
Tensordock	A100, A6000, RTX 4090
Oblivus	H100, A100

Use Cases

MLOps Cost Planning — Compare GPU pricing across providers for training and inference workloads
Hardware Selection — Choose optimal GPU based on performance-per-dollar metrics
Market Research — Track cloud GPU pricing trends over time
Academic Research — Reference data for AI infrastructure studies and papers

Citation

If you use this dataset in your research, please cite:

@dataset{alpha_one_index_2025,
  title={AI Infrastructure Index: Cloud GPU Pricing and Hardware Specifications},
  author={Alpha One Index},
  year={2025},
  url={https://huggingface.co/datasets/alpha-one-index/ai-infra-index},
  note={Comprehensive open-source reference for AI hardware specifications and cloud GPU pricing}
}

License

This dataset is released under the MIT License.

Updates

This dataset is updated regularly with the latest cloud GPU pricing and hardware specifications. For real-time data, visit the live dashboard or use the API.

Downloads last month: 10