AI Compute Tracker

Compare GPU rentals.
Understand what you’ll pay.

Compare GPUs, providers and billing terms—with dated sources.

16 accelerators · 13 provider brands

Start with a GPU.

Selected posted on-demand offers. Compare the terms behind each price.

H100

On-demand from

$3.20/GPU·hr

Hyperstack · lowest of 10 on-demand providers

Observed

Offer details

Exact rate $3.20/GPU·hr

Instance size and total not recorded

NVIDIA H100 SXM 80GB; maximum per GPU: 24 pCPUs and 240 GB RAM; actual allocation not specified; published per GPU-hour

Region: Provider regions vary

On-demand

AI Compute Tracker (aicomputetracker.com), CC BY 4.0; third-party terms apply to data from GPU cloud pricing pages

Compare offers

H200

On-demand from

$3.99/GPU·hr

Hyperstack · lowest of 8 on-demand providers

Observed

Offer details

Exact rate $3.99/GPU·hr

Instance size and total not recorded

NVIDIA H200 SXM 141GB; maximum per GPU: 22 pCPUs and 225 GB RAM; actual allocation not specified; published per GPU-hour

Region: Provider regions vary

On-demand

AI Compute Tracker (aicomputetracker.com), CC BY 4.0; third-party terms apply to data from GPU cloud pricing pages

Compare offers

B200

On-demand from

$6.00/GPU·hr

Hyperstack · lowest of 7 on-demand providers

Observed

Offer details

Exact rate $6.00/GPU·hr

Instance size and total not recorded

NVIDIA B200 192GB; maximum per GPU: 31 pCPUs and 256 GB RAM; actual allocation not specified; published per GPU-hour

Region: Provider regions vary

On-demand

AI Compute Tracker (aicomputetracker.com), CC BY 4.0; third-party terms apply to data from GPU cloud pricing pages

Compare offers

A100 80GB

On-demand from

$1.35/GPU·hr

Hyperstack · lowest of 7 on-demand providers

Observed

Offer details

Exact rate $1.35/GPU·hr

Instance size and total not recorded

NVIDIA A100 80GB; maximum per GPU: 28 pCPUs and 120 GB RAM; actual allocation not specified; published per GPU-hour

Region: Provider regions vary

On-demand

AI Compute Tracker (aicomputetracker.com), CC BY 4.0; third-party terms apply to data from GPU cloud pricing pages

Compare offers

Listed rates are normalized per accelerator-hour. Providers may require a multi-GPU instance; posted prices do not confirm available capacity.

Using a model API?

Compare the cost per token.

Input and output prices, with the terms that change your bill.

Compare API prices

Frontier example · xAI

Grok 4.7

Input / 1M tokens
$2.00
Output / 1M tokens
$6.00

Reviewed

Billing terms and source

Below 200K prompt tokens; at/above 200K $4/$12

Cache-hit input: $0.50 per 1M tokens.

Batch: $1.60 input / $4.80 output per 1M tokens.

Provider pricing source ↗

Workhorse example · xAI

Grok Build 0.1

Input / 1M tokens
$1.00
Output / 1M tokens
$2.00

Reviewed

Billing terms and source

Below 200K prompt tokens; at/above 200K $2/$4

Coding model; replaced grok-code-fast-1; batch has no discount

Cache-hit input: $0.20 per 1M tokens.

Batch price not recorded.

Provider pricing source ↗

Budget example · Anthropic

Claude Haiku 5.5

Input / 1M tokens
$0.10
Output / 1M tokens
$0.50

Reviewed

Billing terms and source

≤100K prompt tokens; above 100K $0.50/$2.50

Cache-hit input: $0.01 per 1M tokens.

Batch: $0.05 input / $0.25 output per 1M tokens.

Provider pricing source ↗

Examples use the lowest blended list price in each editorial tier among the latest tracked versions, with a 3:1 input/output mix. Tiers are not capability rankings. Cache, batch and longer prompts can change the bill.

Estimate your monthly API bill

Before you commit

Check the hardware. Check the quote.

What changed

Full log →
  1. Provider coverage

    B300 typical price: Verda added between the Oct 20 and Nov 16, 2026 quote snapshots

    · Venue change

  2. Data corrections

    Correction: Compare pages, badges and the reference-price file kept the Oct 20 quote table

    · Correction

  3. Market estimates

    H100 anchor repriced +5.6% between sessions

    · Anchor move

Latest recorded changes by category. Each links the dated comparison and evidence; these are not live 24-hour price moves.

Research behind the prices

A wider view of AI compute.

Explore Kalshi-implied GPU rental estimates, recorded price history and the infrastructure behind AI compute.

Forward-market snapshot · . Estimates are separate from posted rental offers.

Explore market research

Read the price with confidence

Know what you’re comparing.

Source links and observation dates travel with every offer.

What does a rental price include?

Rates are shown per GPU-hour, or per chip-hour for other accelerators. A provider may require a multi-GPU instance or charge separately for CPU, memory, storage and network traffic. Compare the recorded configuration and whole-instance cost before choosing.

How recently was a price observed?

Each offer keeps its own observation date. Daily collections and reviewed quotes can have different dates; an older retained quote is labeled rather than made to look new. A site update does not refresh every offer.

Does a posted price mean capacity is available?

No. A published price does not confirm stock, region availability or that an instance can be launched. Check capacity and the final total with the provider.

How are offers compared?

The rental comparison keeps billing models and units explicit. Filter by GPU, memory, provider, region or recorded instance size, then compare matching offers. The lowest listed offer is separate from a market reference price; unknown configuration details stay unknown.