gpu.fund

gpu.fund is an independent, live cloud GPU price index that tracks rental prices across 15 cloud providers, 49 GPU models, and 177+ active listings — updated every hour. It exists to answer one question before you start a job: where is the cheapest place to rent the GPU you actually need right now?

The index covers everything from consumer-grade RTX 3080s at $0.07/hr to enterprise Blackwell B300s and GB300s at the top end, across providers including Vast.ai, RunPod, Hyperstack, CoreWeave, AWS, Azure, Google Cloud, Nebius, Paperspace, Together.ai, Verda, Vultr, TensorWave, Latitude, and DigitalOcean. Every listing shows the real hourly cost, monthly equivalent, VRAM, GPU count, availability status, and region — sortable and filterable in one table.

Beyond the price table, gpu.fund includes a GPU Finder that recommends the right GPU for your workload, exact price drop alerts with no account required, a cost calculator to estimate spend before a long job, and a direct GPU sourcing service for teams that need reserved capacity or bulk deals.

Built for AI engineers, ML researchers, indie developers, and anyone who rents cloud compute and wants to stop overpaying.

Key Features

📊 Live Price Index — Updated Every Hour 177+ listings across 15 providers and 49 GPU models, refreshed hourly. Every row shows GPU model, architecture tier (consumer / prosumer / enterprise / AMD), provider, real hourly rate, monthly equivalent, 7-day trend, VRAM, GPU count, availability, and region. Sort by price, VRAM, or provider. Filter by GPU tier, specific model, specific provider, or price range. The cheapest listing for each major GPU is highlighted automatically.

🔍 GPU Finder — Workload-Matched Recommendations Tell gpu.fund what you’re doing — LLM inference, fine-tuning, image/video generation, training, cheapest CUDA box, or reliable production — and it ranks live listings by workload fit, price, VRAM, and provider reliability. Set your priority (balanced, lowest price, reliability, or performance), max hourly budget, minimum VRAM, estimated runtime, and preferred region. The finder returns ranked recommendations with a plain-English reason for each pick.

🔔 Exact Price Drop Alerts — No Account Required Set a target hourly price for any GPU model and leave an email. gpu.fund pings you once when that target is hit — no account, no provider spam, no recurring emails. The default target is set 10% below the current best live price, adjustable before saving. Works for any of the 49 tracked GPU models.

💰 Cost Calculator Estimate total spend before committing to a long job. Input GPU model, hourly rate, and estimated runtime to get a projected cost before the job starts — avoiding surprise bills on multi-day training runs.

📈 Price Trends 7-day price trend data on every listing. See whether a GPU’s price is rising, falling, or stable before you rent — useful for timing a job or deciding whether to wait for a better rate.

⚡ Cheapest GPU Highlights Three live callouts on the homepage: cheapest H100 right now, cheapest A100 right now, and cheapest RTX 4090 right now — with provider, price, and a direct link to the deal. Updated continuously.

🖥️ Per-GPU and Per-Provider Pages Dedicated pages for each GPU model (H100 prices, A100 prices, RTX 4090 prices, cheapest GPUs) and each provider (Vast.ai, RunPod, Nebius, Hyperstack, Verda rows) — useful for bookmarking a specific comparison or sharing a filtered view.

🤝 Direct GPU Sourcing For teams that need more than the public listings offer — reserved capacity, short leases, bulk deals, or help underwriting a cluster — gpu.fund has provider and hardware-side connections. Submit a spec (H100, A100, RTX 4090, MI300X clusters) and they source directly.

GPU Models Covered

Enterprise: GB300, B300, B200, GB200, H200 SXM, H200 NVL, H100 SXM5, H100 PCIe, A100 SXM4 80GB, A100 PCIe 80GB, A100 PCIe 40GB, A40, A10G, L40S, L40, L4, V100 32GB, V100 16GB

AMD: MI355X, MI350X, MI325X, MI300X, MI300A, MI250X, MI250, MI210

Prosumer: RTX PRO 6000 Blackwell, RTX PRO 5000 Blackwell, RTX PRO 4500 Blackwell, RTX PRO 4000 Blackwell, RTX 6000 Ada, RTX A6000, RTX A5000, RTX A4000, RTX 5090, RTX 5080, RTX 4090, RTX 4080, RTX 3090 Ti, RTX 3090

Consumer: RTX 5070 Ti, RTX 5070, RTX 5060 Ti, RTX 4070 Ti, RTX 4070, RTX 4060 Ti, RTX 3080, RTX 3070, AMD RX 7900 XTX

Providers Tracked

AWS · Azure · CoreWeave · DigitalOcean · Google Cloud · Hyperstack · Latitude · Nebius · Paperspace · RunPod · TensorWave · Together.ai · Vast.ai · Verda · Vultr

Who Is gpu.fund For?

✅ AI engineers and ML researchers who rent cloud GPUs regularly and want to find the cheapest option before each job ✅ Indie developers and solo builders running LLM inference, fine-tuning, or image generation on a budget ✅ Startups and small teams who can’t afford to overpay on AWS or Azure when cheaper options exist ✅ Data scientists who need to estimate compute costs before committing to a long training run ✅ DevOps and infrastructure teams managing cloud GPU spend across multiple providers ✅ Anyone who has ever refreshed multiple provider pricing pages manually to find the best rate

Pros

• ✅ Completely free — no account, no signup, no paywall
• ✅ 177+ live listings across 15 providers updated every hour
• ✅ 49 GPU models covered from consumer RTX to enterprise Blackwell and AMD
• ✅ Price drop alerts with no account required — one email, one ping
• ✅ GPU Finder matches workload type to the right live listing
• ✅ Covers both hyperscalers (AWS, Azure, GCP) and neo-cloud providers (Vast.ai, RunPod, Hyperstack)
• ✅ Per-GPU and per-provider pages for easy bookmarking
• ✅ Direct sourcing service for reserved capacity and bulk deals
• ✅ Independent — not affiliated with any provider
• ✅ 7-day price trend data on every listing

Cons

• ⚠️ 15 providers tracked — does not cover every cloud GPU provider (Lambda Labs, Spheron, Oracle, and others are absent)
• ⚠️ Prices are indicative — actual billed cost may vary with storage, bandwidth, and spot vs on-demand terms
• ⚠️ 7-day trend data is still collecting on many listings — historical depth is limited for newer entries
• ⚠️ No API or programmatic price feed for developers who want to query prices in their own tools
• ⚠️ GPU Finder recommendations are based on live listings only — does not account for provider uptime history or SLA differences

FAQs

Q: What is gpu.fund? 

gpu.fund is a free, independent live cloud GPU price index that tracks rental prices across 15 cloud providers, 49 GPU models, and 177+ active listings — updated every hour. It lets you compare real hourly and monthly costs for GPUs including H100, H200, B200, B300, A100, RTX 4090, and MI300X before you start a job, so you can find the cheapest available option without manually checking each provider.

Q: Which cloud GPU providers does gpu.fund track? 

gpu.fund currently tracks 15 providers: AWS, Azure, CoreWeave, DigitalOcean, Google Cloud, Hyperstack, Latitude, Nebius, Paperspace, RunPod, TensorWave, Together.ai, Vast.ai, Verda, and Vultr. Every listing shows the real hourly rate, monthly equivalent, VRAM, GPU count, availability, and region.

Q: How often are gpu.fund prices updated? 

gpu.fund updates its price index every hour. The timestamp of the last update is shown at the top of the page so you always know how fresh the data is.

Q: How do price drop alerts work on gpu.fund? 

Select a GPU model, enter your target hourly price, and leave your email address. gpu.fund will send you one email when that target price is hit — no account required, no recurring emails, and no provider spam. The default target is set 10% below the current best live price, and you can adjust it before saving the alert.

Q: What is the GPU Finder on gpu.fund? 

The GPU Finder is a workload-matching tool that recommends the best live GPU listing for your specific use case. Tell it what you are doing — LLM inference, fine-tuning, image or video generation, training, cheapest CUDA box, or reliable production — and set your priority, maximum hourly budget, minimum VRAM, estimated runtime, and preferred region. It returns ranked recommendations from live listings with a plain-English reason for each pick.

Q: Is gpu.fund free to use? 

Yes. gpu.fund is completely free. No account, no signup, and no paywall. The full price table, GPU Finder, price drop alerts, cost calculator, and all provider and GPU pages are accessible without creating an account.

Q: What GPU models does gpu.fund cover? 

gpu.fund tracks 49 GPU models across enterprise, prosumer, consumer, and AMD tiers. Enterprise models include GB300, B300, B200, H200, H100, A100, A40, A10G, L40S, L40, L4, and V100. AMD models include MI355X, MI325X, MI300X, and MI250X. Prosumer models include RTX PRO 6000 Blackwell, RTX 5090, RTX 4090, RTX A6000, and RTX 6000 Ada. Consumer models include RTX 5070, RTX 4070, RTX 3090, RTX 3080, and RTX 3070.

Q: What is the cheapest H100 available on gpu.fund? 

The cheapest H100 SXM5 80GB tracked by gpu.fund is currently $2.00perhourviaTogether\mathrm{.}ai,equivalenttoapproximately$1,460 per month. Prices change frequently — check gpu.fund for the live current best price.

Q: Can gpu.fund help me source reserved or bulk GPU capacity? 

Yes. For teams that need more than the public listings offer — reserved capacity, short leases, bulk deals, or help underwriting a cluster — gpu.fund has provider and hardware-side connections. You can submit a GPU spec (H100, A100, RTX 4090, MI300X clusters) directly through the site and they will source options on your behalf.

Q: Are the prices on gpu.fund the final billed cost? 

gpu.fund shows the advertised hourly rental rate for each listing. Actual billed costs may vary depending on storage fees, bandwidth charges, spot vs on-demand pricing terms, and any provider-specific surcharges. gpu.fund notes that prices are indicative and recommends reviewing each provider’s full pricing terms before committing to a job.

How We Rated It