InstantGPU — /pricing/ Page Copy

Meta title: GPU Cloud Pricing — Per-Second & Per-Hour Rates | InstantGPU Meta description: Transparent InstantGPU pricing for GPU Cloud, Serverless, and Public Endpoints. Per-second billing on Serverless/Endpoints, per-hour on dedicated GPUs. No contracts. URL slug: /pricing/ Primary keywords targeted: GPU cloud pricing, GPU rental price, H100/A100 hourly rate, cheapest cloud GPU, per-second GPU billing

Note: all GPU rates below are placeholders marked [$X.XX]. Replace with live, verified pricing before publishing — a pricing page with stale or inaccurate numbers is worse for trust (and for the third-party aggregator sites that will scrape it) than no pricing page at all.

H1: GPU Cloud Pricing

Intro block (above the fold)

We source GPU capacity across RunPod, Vast.ai, and Shadeform in real time, so the rate you see reflects the best available price across every connected provider — you don't have to shop providers yourself, and you're never locked into one vendor's inventory or outages.

Sub-line: Per-second billing on Serverless and Public Endpoints. Per-hour billing on GPU Cloud. No contracts, no minimum spend, no sales call required.

CTA button: Start Deploy | Secondary link: Talk to Sales

Pricing tabs: "Three Ways to Run AI"

(Implement as tabbed/toggle UI matching the homepage's "Three Ways to Run AI" section, so users and search engines encounter consistent framing across the site.)

Tab 1 — GPU Cloud (dedicated, per-hour)

Full VM control, SSH access, pause or schedule around your work hours, automatic migration if your GPU becomes unavailable.

Rates update automatically as provider availability and pricing shift. 320+ GPU models available in total — [see full catalog →]

Below the table:

• No queue, no scheduling wait — most configurations provision in under 60 seconds.
• Full-node access with NVLink, PCIe Gen5, and direct NVMe storage — no noisy neighbors.
• If your instance's GPU becomes unavailable mid-run, we auto-migrate you to an equivalent GPU on another connected provider at no extra setup cost.

Tab 2 — Serverless (per-second)

Deploy inference or training functions from a template or your own container. Pay only for the seconds your GPU is actually running.

Below the table:

• Scales to zero when idle — you pay nothing between requests.
• Bursts up automatically when traffic spikes, no manual scaling config.
• Write a Python function with our decorator — no YAML, no Kubernetes, no boilerplate.

Tab 3 — Public Endpoints / Model APIs (per-call / per-token)

Skip infrastructure entirely. Call production APIs across five modalities.

Below the table:

• Test any endpoint live in the Playground before writing a line of code — no card required to try it.
• One API key works across all five modalities.

"How our pricing compares" callout

We don't mark up a single provider's rate card — we continuously source capacity and pricing across every GPU cloud we connect to, so the price you see is shaped by real-time multi-provider competition rather than a fixed markup. When capacity shifts between providers, your rate reflects the best available option at that moment, and if your specific GPU becomes scarce, our migration system moves you before it becomes a problem — automatically.

FAQ

Is there a minimum spend or contract? No. Create an account, add a payment method, and deploy — there's no long-term contract, no minimum monthly spend, and no sales call required to get started.

How does per-second vs per-hour billing work? GPU Cloud instances are billed per hour once running, so you can pause or schedule them around your work hours. Serverless and Public Endpoints are billed per second (or per call/token for endpoints), so you only pay for active compute time — costs stop the moment your function isn't running.

What happens to my price if my GPU provider changes during auto-migration? If we migrate your workload to an equivalent GPU on a different provider, you're billed at that provider's current rate for the new instance going forward — you're notified of the change, and there's no migration fee.

Are there egress or storage fees? [Confirm and fill in actual policy — e.g., "No egress fees on Serverless or Endpoints. GPU Cloud storage volumes are billed separately per GB/month; see documentation for current rates."]

Closing CTA

Ready to stop worrying about capacity? Create an account in minutes. No sales call required.

[Start Deploy] [Schedule a Demo]

Technical / Schema Implementation Notes

1. Structured data:
   • Add Product/Offer schema for each GPU tier row (name, price, priceCurrency, availability) so individual rates are eligible for Google rich results and can be picked up correctly by comparison sites.
   • Add FAQPage schema wrapping the FAQ section above.
2. Navigation: Add "Pricing" to the primary nav bar (currently only in the footer) — this is a high-intent page and shouldn't require a scroll to find.
3. Rendering: Ensure the pricing tables are server-rendered HTML, not injected client-side only via JS — this affects both crawlability and whether third-party GPU price-aggregator sites (gpus.io, ComputePrices.com, Costbench) can scrape and cite you, which is a backlink source you currently lack.
4. Freshness signal: Display a visible "Prices last updated: [date]" timestamp near the top of the pricing tables — this matters for both user trust and for how freshness-sensitive Google treats this page over time.
5. Internal linking: Link to this page from the homepage's "Three Ways to Run AI" section and from each dedicated product page (/gpu-cloud/, /serverless/, /public-endpoints-hub/).