Why this exists

Cloud GPU prices move. Catalog specs and VRAM estimates are useful only after you can see what the market is asking today.

HardwareHQ is a live AI hardware market desk: instance prices first, then catalog specs, then fit estimates. Paying does not change catalog numbers, fit results, or cloud sort order.

Core tools are free with no signup. Revenue comes from disclosed affiliate links (cloud GPUs and hardware retailers) and optional labeled sponsorships.

What we offer

Live cloud prices

Hourly instances where a provider feed works. Duplicate SKUs are collapsed. If a number is unknown, we omit it.

GPU specs

Consumer and enterprise cards from the compiled catalog. Specs, not invented street prices.

AI model catalog

Parameter counts and estimated VRAM. Homepage, VRAM calculator, and /can-it-run share one estimator (default Q4_K_M, 4k context, batch 1).

Fit estimates

Supporting tool. Instant fit vs the estimator. Estimates, not benches.

Data sources and limits

GPU catalog — compiled from vendor specs and internal IDs. Counts on this page, the homepage, /gpus, /llms.txt, and GET /api/v2/gpus come from the same unique catalog via getSiteStats().
VRAM / fit — one shared estimator (Q4_K_M · 4k · batch 1). For dense 7B–70B the locked total sits ~5–8% above the Q4_K_M GGUF file. It is a planning number, not peak runtime VRAM and not a measured run. Implausible 70B-class Q4 figures are hidden rather than shown as 8B-sized. MoE / PLE load estimates use total (storage) parameters; active or effective counts are architecture metadata.
Cloud — provider APIs and listings, live when the feed works. Not every region or SKU is covered.
Retail — street prices stay gated until a quote is validated. Launch MSRP is not presented as what you would pay today. We do not invent a 5090 at $599.
Cadence — feeds refresh as sources allow. We do not promise same-day coverage of every new model.

Transparency

No account required

Core tools work without signup. Optional Google/GitHub sign-in exists for account features; it is not a paywall.

What we measure

We run first-party product events (compatibility checks, tool opens, studio and guide clicks, outbound hops) plus Vercel Analytics. Google Analytics fires only after you accept cookies — Consent Mode defaults to denied. We do not sell a visitor list or your hardware choices. Privacy policy.

Affiliate disclosure

Some hardware and cloud links are affiliates. We may earn a commission at no extra cost to you. Links are disclosed and do not change fit results, catalog counts, or organic cloud ranking.

Public API

GPU and model data is available via REST API for free use. Rate limits may apply to protect the service. For higher-volume access, email hardwarehq.dev@gmail.com. Core web tools stay free for everyone.

GET/api/v2/gpus— Unique GPU catalog (same count as this page)

GET/api/hardware-database— Hardware specs

GET/api/ai-models— AI model data

Contact

Found a bug? Have a feature request? Want to contribute data?

hardwarehq.dev@gmail.com

HardwareHQ has been a free AI hardware desk since 2024.

HardwareHQ

About HardwareHQ

Live cloud GPU prices, catalog specs, and fit estimates. Market tracking is the primary job. Fit tools are supporting. Core tools stay free — no signup. Catalog currently covers 909 GPUs, 313 AI models, 24 cloud providers.