AI builders
Compare VRAM-heavy cards quickly when you care about local inference, CUDA access, or open-model experimentation.
Public MVP
Compare GPU prices and stock without the usual retailer tab chaos.
gpu.watch gives enthusiasts a clean market view: seeded pricing, stock status, and recent movement across major retailers, with architecture ready for live feeds later.
Tracked GPUs0
Retailers0
In-stock listings0
Who it is for
Built for practical enthusiasts
The first release focuses on public, indexable market pages instead of accounts, alerts, or scraping infrastructure.
Gamers
Scan current pricing and stock without wading through noisy retailer merchandising pages.
Workstation shoppers
Keep an eye on high-memory options that make sense for rendering, creator tools, and mixed-use rigs.
Featured cards
Useful starting points for browsing the seeded catalog
A mix of halo GPUs, sensible 16GB picks, and value options for lighter builds.
The catalog is ready, but the database is still empty.
Thread 2 moved the public site onto persisted Postgres data, so local development now needs the Supabase stack plus the bootstrap seed step.
supabase startsupabase status -o envnpm run db:migratenpm run db:seedSet NEXT_PUBLIC_SUPABASE_URL and SUPABASE_SECRET_KEY in .env.local before running the seed step. You can map them from API_URL and SECRET_KEY in `supabase status -o env`.
Recent price drops
Where the seeded market moved the most
These examples are driven from recent snapshots so the homepage already reflects trend-aware data shaping.
No persisted price moves yet.
Once the database is seeded, recent price drops will be derived from the persisted snapshot history instead of the in-memory seed module.
How it works
A deliberately clean first release
This thread is about public architecture and SEO, not live data plumbing.
1. Seed the catalog
GPU core records, retailer offers, and price history snapshots all live in a local data layer that mirrors the future live-feed shape.
2. Derive the market view
Lowest price, average price, best in-stock offer, stock counts, and recent drops are isolated in selectors so they can be tested and reused later.
3. Publish indexable pages
Each route ships with metadata, canonical URLs, and structured data so live retailer ingestion can plug into a credible public surface later.