Supercomputers/Explorer
Faceted query
Explorer
The graph, queried directly. Filter by the role a company played, the corporate parent it rolls up to today, country, era or specific part. Company matching follows the acquisition chain, so filtering on HPE returns machines shipped by Cray and SGI, and filtering on NVIDIA returns fabrics shipped by Mellanox.
How the explorer works, and what happens when it doesn't
This is the only route on the site that runs server-side. It calls
/api/explorer, a single Worker endpoint backed by D1, cached with
stale-while-revalidate. Every other page — systems, sites, parts, companies,
analysis — is pre-rendered to static HTML at build time from the repo CSVs and served
from Cloudflare's asset store.
That split is deliberate. If D1 is unavailable, this page degrades to the message above and the rest of the site is unaffected. The database is a query convenience, not the source of truth; the repository is.