Glossary term

Sub-Index

A smaller index carved out of a Closelook index by sector, layer, region or tier — for example the Rubin index’s memory, interconnect or Japan groups — each computed with the same equal-, cap- and modified-weight methods as its parent. Sub-indices show where inside a theme the money moved.

AI-generated — produced automatically by Closelook’s systems under this site’s editorial policy.

What it means

A headline index answers whether a theme worked; its sub-indices answer where. The Rubin Build-Out 100 publishes eight layer sub-indices (from architects and foundry to data-centre systems), around twenty sector sub-indices (memory, packaging, lithography, interconnect, power semiconductors, thermal, grid and so on) and four regional ones (US, Japan, Europe, Asia ex-Japan); HALO, Euro-AI, the Agentic Winners and the Agentic Ecosystem publish their own by sector, tier and region. Each uses the parent’s three weighting methods and the same quarterly anchors.

Small sub-indices are labelled by member count — a group of one or two names is shown as such, never hidden — because a two-name sub-index is a pair of stocks, not a sector.

Why it matters for the AI trade

The sort inside the AI trade is invisible at the index level and obvious one layer down. On 14 September 2026 the Rubin index fell 5.25% while its sub-indices ranged from interconnect −8.5% and data-centre power −7.2% to architects −1.4% and wafers flat; by region the US constituents fell 7.0% and Japan’s 0.5%. That dispersion is the reading — a valuation event in the parts of the stack hardest to reprice quickly, not a demand event across the board.

How Closelook uses it

Each index dashboard — Rubin, HALO, Euro-AI, AW40, Agentic Ecosystem — carries a sector grid, a chart selector by sub-index and a rotation radar; every stock page shows its own sub-index as a benchmark line. The weighting entry explains the three methods.

Common questions

How many sub-indices does the Rubin index have?
Eight by layer, about twenty by sector and four by region, each in equal-, cap- and modified-weight form, plus the headline in all three.
What happens when a sub-index has only one or two members?
It is published with its member count in the label — “(1 name)” or “(2 names)” — and read as the stocks it contains rather than as a sector. Nothing is hidden, because the gaps themselves say something about the theme’s public listings.
Are sub-indices rebalanced?
At the same quarterly anchors as the parent, with the same cap rules. Between anchors the weights drift with prices, as in any index.