Silicon Valley developer market — July 2026
This is the monthly read on where the teardown, spec-build, and value-add opportunity is concentrated across Silicon Valley — the San Mateo County Peninsula and the South Bay (Santa Clara County, including San Jose). We track what the dirt is selling for, where demolition and new-construction permits are clustering, and which live listings pencil after demo-and-rebuild costs are netted out.
Where the activity is. Across the 11 Silicon Valley cities with classifiable permit activity this window, builders pulled 50 demolition and 59 new-construction permits. San Mateo led on combined teardown + new-build activity; Burlingame ran the hottest at ~107% of list.
Coverage note. 10 of the 21 tracked cities are not ranked above: they recorded no classifiable demolition or new-construction permit in this window, so we publish no count for them rather than a zero we cannot stand behind. The reason varies by city and we do not guess at it. Some file work under a generic permit type that never says what the job is (Palo Alto files most work as "Project - Building Permit"; San Jose reports the building type rather than the work), so a demolition there would be invisible to us. Others simply had a quiet window, and others we do not yet ingest fully. Not ranked: Belmont, Burlingame, Campbell, East Palo Alto, Los Altos, Los Altos Hills, Mountain View, San Jose, Santa Clara, Woodside.
What the dirt costs. Recent land buys set the basis: a 1,619 sqft home in San Jose on a 13,068 sqft lot traded at ~$120/lot-sqft. Builders here are buying dirt and entitlements, not houses.
This month's opportunities. Our screen surfaced 10 live candidates worth underwriting as a land-and-rebuild or value-add. The full list with street addresses is below (free with email). These are screened candidates for your own underwriting — we publish the recorded figures, not a projected return. This edition originally carried a model-estimated developer-ROI and teardown score; those were withdrawn on 2026-07-31 because the underlying model has not demonstrated a validated predictive edge.