Vercel's Elastic Build Machines now automatically upgrade to a higher memory tier when a build approaches its memory limit, preventing out-of-memory failures without manual intervention.
Three behaviors define the system: fast but memory-intensive builds no longer get downgraded to smaller machines, builds nearing memory exhaustion trigger an automatic tier upgrade mid-run, and builds that already failed from OOM errors are promoted to a higher tier on the next deployment attempt. Thresholds are calculated using only your build's memory consumption, not Vercel's own infrastructure overhead, and are tuned conservatively to balance reliability against cost.
The feature is off by default. Enable it in team or project settings under Build and Deployment. The full changelog details the tier logic and edge cases worth understanding before you opt in.
[READ ORIGINAL →]