next-server consuming over 600% of my CPU locally

Currently when running pnpm dev instead apps/web in my Turborepo monorepo the next-server is consuming over 600% of my CPU. For context, I have an M1 Pro MacBook Pro with 16 GB of RAM.

It is also using roughly 12 GB of RAM as well

screenshot_2026-05-21T15:16:12

Same here.

Also experiencing this on an idle dev server. Hope this is resolved soon.

Same here. M4 Pro, 48GB Ram. Bought so much ram to finally have nextjs server work smoothly. But it’s still a huge pain.

Same here, over 800% CPU usage on my M1 macbook pro. Running v16.2.6 as well.

Facing this same issue today on M4 16GB Ram with above 600% cpu usage. I see no workaround here yet but documenting.

Please run the following command immediately

"dev:clean": "rm -rf .next && next dev --turbopack"

Root Cause

The .next cache was corrupted.

Why This Happens

Turbopack has a different cache structure from webpack, and the cache tends to get corrupted in the following situations:

  • Repeatedly switching between webpack ↔ Turbopack causes cache conflicts

  • After changing dependencies, the .next cache still references the old module graph

  • After a Turbopack version upgrade, the old cache becomes incompatible

Hi James,

I’d treat the .next cache suggestion as a good first check, but not the whole diagnosis yet. If deleting .next fixes it only temporarily, the cache may be a symptom rather than the root cause.

I’d test these separately from apps/web:

rm -rf .next
pnpm dev

Then compare Turbopack vs webpack:

pnpm dev --webpack

If CPU/RAM is normal with --webpack but spikes with the default dev server, that points more toward a Turbopack/module graph issue. In that case I’d check for things that make the dev bundler crawl too much of the monorepo: large barrel exports, importing from a package root that re-exports everything, icon libraries imported broadly, generated folders included in source globs, or custom Turbopack/webpack config.

If both Turbopack and webpack spike, I’d look outside the bundler first: Tailwind content globs scanning too much, a watcher loop from generated files, or a server component / route doing expensive work during dev compilation.

Next’s local dev performance guide has a useful checklist for this:
https://nextjs.org/docs/app/guides/local-development

Can you share whether pnpm dev --webpack has the same CPU/RAM behavior after deleting apps/web/.next?