Why We Migrated to Astro (And Why You Should Consider It)

For years we assumed something we never questioned: if we build a website, we have to ship JavaScript to the browser. Lots of it. React, Vue, full hydration, 300KB bundles just to render a static landing page. We normalized it.

Astro breaks that. And once you get it, it’s hard to go back.

The paradigm: zero JS by default

The core idea of Astro is simple: HTML is generated at build time, and not a single byte of JavaScript is sent to the client unless you explicitly ask for it. Interactive islands (client:load, client:visible, client:idle) are the exception, not the rule.

Compare that to Next.js or Nuxt, where hydration is the default. There, you fight the framework to not ship JS. In Astro, you fight to ship it. The incentive is inverted.

What we actually gained

Migrating the Nakea website from Next.js to Astro gave us:

• Lighthouse Performance: 98-100 across every page, no tricks.
• Build times 3x faster: from ~45s down to ~14s.
• Initial bundle ~0KB on static pages (blog, about, legal).
• Surprisingly good DX: .astro components that feel like JSX without the complexity of a reactive runtime.

And something harder to measure: the feeling that the framework gets out of your way. You write HTML, CSS, and a bit of TS. No weird rules, no memorizing 15 hooks.

When NOT to use Astro

Let’s be honest. Astro isn’t the answer to everything:

• Highly interactive apps (dashboards, editors, SaaS tools). React/Vue still win there.
• Complex global state shared across pages. Islands don’t trivially share state with each other.
• Heavy dynamic SSR with lots of per-request data. You can do it, but it’s not the sweet spot.

Astro shines for content sites: landings, blogs, docs, marketing, e-commerce with targeted islands. That’s about 70% of what most studios actually build.

The bottom line

Don’t migrate because of hype. Migrate because the paradigm fits what you’re building. If your site ships 500KB of JS to render static text, something’s wrong. Astro forces you to ask what actually needs to be interactive.

And that question, by itself, is already worth the switch.