Server-side rendering services decide where and when each page of your site or app is turned into HTML, then build and tune that setup. We choose between server rendering, static generation, incremental regeneration and client rendering route by route, so pages load fast, rank in search and do not cost more to host than they need to.
SSR, SSG, ISR and client rendering compared
Four rendering modes cover almost every page. Most real projects use at least three of them.
| Mode | When the HTML is built | Best for | Watch out for |
|---|---|---|---|
| Server-side rendering (SSR) | On every request | Personalized pages, search results, prices and stock that change by the minute | Server cost per request and slow APIs delaying the page |
| Static site generation (SSG) | At build time | Marketing pages, docs, blog posts | Long builds on large sites, content stale until the next build |
| Incremental static regeneration (ISR) | At build time, then refreshed in the background | Large catalogs and CMS pages that change through the day | Choosing revalidation times and purging on publish |
| Client-side rendering (CSR) | In the browser, after JavaScript loads | Dashboards and tools behind a login | Blank first view and weak SEO for public pages |
In Next.js all four modes live in one app and can be set per route. Astro is static by default and renders individual pages or server islands on demand when they need fresh data.
Streaming and React Server Components
Classic SSR waits for every data call before it sends the first byte. One slow API holds up the whole page. Streaming changes that: the server sends the layout and fast content at once, then streams in slower sections as their data arrives, using Suspense boundaries as the split points.
React Server Components go a step further. Components that only display data run on the server and send HTML, so their code never reaches the browser. Only interactive parts, such as forms, filters and charts, ship JavaScript. On data-heavy pages that cuts the JavaScript the browser has to download and parse.
We decide where the Suspense boundaries go, which components stay on the server and which need the client. Those choices decide what a visitor sees first.
How rendering affects SEO and performance
Search engines and AI crawlers read the HTML they receive. Google can render client-side JavaScript, but it does so later and not always fully, and many other crawlers do not run JavaScript at all. Pages rendered on the server or at build time give every crawler the full content, links and structured data in the first response.
For speed, the rendering mode mostly shows up in Largest Contentful Paint. Server or static HTML lets the browser paint the main content early. Hydration then has to finish before the page responds to input, so too much client JavaScript can still hurt Interaction to Next Paint on a server-rendered page. Our guide to Core Web Vitals explains each metric and its thresholds.
Caching at the edge
Server rendering without caching sends every visitor to your origin. We set the cache up in layers:
- CDN cache with
Cache-Controlheaders.s-maxagekeeps a page at the edge, andstale-while-revalidateserves the cached copy while a fresh one renders in the background. - Revalidation on publish. When an editor publishes in the CMS, a webhook clears exactly the affected pages instead of the whole site.
- Data cache. API responses that many pages share are cached on the server, so one slow service does not slow every page.
- No shared cache for personal data. Pages that depend on cookies or the logged-in user are marked private so they never reach another visitor.
On Vercel much of this is built into the platform. On AWS or a self-hosted Node.js server we configure the CDN and headers ourselves and document them.
When SSR is overkill
Sometimes the right move is to take server rendering out:
- Pages that change a few times a week: static generation with revalidation on publish gives the same HTML with no server work per request.
- Apps behind a login: search engines never see them, so a client-rendered app is simpler to build, test and host.
- Small content sites: a static Astro site on a CDN has no server to patch or scale.
Our Next.js vs Astro vs Remix comparison shows how each framework handles these trade-offs. If your team has already picked Remix, see our Remix development page.
Proof: static and server rendering in one healthtech replatform
A European enterprise healthtech organization moved off Drupal onto Payload with a Next.js frontend. We used static generation for clinical articles and server-side rendering for dynamic patient-facing pages. LCP came down to 1.5s from 4.3s, with 0 SEO ranking regressions at cutover. Read the Drupal replatform case study.
Get a rendering audit
We start with a short audit: every route, its data sources, how often it changes and what crawlers see today. You get a rendering and caching plan per route before any code changes. Book a call to start, or see our wider frontend development services.