Vercel Data Cache to Runtime Cache: what AI-built app founders should do before migration

Vercel Runtime Cache is replacing Data Cache, and for founders who shipped a Lovable, Bolt, Cursor, Replit or v0 app to Vercel, that is a billing and behavior change you should plan for — not a rename you can ignore. On October 30, 2026, Vercel published that Data Cache is exiting beta and moving to Runtime Cache, with Next.js fetch caching and unstable_cache switching over without code changes, larger per-project storage on paid plans, and — critically — cache reads and writes that were previously unbilled starting to draw from Runtime Cache pricing after you migrate.

If your Vercel bill already feels noisy, treat this as a checkpoint. Our Vercel bill too high review is the offer path when you want a human to map which lines are growing. Below is a founder-facing brief: what changed, who is affected when, and a practical checklist you can run this week.

What Vercel announced (Oct 30, 2026)

From Vercel's own changelog:

• Data Cache moves to Runtime Cache. Same Next.js caching APIs keep working; the storage layer underneath changes.
• More room on paid plans. Pro gets a much larger Runtime Cache allocation per project and environment than Data Cache's old team-wide pool (Vercel cites 8×). Enterprise gets a larger bump (Vercel cites 10× versus prior Data Cache storage).
• Billing changes after migration. Cache reads and writes that were unbilled under Data Cache are charged at Runtime Cache rates once your team is on the new system. Vercel publishes regional unit rates on the changelog; check that page for current numbers rather than trusting a screenshot in a blog post.
• Hobby free allowance. Hobby teams get a monthly free allotment of read and write units across projects (amounts are listed on the changelog).
• Pro credit. On Pro, usage draws from the team's infrastructure credit first, then bills at regional rates.

None of that requires you to rewrite your app for the migration itself. It does require you to understand whether your app is a quiet cache user or a chatty one.

Who moves when

Per the same changelog:

• New teams created on or after October 30, 2026 already use Runtime Cache by default.
• Existing Hobby teams are scheduled to migrate on November 29, 2026.
• Existing Pro teams get a notified migration date.
• Enterprise is contacted by the account team.
• Early opt-in is available from October 30 via team CDN settings. Pricing applies as soon as you opt in — so "try it early" is not free curiosity if you are already cache-heavy.

If you are a Hobby founder with an AI-built SaaS on a custom domain, November 29 is the date on your calendar. If you are on Pro, watch for the email and the dashboard notice; do not assume the Hobby date applies to you.

Why AI-built apps should care

Vibe-coded Next.js apps often lean on default caching more than a hand-tuned app does:

1. AI builders love `fetch` in Server Components. Cached fetches that felt "free" under Data Cache become metered reads and writes after migration. 2. Preview + production both count. Pro allocations are per project and environment, which is better isolation than a shared team pool — and also means a noisy preview environment has its own billable footprint once you are on Runtime Cache pricing. 3. Chatty dashboards. Admin pages that re-fetch large lists on every navigation can amplify cache writes if something keeps missing or revalidating. 4. Bots and scrapers. An open marketing page that triggers server work on each hit can turn into cache churn you never intended. 5. You may already be watching the wrong lines. Edge Requests, Function Duration and ISR/revalidation noise showed up in earlier bill-shock posts; Runtime Cache is an additional line to learn before it appears as a surprise.

This is not an argument against caching. Caching still protects your origin and your Supabase bill. It is an argument for measuring cache traffic before Vercel flips your team.

A 45-minute founder checklist (do this before you opt in)

Step 1 — Find whether you use Data Cache today

In the Vercel dashboard, open the project that hosts the AI-built app. Check Observability / cache-related views for Data Cache usage. If the charts are flat empty, you may feel little after migration. If you see steady read/write volume, keep going.

Step 2 — Note preview vs production

AI builders often deploy every prompt iteration. Confirm which environment shows cache activity. A frantic preview can look "busy" while production is calm — or the reverse.

Step 3 — List the chatty routes

From your own knowledge of the app, write down:

• Pages that load large tables (CRM lists, message inboxes, admin user lists).
• Pages that call Supabase on every load without pagination.
• Anything that "revalidates" on a short timer.
• Image-heavy or JSON-heavy responses that might be cached at the data layer.

You do not need perfect telemetry for a first pass. You need a shortlist of suspects.

Step 4 — Decide: optimize first, or wait for the scheduled migration

• Optimize first if Observability already shows heavy Data Cache traffic, or if last month's Vercel usage was already climbing.
• Wait for the scheduled migration if usage is tiny and you have alerts + a spend ceiling in place — but still complete Steps 5–7.
• Do not casually opt in early on a production team just to "see the new UI" unless you have reviewed usage. Early opt-in starts Runtime Cache pricing immediately.

Step 5 — Cut obvious waste (safe edits)

Without inventing micro-benchmarks:

• Paginate large list fetches.
• Avoid caching personalized responses that must differ per user (cache shared public data; keep private data keyed correctly or uncached).
• Lengthen revalidation where a five-second freshness window is vanity, not product need.
• Stop double-fetching the same resource from layout and page.

Vercel's changelog points to an optimization guide and an agent prompt you can paste into a coding agent. Use those against your repo; then re-check Observability.

Step 6 — Put a ceiling on the bill

Before November 29 (Hobby) or your Pro date:

• Confirm Spend Management / spend alerts are on for the team.
• Confirm you will get notified before a bad weekend becomes an invoice story.
• Re-read which lines you already capped after earlier bill work (see Vercel bill shock: the 7 usage lines).

Runtime Cache does not replace those other lines. It sits beside them.

Step 7 — Write a one-line rollback plan

If you opt in early and cache cost spikes: who turns what off, and where? Vercel notes you can disable fetch and unstable_cache caching in CDN settings (which pushes more load to origins — including Supabase — so it is a trade, not a free escape). Know that trade before you need it.

How this interacts with Supabase cost

Cutting Vercel cache traffic sometimes increases origin reads against Supabase. That can move spend from one vendor to the other. When you change caching:

1. Watch Supabase egress and database CPU the same day. 2. Prefer fixing chatty queries (select only needed columns, paginate, avoid N+1 patterns the AI generated) over simply disabling cache. 3. If both bills are climbing, you want a combined pass — host and database — not two separate panics a week apart. That is the spirit of a bill cut review when the stack is Vercel + Supabase.

What not to do this week

• Do not paste invented unit prices into investor updates; link Vercel's changelog.
• Do not grant a coding agent production env write access just to "flip CDN settings" without a human watching.
• Do not disable all caching as a first move on a launch-week app — you may trade a small cache line for a large Functions + Supabase spike.
• Do not ignore the Hobby November 29 date because "we barely use cache." Confirm in Observability.

Practical takeaways

• Oct 30, 2026: Data Cache → Runtime Cache is official; APIs stay, billing changes after migration.
• Hobby existing teams: November 29, 2026. Pro: watch your notice. New teams: already on Runtime Cache.
• Early opt-in starts paid Runtime Cache pricing immediately.
• Measure Data Cache usage before you opt in; optimize chatty routes first.
• Keep spend alerts on; remember origin load can shift to Supabase if you disable caching bluntly.
• Pair this news with the older seven usage lines guide so Runtime Cache is not the only line you understand.

FAQ

Do I need to change my Next.js code for Runtime Cache?

Vercel says fetch caching and unstable_cache switch over without code changes. You may still change code to reduce unnecessary cache traffic once you see the meters.

Is this only a problem for huge apps?

No. Small AI-built apps with aggressive revalidation, admin list pages or bot traffic can create surprising cache churn. Size of the company is a weak predictor; chattyness is a strong one.

Should every founder opt in early?

No. Opt in early when you want the new allocation model and you have already reviewed usage. Otherwise wait for your scheduled migration and use the time to clean hot routes.

Where do I verify the current rates?

Vercel's October 30, 2026 changelog entry for the Data Cache → Runtime Cache move. Rates vary by region; use the primary source, not a secondary summary.

---

If your Vercel usage is already hard to explain — or you want a senior pass before the Hobby migration date — start at Vercel bill too high. Book a free technical assessment and we will help you see which lines deserve a ceiling before Runtime Cache joins them.