Dora NodaIf your DigitalOcean App Platform app has been running since before May 7, 2024, you're on a "legacy" pricing plan — and DigitalOcean's own documentation says it "intends to deprecate legacy plans soon." No calendar date. No fixed cutover. Just a plan to move you off a tier you may have never looked at twice, at a time DigitalOcean picks, not you.
The good news, mostly, is that the new plan structure is a genuine upgrade: same-or-lower price on most tiers, and egress allowances that are 2-8x larger across the board. The catch isn't the math — it's that nobody who's been quietly running a legacy app for two years gets to choose when that migration happens to their production traffic. Here's the exact bill difference, tier by tier, and what "your plan is safe" actually means when there's no grandfather clause behind it.
The Full Legacy-vs-New Math, Tier by Tier
DigitalOcean's legacy structure had two tiers — Basic (shared CPU, fixed instance, no manual scaling) and Professional (shared or dedicated CPU, manual scaling) — each capped at a flat bandwidth allowance regardless of instance size: 40 GiB for every Basic plan, 100 GiB for every Professional plan. The current structure drops the tier names entirely and prices by instance type, with egress scaled to the instance size itself. Mapping every legacy plan to its closest current equivalent:
| Legacy Plan | Legacy Price | Legacy Egress | Current Equivalent | Current Price | Current Egress | Price Δ | Egress Δ |
|---|---|---|---|---|---|---|---|
| Basic-XXS (512 MiB, shared) | $5/mo | 40 GiB | Shared Fixed 512 MiB | $5/mo | 50 GiB | $0 | +25% |
| Basic-XS (1 GiB, shared) | $10/mo | 40 GiB | Shared Fixed 1 GiB | $10/mo | 100 GiB | $0 | +150% |
| Basic-S (2 GiB, shared) | $20/mo | 40 GiB | Shared Scalable 2 GiB | $25/mo | 200 GiB | +$5 | +400% |
| Basic-M (4 GiB, shared) | $40/mo | 40 GiB | Dedicated 2 vCPU/4 GiB | $50/mo | 250 GiB | +$10 | +525% |
| Professional-XS (1 GiB, shared) | $12/mo | 100 GiB | Shared Scalable 1 GiB | $12/mo | 150 GiB | $0 | +50% |
| Professional-S (2 GiB, shared) | $25/mo | 100 GiB | Shared Scalable 2 GiB | $25/mo | 200 GiB | $0 | +100% |
| Professional-M (4 GiB, shared) | $50/mo | 100 GiB | Dedicated 2 vCPU/4 GiB | $50/mo | 250 GiB | $0 | +150% |
| Professional-1L (4 GiB, 1 dedicated vCPU) | $75/mo | 100 GiB | apps-d-1vcpu-4gb | $49/mo | 400 GiB | -$26 | +300% |
| Professional-L (8 GiB, 2 dedicated vCPU) | $150/mo | 100 GiB | apps-d-2vcpu-8gb | $98/mo | 600 GiB | -$52 | +500% |
| Professional-XL (16 GiB, 4 dedicated vCPU) | $300/mo | 100 GiB | apps-d-4vcpu-16gb | $196/mo | 800 GiB | -$104 | +700% |
The honest read is not uniformly rosy, and it shouldn't be flattened into one. Every Professional-tier and dedicated-CPU migration is flat-price-or-cheaper, with egress up 50% to 700%. The two lowest shared-CPU tiers — Basic-S and Basic-M, the plans a solo side project or a cheap staging environment is most likely sitting on — are the only bands where the bill actually goes up, by $5 to $10 a month, in exchange for 4-6x more bandwidth. If you're on Basic-XXS or Basic-XS, you get more for the same money. If you're on Basic-S or Basic-M, you pay a bit more for a lot more headroom. Nobody migrating off Professional tiers, shared or dedicated, ends up worse off on price.
Egress Allowance Pooling — the Catch in "Better"
"Substantially higher egress" is real, but it's not unlimited, and it doesn't work quite the way a bigger number might suggest. DigitalOcean pools data-transfer allowances across every app on a team, on a monthly cycle, and unused allowance from one month doesn't carry into the next. A team running ten small apps under the new structure isn't stacking ten independent 50-250 GiB allowances in isolation from each other's traffic spikes — they're drawing from one shared team-wide pool, sized by the sum of each app's instance allocation.
Overage past the pool is billed at $0.02 per GiB, and DigitalOcean's own announcement of the new structure cites an 80% cut to the old bandwidth overage fee, which is real relief if you were previously blowing through a fixed tier's cap. But "substantially higher" is a monthly, poolable, non-rolling number — worth checking against your actual traffic shape before assuming the new plan quietly solves a bandwidth problem you've been having.
No Fixed Migration Date Is Not a Grandfather Clause
Here's the part the pricing math doesn't cover: DigitalOcean's current public documentation states only that it "intends to deprecate legacy plans soon" and "strongly encourages" moving to the new structure voluntarily. There is no calendar date, no published cutoff, and no forced-migration announcement in DigitalOcean's product blog as of this writing. That's not a criticism of DigitalOcean specifically — it's the structurally honest position any hosted platform is in once it decides an old pricing tier isn't worth maintaining indefinitely. Someone has to eventually flip the switch, and "soon" with no date attached means the switch flips on the vendor's internal schedule, not on a date any tenant can plan around.
For most tenants on this list, per the table above, that eventual move will be a non-event or a modest improvement. But "the math works out fine for most people" and "you get to choose when it happens to your production app" are two different guarantees, and only the first one is actually being offered here. A tenant who set up an App Platform deployment in early 2024 and hasn't touched the plan dropdown since is, by definition, not watching for a deprecation notice — they're going to find out when it happens, not before.
This Isn't a DigitalOcean-Only Problem
It's tempting to read all of this as an argument for owning your own hardware instead of renting a managed platform's pricing tier. That argument only holds up halfway. Hetzner — the bare-metal and cloud provider a lot of self-hosted PaaS setups (including bex's own default fleet target) run on — raised its own cloud prices on June 15, 2026, and it wasn't a rounding error.
CCX dedicated-vCPU instances went up 113-169% in EUR terms, and CPX shared-vCPU instances rose 144% to as much as 204% in USD pricing, driven by industry-wide RAM, NVMe, and GPU cost pressure. Owning the machines does not mean the machines' prices are frozen in amber. Nobody self-hosting on rented Hetzner capacity gets a permanent price floor either.
What owning the infrastructure actually buys is different from price immunity: control over the timeline. Hetzner's price adjustment applied to new orders and rescale operations starting on its effective date — existing servers kept their existing pricing unless and until you chose to resize or add capacity. A self-hosted operator decides when to reprovision onto new pricing, because reprovisioning is something they initiate. A hosted-PaaS tenant on a deprecated tier gets moved on the platform's timeline, inside their production environment, whenever the vendor decides "soon" has arrived. Both worlds see prices change. Only one of them lets the tenant pick the day.
What to Actually Do About It
If you're running production traffic on DigitalOcean App Platform and you're not sure which era you're in, the fix costs nothing and takes five minutes:
- Check your app's creation date. Anything provisioned before May 7, 2024 is very likely still on a legacy plan.
- Match your current tier against the table above to see whether your specific move is a straight upgrade or a small, egress-justified price increase.
- Opt in on your own schedule. DigitalOcean lets you switch to the new plan structure manually today — there's no reason to wait for an unannounced forced migration to find out what your new bill and bandwidth allowance actually look like.
- Re-check your team's pooled egress usage against the new allowance before assuming "substantially higher" means "no longer a constraint."
None of that requires leaving DigitalOcean. But it's worth noticing that the entire exercise — tracking a vendor's undated deprecation notice, reverse-engineering which of ten tier mappings applies to you, hoping the migration lands on a quiet traffic day — is bookkeeping that exists specifically because the pricing tier is something someone else controls.
Bex.co is the open-source, AI-native Render alternative — push a git repo, get a running HTTPS service on machines you own, with no vendor-controlled plan-migration clock to track in the first place. Star the repo on GitHub or deploy your first app today.
