Oracle Cloud 'out of capacity' error: every fix that actually works

What the error actually means

Oracle's own Always Free resources documentation says the "out of host capacity" error "indicates a temporary lack of Always Free shapes in your home region," and the Free Tier FAQ adds that "it might take several days before additional capacity is available." In plain terms: Oracle carves out a small pool of free ARM capacity per availability domain, demand exceeds it, and instances appear only when someone else's instance is terminated.

Two things this is not: it is not a quota problem (that is the separate "LimitExceeded" error), and it is not something you can fix by raising a service limit. Knowing the difference saves you the two most common dead ends, covered in the myths section below.

First, check your numbers: the free ARM limit is now 2 OCPUs / 12 GB

Before touching anything else, make sure you are requesting a shape that can exist. In mid-June 2026 Oracle quietly halved the free ARM allowance from 4 OCPUs / 24 GB to 2 OCPUs / 12 GB (1,500 OCPU hours and 9,000 GB hours a month), with enforcement — including automatic termination of over-limit instances — starting August 18, 2026. The change was covered by WebProNews and InfoQ.

This matters twice over. First, stale tutorials still show 4/24 — if you are following one from before June 2026, its numbers are wrong and the official docs page is the source of truth. Second, provisioning above the limit does not just fail: over-limit instances get disabled and deleted after 30 days. Request exactly 2 OCPUs and 12 GB — or start at 1 OCPU / 6 GB and resize up after the instance exists, since resizing an existing instance does not need new capacity.

Fix 1: cycle the availability domains

Capacity is tracked per availability domain, not per region — Oracle's own first recommendation is to try a different AD. If AD-1 is full, AD-2 or AD-3 may have space. In the create-instance form, change the availability domain under Placement and retry.

One caveat: not every region has three ADs. Frankfurt has three; San Jose has only one, so there is nothing to cycle. If your home region has a single AD and it is dry, skip to the retry automation below.

Fix 2: launch small, then resize

A smaller instance is easier to place. Launch the A1 shape at 1 OCPU and 6 GB of memory first — the minimum — and once it exists, resize it up to the full 2 OCPUs / 12 GB. Resizing an existing instance does not require new capacity the way creating one does, so the resize usually succeeds even when fresh launches keep failing.

This trick is documented in community deployment guides and works because the capacity check happens at creation time. Just never exceed the 2/12 ceiling: Oracle's enforcement now terminates over-limit instances.

Fix 3: automate the retry (the fix that works for most people)

Capacity frees in random bursts when other users terminate instances. The reliable move is a script that polls the launch API every 60 seconds and rotates across ADs until one succeeds — then notifies you. This is the single most-reported working approach in the community, and several public implementations exist:

• oci-arm-catcher — bash and PowerShell, retries on the capacity-error family and rotates ADs, with a writeup of the approach.
• oci-arm-host-capacity-py — Python port of the well-known retry script, already updated for the June 2026 2/12 limits.
• oracle-freetier-instance-creation — Python, attempts every 60 seconds and sends a Gmail notification on success.
• oracle-vm-creator — web wizard with a background launch loop that backs off on rate limits.

Fix 4: retry at off-peak hours

Capacity frees when other people terminate instances, which clusters outside business hours. Community guides consistently report better luck retrying late at night or early morning in the region's local time. There is no official source for this — it is community-reported pattern-matching — but it costs nothing to try alongside the automation above.

Fix 5: upgrade to Pay As You Go (the reliable one, with the honest caveat)

Oracle's own docs state that Pay As You Go accounts are subject to different capacity limits than Always Free accounts, and upgrading "gives you access to more types of Compute resources." Community experience is blunt: upgrading is the fix that ends the capacity lottery, because paid tenancies draw from a different capacity pool.

The honest caveat: PAYG removes the guardrails. Oracle does not charge for Always Free resources after you upgrade and only bills usage above the Always Free limits — but usage above the limits gets billed for real. Set a billing budget alert (even $1) the moment you upgrade, and double-check your instance stays within 2 OCPUs / 12 GB. One more wrinkle: some users report a temporary card-verification hold of around $100 at upgrade time; it is reversed.

Upgrading has a side benefit the docs confirm: idle Always Free instances may be reclaimed by Oracle, and PAYG tenancies are exempt from that reclamation.

Fix 6: fall back to the AMD micro instances

If you just need a tiny server, the two Always Free AMD micro instances (VM.Standard.E2.1.Micro, 1 OCPU and 1 GB of RAM each) are almost always available in every region. They will not run anything hungry, but for a DNS server, a small bot, or a static site they are the path of least resistance — no capacity lottery at all.

Your home region cannot be changed — choose it carefully at signup

Always Free compute must be created in your tenancy's home region, and Oracle's documentation is explicit: it cannot be changed after signup. If you picked a congested region, the only fix is a brand-new tenancy — and creating multiple free accounts is prohibited, so do this once and do it deliberately.

Community consensus (anecdotal, and it shifts) is that US regions like Ashburn and Phoenix are the most frequently congested, while EU and APAC regions such as Frankfurt, Singapore, and Tokyo provision more often. Treat that as a hint to check recent reports, not as fact. If free cloud is the goal and Oracle keeps fighting you, Azure for Students is a calmer alternative with no capacity lottery.

What does NOT work (stop wasting time on these)

Four dead ends, each with its source:

• **Opening a support ticket for free-tier capacity.** Oracle's FAQ is explicit: "Customers using only Always Free resources are not eligible for Oracle Support... you will not be eligible to raise a service request."
• **Requesting a service limit increase.** The LimitExceeded error text suggests it, but Always Free limits are fixed entitlements, not adjustable quotas. One user who tried was simply told the region was under heavy load and to go PAYG (Oracle community thread).
• **Changing your home region after signup.** Cannot be done — see above.
• **Following pre-June-2026 tutorials at face value.** Any guide quoting 4 OCPUs / 24 GB is stale. Our Oracle Cloud free tier overview and deal page track the current 2/12 numbers.

Ready to set it up?

We keep a verified, screenshot by screenshot walkthrough of the signup, including the budget alert and the Always Free shape settings.

Frequently asked questions