A Mac’s storage is soldered to the board. You choose the size on the day you buy the machine, and after that the only way to get more is a new Mac. On Apple’s configurator each step up costs $200 to $400, checked on 15 September 2026: 256 to 512 GB is $200, 512 GB to 1 TB is $200, 1 TB to 2 TB is $400. That has been true for years, and this year the flash inside every one of those steps got two to three times more expensive to make, so the pressure is upward, on new Macs and on the used ones people buy instead.
An external drive over USB is not the same thing. It is slower, it is a cable, and the folders that fill a developer’s disk are caches and build outputs that need to be fast and local. So the real question for a full Mac is a bigger machine or making room on the disk you have. This article is about the second one, and about how much room there usually is.
Why storage got expensive
The chips that make up an SSD are NAND flash. In 2025 cloud providers began buying enterprise SSDs for AI servers at a scale the industry had not planned for, and the makers of NAND sold out their 2026 capacity early. Contract prices for NAND rose by more than 100% in the first half of 2026 and are expected to climb another 10 to 15% in the third quarter, from a level that was already a record. Consumer drives follow contract prices with a lag of a few months, which is why the drive you looked at in November costs so much more now. New factories are years away, so most forecasts put relief in 2027 or later.
Sources, checked on 15 September 2026:
- Tom’s Hardware, State of play: SSD pricing one year into the AI component crisis
- TrendForce, AI server demand continues to support memory prices in 3Q26
- TrendForce, Contract prices surged more than 100% in 1H26
None of this reaches a Mac as a cheaper drive you could fit. It reaches it as the price of the next machine, and as the reason the one you have is worth keeping full of your own work rather than caches.
What developer tools leave behind
Every toolchain keeps a cache somewhere in your home folder, and almost none of them clean up after themselves. The caches are useful while you work on a project. After the project ends, they sit there. Over a year or two of daily work they add up to tens of gigabytes, and on a 512 GB MacBook that is a meaningful share of the drive.
These are the figures Cache Goblin finds on typical setups after a year or two of use. Each is a real scan of a working Mac, rounded.
| Setup | Safe to delete | Back with a download or build | Total |
|---|---|---|---|
| iOS developer | 27.7 GB | 45.9 GB | 86.7 GB |
| Android developer | 23.3 GB | 33.7 GB | 75.4 GB |
| Web developer | 14.7 GB | 52.3 GB | 71.6 GB |
| Python and ML engineer | 12.8 GB | 65.3 GB | 82.7 GB |
| LLM developer | 9.6 GB | 237.9 GB | 249.6 GB |
| Rust developer | 8.2 GB | 66.6 GB | 80.3 GB |
| Backend and DevOps | 16.2 GB | 60.2 GB | 84.7 GB |
| Flutter developer | 25.5 GB | 42.0 GB | 85.6 GB |
| Game developer | 28.4 GB | 57.0 GB | 89.7 GB |
The largest single items, by toolchain:
| Toolchain | What takes the space | Typical size |
|---|---|---|
| Xcode | DerivedData, module caches, unused simulator runtimes at 7 to 8 GB each | 15 to 60 GB |
| Docker, OrbStack | The virtual disk image behind your containers | 20 to 60 GB |
| AI models | Ollama, LM Studio, Hugging Face and MLX weights, 4 to 9 GB per model, and every model you tried once and never removed | 30 to 80 GB |
| Node | node_modules in projects you stopped touching, plus the npm, yarn and pnpm stores | 30 to 50 GB |
| Android | Gradle caches, system images, emulator disks | 20 to 50 GB |
| Rust | target directories, one per project, 2 to 5 GB each | 10 to 50 GB |
| Unity | Library folders, one per project, and old editor versions | 20 to 40 GB |
| Everything else | pip and conda, Gradle and Maven, Homebrew, browser dev profiles, editor caches | 5 to 15 GB |
Two kinds of space
The distinction that matters is how the space comes back if you need it.
Some caches are rebuilt by the tool on its own the next time you run it. Xcode’s DerivedData, the npm cache, the Go build cache, Homebrew’s downloads. Deleting them costs you one slower build. Cache Goblin calls these Safe and removes them with one click.
Other things come back with a download or a full build. Simulator runtimes, Docker images, model weights, node_modules in a project you have not opened in months. These are safe to remove too, and you should think for a second before you do, because getting them back takes time or bandwidth. Cache Goblin calls these Restorable, lists them, and asks before it moves them to the Trash.
A third group holds work you might still want. Xcode archives, virtual machine disks, database data, old toolchain versions you may need for an old project. Cache Goblin shows these and never touches them on its own.
What to do before you buy a drive
- Run a scan. The free version of Cache Goblin shows every cache it finds, graded Safe, Restorable or Risky, with the size of each. It does not delete anything.
- Take the Safe total. On the setups above that is 8 to 28 GB, gone in one click, back on its own the next time you build.
- Read the Restorable list. Simulator runtimes for iOS versions you no longer test, Docker images from a project that shipped, a model you tried once. This is usually the larger number.
- Decide about the drive with the real figure in front of you. A 512 GB Mac with 80 GB of caches on it is a different machine from a 512 GB Mac that is genuinely full of your own files.
If the scan comes back small, you do need more storage, and you can buy the next Mac knowing the one you have is full of your own work. If it comes back with 60 or 80 GB, you have the better part of a $200 configurator step sitting in your home folder, and a $29 license clears it in one click.
Download Cache Goblin free for Mac and see your own numbers. Scanning is free. Cleaning is a one-time license.