Table of Contents
Two developers I know bought laptops in the same week last year. One spent $2,400 on a machine with a 4K OLED panel and a discrete GPU. The other spent $850 on a refurbished business laptop with a boring 1080p screen. Six months on, the cheap one is the machine happily running three Docker containers, an Android emulator and twenty browser tabs — because it had 32GB of RAM and the expensive one had 16.
That gap is the whole game. Choosing a computer for programming is mostly about matching the machine to a specific workload, then pouring your budget into the two or three components that actually slow you down. Everything else on the spec sheet is marketing aimed at someone else.
Start With the Workload, Not the Spec Sheet
A script that reads a CSV and a monorepo with 400,000 lines of TypeScript are not the same problem. Neither is a Flutter app with an iOS simulator running next to it. Before you compare processors, write down what a normal Tuesday looks like: which editor, how many containers, whether you compile native code, whether you train models.
The language you work in shapes that answer more than you’d expect. Rust and C++ punish slow CPUs during full rebuilds. Python and JavaScript punish low RAM once your tooling layers stack up. If you’re still deciding where to start, our breakdown of the top programming languages in 2025 is a reasonable map of what each ecosystem demands from your hardware.
The CPU: Cores Matter Less Than the Reviews Suggest
Most of a programmer’s day is single-threaded. Typing, language-server autocomplete, incremental compilation, running a test file — all of it leans on one fast core. Then, twice a day, you do something embarrassingly parallel: a clean build, a full test suite, a Docker image rebuild. That’s when cores pay off.
For a solo developer, a recent 8-core chip is the sweet spot. An AMD Ryzen 7, an Intel Core Ultra 7, or any Apple M-series chip from the M2 onward will chew through nearly anything you throw at it. A 24-core desktop monster for a Node project is money burned on heat and electricity.
One thing reviews rarely capture: sustained performance. A slim 13-inch laptop with a 28W chip may post great benchmark numbers for 30 seconds, then throttle hard for the next ten minutes of compiling. If you build big projects often, a slightly thicker chassis with real cooling is worth more than a faster chip in a thin shell.
RAM Is the Spec You Will Regret Skimping On
Here’s roughly where 16GB disappears on a working machine:
- VS Code with a handful of extensions: 1.5–2GB
- Chrome or Firefox with 25 tabs: 3–4GB
- Docker Desktop and two or three containers: 4–6GB
- An Android emulator or a local database: 2–4GB
- The OS itself, idling politely: 3GB
That’s already past 16GB before you open a second project. Treat 16GB as the absolute floor in 2025 and 32GB as the comfortable default if you touch containers, virtual machines or mobile emulators. On almost every modern laptop the RAM is soldered to the board, so this is the one decision you cannot fix later with a $70 stick.
Storage: Cheap Insurance Against a Bad Afternoon
512GB is the minimum I’d accept; 1TB is the number that stops you thinking about it. The culprits aren’t your source files — they’re everything around them. Node modules across a few projects can quietly hit 15GB. Docker images and volumes eat 30GB without trying. A Windows VM is 40GB before you install anything inside it. Datasets for a machine learning course can run to hundreds of gigabytes.
Check the drive type, not just the capacity. Some budget models still ship eMMC storage that makes an SSD feel like a sports car by comparison. You want NVMe, and ideally a machine where the drive can be swapped in twenty minutes with a screwdriver.
The Screen, Keyboard and Ports You’ll Live With
You will stare at these for eight hours a day, and they’re the parts reviewers gloss over.
Display
1080p at 14 inches is genuinely fine. What matters more is the aspect ratio — a 16:10 panel shows noticeably more code than 16:9 — and the finish. Matte screens beat glossy ones in every room with a window. High-DPI panels look lovely, but if your Linux desktop or older Windows apps don’t scale cleanly you’ll spend evenings squinting at blurry fonts. A 27-inch 1440p external monitor costs around $200 and will improve your working life more than any internal upgrade.
Keyboard
Key travel around 1.5mm, a proper function row, and a layout that puts Home/End/PgUp somewhere sane. If you can, type on the machine in a store before buying — ten minutes tells you more than any review. Otherwise buy from a retailer with a real return window.
Ports
USB-C charging is non-negotiable now; it means one cable for everything and a $25 spare charger instead of a $90 proprietary brick. Beyond that: at least one USB-A for your mouse and hardware keys, and HDMI for the projector in the meeting room you’ll inevitably be late to.
Mac, Windows or Linux?
All three work. The differences that actually bite:
Mac. A real Unix shell out of the box, excellent battery life, and Apple silicon that compiles fast while staying cool. The trade-offs are soldered everything and a price premium on RAM. You also can’t easily virtualise x86 images, which occasionally matters for legacy work.
Windows. WSL2 changed the calculus. You get a genuine Linux kernel alongside your Windows apps, so Rails, Docker and Python tooling behave the way the tutorials say they will. It’s the easiest path if you also build .NET or Windows desktop software.
Linux. The most direct route to a server-like environment, and the best experience on cheaper hardware. Expect the occasional fight with sleep states, fingerprint readers and Wi-Fi drivers on very new laptops.
When you’re genuinely torn, copy your team. Watching how seniors and juniors split work on a codebase tells you a lot about which environment removes friction — and the boring parts of a senior developer’s setup usually exist for a reason that will become obvious to you in about three weeks.
Laptop, Desktop, or an Awkward Hybrid
Dollars per unit of performance strongly favour desktops: upgradeable RAM, a better cooler, a GPU slot you can fill in two years. If you never leave the house to work, build or buy a desktop and spend the savings on a good chair.
If you do move around, don’t overlook refurbished business laptops. Three-year-old ThinkPad T-series and Dell Latitude machines sell for 40–50% of their original price, are built to be serviced, and have parts available for years. Corporate IT departments replace hardware on a fixed schedule regardless of condition, which is exactly the product lifecycle quirk that makes these bargains possible. Buy from a seller offering a warranty and check the battery health report before you commit.
What You Don’t Need to Pay For
A discrete GPU tops the list. Unless you’re training models, doing game development or rendering video, integrated graphics handle everything — your code editor does not care. Local AI assistants are getting more useful, but a 7B model at 4-bit still wants roughly 8GB of VRAM, so a laptop GPU doesn’t buy you much there either; the trends worth tracking in O’Reilly’s September 2025 radar point toward hosted models for most day-to-day work.
Skip 4K on a 14-inch screen, skip RGB lighting, skip the newest generation if last year’s is $300 cheaper. Enthusiast hardware is priced for gaming and prestige, and programming sits in a much narrower band of demand.
A Short Buying Checklist
- 8 cores from the last two generations, or Apple M2 and newer
- 16GB RAM minimum, 32GB if you use Docker, VMs or emulators
- 512GB NVMe SSD, 1TB preferred, user-replaceable if possible
- 14–16 inch matte display, 16:10, 1080p or 1440p
- Keyboard you’ve actually typed on, or a generous return policy
- USB-C charging plus one USB-A and HDMI
- Serviceable battery and storage if you plan to keep it five years
Where the Extra $300 Should Go
If you have a fixed budget and a little headroom, spend it in this order: RAM first, storage second, CPU third, screen fourth. Everything else is noise. Doubling from 16GB to 32GB costs less than a GPU upgrade and removes the single most common reason a developer’s machine feels slow. Going from 512GB to 1TB costs about $100 and buys you years of not deleting Docker images at 11pm.
Only after those two are settled does a faster chip make sense — and by then you’ll probably find you didn’t need it. The developers I know who are happiest with their machines didn’t buy the fastest thing available. They bought enough of the right things, and then stopped thinking about hardware entirely.
