Why RAM Should Be Your First Upgrade

Every time someone tells me their laptop feels slow, the conversation goes the same way. They blame the processor, or start eyeing a better graphics card. Almost nobody suspects the actual culprit: not enough RAM. It's the least glamorous spec on the whole sheet, and it's usually the one quietly deciding whether your computer feels instant or feels like it's wading through mud.

What RAM is actually doing

Storage, your SSD, is where data lives when nothing is using it. RAM is where data lives while something is using it: the workspace your CPU reads from and writes to as you work. Every open app, every browser tab, every background service claims a slice of it for as long as it keeps running.

Once that workspace fills up, the operating system doesn't just stop. It starts swapping, quietly shuffling the memory you're not actively using out to disk to free up space, then pulling it back in the moment you switch to it. Even on a fast NVMe SSD, that round trip is orders of magnitude slower than RAM itself. The CPU can be sitting almost idle, the GPU untouched, and the whole machine still feels like it's thinking underwater, because it's spending its time waiting on disk instead of actually computing.

This is why a RAM shortfall feels so different from a CPU or GPU one. A weak CPU slows down CPU-heavy tasks. A weak GPU slows down graphics-heavy tasks. Not enough RAM slows down everything at once, including tasks that have nothing to do with memory in principle, simply because the system is busy fighting for workspace before it can even start.

Why 8GB stopped being enough

8GB was a reasonable baseline a decade ago. It isn't anymore, and not because software suddenly got greedy for no reason: operating systems, browsers, and everyday apps have all quietly grown into whatever memory is typically available, the same way traffic expands to fill a new lane.

Browsers are the biggest offender. Each tab is basically its own small process. A dozen tabs plus a couple of extensions will quietly eat several gigabytes before you've even opened a single "real" application.
The OS itself got heavier. Background indexing, sync services, security scanning, and update agents all sit resident in memory, whether you're using them or not.
Multitasking is the default now, not the exception. A video call, a Slack or Teams client, a code editor with a language server running, a couple of Docker containers, and a browser: that's just an ordinary Tuesday for a lot of people, and every one of those wants its own chunk of RAM at the same time.

16GB is a realistic baseline for a comfortable laptop in 2026. 32GB is the point where you stop thinking about memory at all.

The AI wrinkle: local models are hungry for RAM

The newest reason to care about RAM has nothing to do with browser tabs. Running large language models locally, with tools like Ollama or LM Studio instead of calling a cloud API, has gone from a hobbyist curiosity to something genuinely mainstream people want to do, whether for privacy, cost, or just working offline. And local models are hungry for memory in a way most everyday software simply isn't.

As a rough guide, for quantized models running on CPU or unified memory:

A 7 to 8B parameter model, quantized, wants roughly 8GB just for the model weights.
A 13 to 14B model climbs to around 16GB.
A 30 to 34B model is comfortably into 32GB territory.
70B-class models push toward 64GB and beyond.

And that's on top of whatever your OS and other apps are already using, not instead of it. If you want real headroom to run a useful local model alongside a normal workload, 32GB is a sensible minimum today, and 64GB is where things get genuinely comfortable. This is one of the fastest-growing reasons ordinary buyers are starting to care about a spec they used to ignore completely.

One nuance worth flagging: on a desktop with a discrete GPU, video memory (VRAM) is a separate pool from system RAM, and it matters enormously for GPU-accelerated inference and training. But on most laptops, where the GPU either doesn't exist or shares memory with the rest of the system, system RAM is the number that decides what you can run at all.

Giving the other specs their due

None of this means CPU, GPU, and SSD don't matter. They do, just for narrower reasons.

CPU sets your ceiling for raw compute: compiling code, exporting video, running heavier calculations. A weak CPU paired with plenty of RAM will still feel snappy day to day, just slow at the specific things that actually lean on the processor.
GPU matters enormously for graphics-heavy work (gaming, 3D rendering, video editing) and increasingly for ML training and GPU-accelerated inference. If your workload is genuinely GPU-bound, it earns the right to jump ahead of CPU in priority.
SSD speed affects boot time, app launch time, and how painful swapping feels when it happens. A fast SSD softens the blow of low RAM, but it never removes the problem, because disk is still vastly slower than memory no matter how fast that disk is.

The real difference is that RAM is a systemic bottleneck: it caps how many things can run smoothly at once, no matter the task. CPU, GPU, and SSD are task-specific bottlenecks that only bite when you're doing the one thing they happen to be good or bad at.

My suggested upgrade priority

If you're shopping for a new laptop, or deciding where to put extra budget on a configurable one, here's the order I'd actually spend money in, for most people:

RAM. The single highest-impact upgrade for everyday responsiveness, and the one most likely to be soldered down and impossible to add later. Get enough of it now.
CPU. Sets your ceiling for anything compute-heavy, and matters more the more "real work" (compiling, data processing, video export) you actually do.
GPU. Essential if your workload is genuinely graphics- or ML-bound, otherwise a perfectly reasonable third priority.
SSD (capacity first, then speed). Make sure you won't run out of space, then worry about which NVMe generation it is. It's the upgrade with the smallest impact on how the machine feels to use.
Everything else: display quality, battery life, build, ports. All real considerations, none of them change how the machine performs.

The one exception: if you already know your workload is dominated by one thing, competitive gaming, 3D rendering, model training, let that specific bottleneck jump the queue. But for the median buyer just trying to make a laptop feel fast for years, RAM first is the right default, and that's only gotten truer as local AI tools have entered the picture.

Why RAM got expensive

If you've priced out RAM recently and it seemed pricier than you remembered, that's not your imagination. Industry analysts have tracked a sharp run-up in DRAM pricing through 2025 and into 2026, driven mostly by one cause: AI data centers buying up an increasingly dominant share of global memory production. Fab capacity at the major manufacturers, Samsung, SK Hynix, and Micron, has been shifting toward the higher-margin memory that AI infrastructure needs, which leaves a shrinking slice of supply for laptops, phones, and everyday consumer devices, and pushes consumer DRAM prices up in the process. Most analyst outlooks don't expect meaningful relief until new fabrication capacity comes online, which looks like a 2027 or 2028 story rather than a "wait a few months" one.

The practical takeaway is almost the opposite of what you'd hope for: RAM is more expensive right now, at the exact moment it matters more than ever thanks to local AI. If you're buying a laptop and can afford to size up the memory, doing it now, rather than hoping to add it later or hoping prices drop, is the safer bet. Especially since a growing share of thin-and-light laptops solder RAM to the board and don't allow any upgrades after purchase at all.