Why I Don't Use Apple

My last Apple device was an iPhone 3GS.

It was great, actually. The first app I installed was Angry Birds. I played it constantly. The phone felt premium — expensive, sure, but premium. I was sold.

Then, not even a year in, I plugged in the charger — the same factory charger I'd been using since day one — and the screen said:

"This accessory is not supported by this device."

Same cable. Same wall adapter. Same everything. My phone just decided it didn't want to charge anymore.

The Genius Bar Moment

I went to Apple. Showed them the phone. Showed them the charger. Their charger.

They wouldn't fix it.

Instead, they asked me to book a Genius Bar appointment. Another $100-ish, on top of the overpriced phone that broke with its own accessories.

That was the moment.

Not a dramatic boycott. Not a philosophical stance. Just — why am I paying premium prices for a device that rejects its own charger, from a company that won't fix it without another fee?

I switched to Android. Before the iPhone I'd been on BlackBerry. After it, I never looked back.

The Sales Pitch Is Incredible Though

I'll give Apple this: nobody sells technology better.

The keynotes are cinematic. The product shots are flawless. The messaging is tight. If you don't understand the specs — and most people don't — you'd walk out of an Apple Store believing you just bought the most advanced machine on the planet.

And that's the trick. The marketing is the product. The hardware is just... there.

Overpriced. Over-advertised. Overvalued. But beautifully presented.

The "Pro" That Isn't

Apple sells laptops starting at $1,699 and scaling past $7,000 — all with soldered RAM, soldered storage, and zero upgrade paths. They call the high-end ones "Pro" and convince people it's revolutionary.

Their upgrade path is simple: buy next year's model. That's it. That's the path.

The hardware improves each generation — I'll give them that. The M-series chips are genuinely faster year over year. But you can't swap the GPU. You can't add RAM. You can't replace storage without replacing the entire machine. And that forced upgrade cycle means more devices in landfills. A swappable GPU isn't just better for my wallet — it's better for hardware lifespan.

That's not "Pro." That's a subscription with a unibody shell.

Then I Started Building AI Systems

This is where my opinion stopped being just sarcasm and started being informed.

When I got into local LLM inference — running models on my own hardware — I had to actually understand what these chips do. Not the keynote version. The real version.

Apple's unified memory architecture? It's not a gimmick. An M4 Max with 128 GB can hold a 70B parameter model at Q4 quantisation entirely in memory — and actually run it at 10-18 tokens per second. An RTX 4090 with 24 GB VRAM can't fit it without offloading layers to system RAM, which significantly degrades throughput. For running large models locally, Apple Silicon has a genuine structural advantage.

I didn't expect to say that. But it's true.

The problem is everything else. Training, fine-tuning, anything that needs the CUDA ecosystem — PyTorch, DeepSpeed, Flash Attention, bitsandbytes — all of it is built and optimised for NVIDIA first. Metal and MPS backends exist but they're behind. The software gap is the real issue, not just the hardware.

So Apple's not useless for AI. I just wouldn't build a workstation around it.

The Build I'd Actually Want

If someone handed me a top-spec MacBook Pro, I'd thank them. Then I'd sell it.

The build would cost more than a base MacBook Pro — let's not pretend otherwise. But it would cost less than a maxed-out one, and the difference in what I can do with it is enormous:

Component	Spec
CPU	AMD Ryzen 9 9950X
GPU	RTX 4090 (24 GB VRAM)
RAM	64 GB DDR5
Storage	4 TB NVMe Gen4
Cooling	360mm AIO liquid cooler
Upgrade path	Everything

For AI workloads, the metric that matters is tensor core throughput — the RTX 4090 delivers 165 TFLOPS at FP16 dense, 330 with structured sparsity. Those are the numbers that actually move tokens. Headline figures — NVIDIA's or Apple's — are all marketing-friendly until you benchmark a real workload.

What I can do with this: fine-tune models up to ~13B with QLoRA, train smaller models from scratch, run inference on anything that fits in 24 GB VRAM. When it doesn't fit, I upgrade the GPU. Not the entire machine.

No unified memory pool. No macOS. No ecosystem magic. For what I do — that's fine.

But People Love It

They do. And I get it. My family still uses Apple devices. I'm the one who bought them.

The iPhone camera is genuinely top-tier — Samsung, Pixel, and Chinese brands like Huawei and Xiaomi have caught up or surpassed it on stills, but for video, iPhone still wins. Retina displays are gorgeous. The ecosystem is seamless if you're all-in.

If someone asks me what phone to buy, I'll push them away from Apple. But I'll be honest about where it excels. Probably less enthusiastically than where it doesn't. I'm biased. I know.

There's a practical reason I keep them around. I needed it for WebKit testing, Safari QA, and Xcode deployments. Some markets still skew heavily Apple — you can't just ignore that as a developer.

So I keep them. I just don't use them as my daily driver.

The "Totally Fair" Comparison

Here's my completely unbiased breakdown. I promise.

	MacBook Pro / iMac	Custom PC / Laptops (Lenovo, ASUS, etc.)
Build quality	Beautiful. Genuinely premium.	Varies wildly. Mine has RGB though. That counts, right?
Display	Retina is gorgeous. Designers love it.	You pick your panel. 4K OLED? Ultrawide? Mini-LED? Your call.
Upgrade path	Buy a new one next year.	Swap a part. Keep the rest. Desktop and some laptops.
RAM	Soldered. Choose wisely at checkout.	Slotted on desktops, often on laptops too. Change your mind whenever.
GPU compute	Fine for Figma.	Desktop: fine for fine-tuning up to ~13B. Laptop: depends.
Repairability	"Visit the Genius Bar" ($$$)	Desktops: fully repairable. Laptops: ThinkPads are great, others vary.
Ecosystem	AirDrop, Handoff, iMessage. Seamless if you're all-in.	USB-C, Bluetooth, open standards. Less magic, more freedom.
Privacy	"What happens on your iPhone stays on your iPhone" (terms apply)	You control the telemetry. Or don't. Your call.
Pricing	Starts at $1,699. Maxes at $7,349. For a laptop.	Same budget, multiple times the specs. Or half the budget, same specs.
AI workloads	Unified memory is genuinely good for large model inference.	CUDA ecosystem is genuinely good for everything else.
Software support	macOS natively. You can virtualise Windows, but you need macOS for Xcode.	Windows, Linux, or both. Run whatever you want.
Marketing	World-class. Oscar-worthy keynotes.	None. The specs speak for themselves.
Target audience	People who want things to just work.	People who want to know why things work.

Look, I said I was biased.

The Real Innovation

Apple's real innovation isn't silicon. It's not the M-series chip. It's not the Neural Engine.

It's convincing millions of people that a sealed, non-upgradeable computer with premium pricing is somehow the pinnacle of professional hardware.

That takes talent.

Still not paying premium prices for sealed hardware. And now that I actually understand what these chips do — I'm even more sure about that.