Nvidia’s Quiet Crisis: It’s Not Who You Think

by

Right, so, three decades ago the internet happened. Honestly, it was a bit of a mess, wasn’t it? But a profitable mess. It changed everything for corporate America, opened up all these shiny new channels for selling things, and, crucially, let regular people start poking around in the stock market. It was… disruptive. And now everyone’s waiting for the next big thing. And, wouldn’t you know it, it looks like AI might be it. Nvidia, naturally, is leading the charge. Because, well, someone has to.

They’re basically giving computers the power to make decisions at the speed of thought. Which sounds amazing, until you consider what terrible decisions I make at that speed. Still, analysts at PwC reckon it’ll add over $15 trillion to global GDP by 2030. Fifteen trillion. I could really use a slice of that. Nvidia’s GPUs are the brains behind it all, powering everything from data centers to, let’s be real, probably some very sophisticated cat videos.

Since the beginning of 2023, Nvidia’s market cap has gained almost $4.3 trillion. That’s… a lot of zeros. It’s clear how essential their hardware has become. But here’s the thing about sitting on top of a pile of money: everyone starts looking at you funny. And they start building things to knock you off.

Broadcom and AMD are definitely trying. They’re formidable, sure. Like that overly enthusiastic colleague who always has a “solution” for everything. But they’re not the real threat. No, the biggest risk to Nvidia isn’t another company making faster chips. It’s something far more…internal. And it’s making me deeply uncomfortable.

Broadcom and AMD: The Usual Suspects

Okay, let’s be fair. Demand for GPUs is insane. Nvidia, Broadcom, and AMD are all having a field day in the AI data center space. Nvidia currently owns the lion’s share, thanks to their compute advantages. Their GPUs – Hopper, Blackwell, Blackwell Ultra – are consistently ahead of the curve. Jensen Huang, the CEO, is basically launching a new chip every year. It’s relentless. And frankly, a little intimidating. He’s not messing around.

Broadcom and AMD are trying to chip away at Nvidia’s dominance (sorry, I had to). AMD’s angle is value and availability. They’re the reliable friend you can always call, even if their solution isn’t the flashiest. Nvidia’s orders are backlogged, so AMD offers a quicker alternative. Broadcom, meanwhile, is specializing in ASICs, custom-built chips for hyperscalers. They’re projecting $60 to $90 billion in sales over the next few years. Not bad, not bad at all.

They’re both competent. They’re both trying. But they’re not the ones keeping me up at night.

The Real Enemy: It’s Complicated

The biggest threat to Nvidia is…its customers. Specifically, the “Magnificent Seven.” Meta, Microsoft, Amazon… these companies have poured billions into Nvidia’s GPUs. They’re fueling their AI ambitions, and Nvidia is happily taking their money. But they’re also all developing their own AI chips. Isn’t that just…rude?

Meta has its Meta Training and Inference Accelerator. Microsoft has Azure Maia 200. Amazon has Inferentia2 and Trainium. Even Alphabet’s Google Cloud has Tensor Processing Units. They’re all building in-house solutions. And while they might not be faster or more efficient than Nvidia’s GPUs right now, they’re significantly cheaper and, crucially, available.

Let’s break this down. Nvidia’s pricing power – and its mid-70% gross margin – is based on scarcity. If the biggest buyers start making their own chips, that scarcity disappears. And that margin? Gone. Plus, if they’re developing their own hardware, they’re less likely to upgrade Nvidia’s GPUs as frequently. They’ll stretch out the lifespan of existing hardware, and that impacts future revenue. It’s like dating someone who suddenly decides they’re perfectly happy with their houseplants. It’s…disconcerting.

And there’s one more thing. Nvidia’s aggressive innovation cycle – launching a new chip every year – rapidly depreciates older GPUs. If customers can complement those older GPUs with their own in-house chips, they have less incentive to buy the latest and greatest from Nvidia. It’s a slow burn, but it’s happening. And it’s terrifying.

Honestly, the biggest competitive risk for Nvidia isn’t a faster chip. It’s the quiet realization that its most valuable customers might not need it as much as they used to. And that, my friends, is a crisis of confidence. And I, for one, am deeply invested in seeing how this plays out. (Mostly because it’s entertaining.)

Loading widget...

Read More

2026-02-03 12:12