Nvidia just started a new chip war | The Vergecast

ELI5 / TLDR

Nvidia, the company that makes the graphics chips everyone wants for AI and gaming, just announced it’s going to make the main brain of regular laptops too — not just the graphics part. The new chip is called the RTX Spark, and basically every major laptop maker (Dell, HP, Lenovo, Asus, Microsoft) has lined up to build machines around it. Nobody knows the price, nobody has tested one, and Nvidia is making big claims with zero charts to back them up. But the fact that the whole industry showed up is a sign this could genuinely shake up what’s inside your next computer.

The Full Story

What “making a chip” actually means here

A laptop has two brains. One is the CPU — the central processor, the part that runs your operating system, opens your apps, does the everyday thinking. The other is the GPU — the graphics processor, originally for drawing pixels in games, now also the workhorse for AI.

Nvidia has owned the GPU side for years. Its graphics cards are so in demand right now that, as one host puts it, “they cost a billion dollars and no one can get any.” But Nvidia has almost never made the CPU — the heart of the machine. That seat has belonged to Intel and AMD for decades, with Apple and Qualcomm joining more recently.

The RTX Spark changes that. Nvidia is now making the whole central chip, the silicon at the core of the computer.

“Now, they’re finally doing that thing where they are the silicon.”

x86 versus ARM, in plain terms

There are two competing “languages” a chip can be built on. Think of them as two different blueprints for how a processor is wired.

The old blueprint is called x86 — that’s Intel and AMD, the chips inside almost every Windows PC of the last 30 years. The newer blueprint is called ARM. It’s simpler and sips less power, which is why it runs your phone, and why Apple’s laptops got dramatically better battery life when they switched to it in 2020.

Nvidia’s new chip is ARM. So is Apple’s. So is Qualcomm’s. With Nvidia joining, the senior editor notes there are now more companies making ARM laptop chips than x86 ones. The decades-long reign of x86 is being challenged on multiple fronts at once.

Why Nvidia wants in — and why it’s scared

The hosts land on two motives. The first is obvious: AI. Nvidia wants to be “the AI computer,” the thing at the center of the whole AI shift.

The second is fear, dressed up as ambition. Nvidia remembers what happened to Intel. Intel ruled PCs, then completely missed the smartphone — there was no “Intel inside” any iPhone — and never recovered that ground. The lesson: if you are not physically inside the device people actually use, you can miss the next era entirely.

“Intel wasn’t in the phone. And that was a big problem for Intel when the tide turned.”

So this is a hedge. Right now AI mostly happens in the cloud — in giant data centers Nvidia already supplies. But if the future tilts toward “local AI” (your laptop doing the AI work itself, no internet round-trip, more private), Nvidia wants to own that too. By putting its chip at the heart of the laptop, it covers both bets.

The chip already existed — this is a repackaging

Here’s the funny part. The RTX Spark isn’t really new. Nvidia had already shipped almost the exact same thing as the DGX Spark — a small gold box, an “AI supercomputer” you put on your desk for four or five thousand dollars to run AI models locally.

What seems to have happened: rival AMD took a similar high-memory chip and repurposed it as a normal computer processor, dropping it into mini desktops and even handheld game consoles. Nvidia apparently looked at that and said, we can do that too — and turned its desk-box chip into a laptop chip.

The top-end spec is hefty: over 6,000 graphics cores, 20 CPU cores, and 128 gigabytes of “unified memory.” Unified memory means one shared pool that any part of the chip can grab instantly, rather than separate memory stashes that have to copy data back and forth. It’s built by TSMC (the Taiwanese foundry that physically manufactures most advanced chips) using a partner design from MediaTek with Nvidia’s technology layered on. So it’s not purely an Nvidia chip — it’s a collaboration.

The strange omission: no graphics card

The puzzling bit: this chip can’t be paired with one of Nvidia’s famous discrete graphics cards. For a company whose entire reputation is graphics cards, that’s odd.

The explanation is that mainstream laptops are moving toward one chip that does everything — CPU, graphics, AI all fused together. It’s more efficient, runs cooler, lasts longer on battery. The trade-off is a ceiling on raw power. The Spark’s graphics land at a respectable mid-tier (roughly an “RTX 5070 mobile” level), not the absolute top a bulky gaming laptop can reach. If you’re a hardcore gamer or video editor, you’ll still buy a thick machine with a separate graphics card. Everyone else gets the all-in-one.

Why the industry pile-on matters

The strongest signal isn’t the chip — it’s the guest list. When Qualcomm does these launches, it shows a slide of vague “partners” and shares no real products. Nvidia showed up with actual shipping devices from Dell, HP, Lenovo, Asus, and a new Microsoft Surface Laptop Ultra. Thirty laptops and ten mini-desktops are reportedly in development.

“This is we are creating new kinds of devices for this chip and they seem to have the whole PC industry on board. That feels huge to me.”

The hosts’ read: Nvidia, like Apple, is unusually good at making the whole industry build to its specs. That alone could finally push “Windows on ARM” — long a stalled, buggy experiment — into something that actually works, because game makers like Riot (Valorant, League of Legends) are already committing to support it.

The honest caveats

Nobody has tested one. Nobody knows the price — when pressed in front of executives, the editor floated “$1,500 to $8,000” and got no answer, then guessed it’ll land “a lot closer to 8,000.” The first wave is for early adopters and “AI bros,” not normal buyers. And the deeper bet — that we’ll soon use computers so differently that the whole PC needs reinventing — is, as the host says, “not settled fact yet.”

Key Takeaways

The RTX Spark is Nvidia’s move to make the CPU (the central brain) of laptops, not just the GPU (graphics) it already dominates.
It’s an ARM chip; ARM is the power-efficient blueprint behind phones and Apple’s laptops, now overtaking the older x86 standard used by Intel and AMD.
Two motives: ride the AI wave, and hedge against missing “local AI” the way Intel missed the smartphone.
The chip is essentially a repackaged DGX Spark desktop box, prompted partly by AMD doing the same trick first.
Top spec: 6,000+ GPU cores, 20 CPU cores, 128GB unified memory (one shared instant-access memory pool), built by TSMC on a MediaTek design.
It can’t pair with a separate Nvidia graphics card — the trend is one fused chip doing everything, trading peak power for efficiency and battery life.
Nearly the entire PC industry (Dell, HP, Lenovo, Asus, Microsoft) launched real products around it — the clearest sign it’s serious.
Could finally make “Windows on ARM” reliable by forcing developers to support it.
Price unknown and untested; first generation is for early adopters, likely several thousand dollars; ships around fall 2026.

Claude’s Take

This is a daily news podcast doing its job: two smart people talking through an announcement on the day it dropped, with no benchmarks and no prices because there aren’t any yet. That’s the ceiling on how useful it can be. Half the value is “here’s what Nvidia said,” and Nvidia said remarkably little of substance — the “most efficient PC chip ever built” claim with “no graphs, no charts, nothing to back it up” is a marketing line, and the hosts are appropriately skeptical about it.

What they do well is the strategic read. The Intel-missed-the-phone framing is the right lens, and the observation that the real story is the partner list rather than the silicon is genuinely sharp. The Apple-2020 comparison is the obvious one but it’s apt. Where it stays honest: they repeatedly flag that nothing is tested, the price is a mystery, and the grander “we’ll reinvent the PC” thesis is unproven.

A 6 because it’s a clear, well-framed primer on a real shift — ARM eating x86, Nvidia going vertical — but it’s commentary on a press release, not analysis of a product anyone has touched. Worth listening if you want the lay of the land; come back in six months for whether any of it was true.