NVIDIA’s ‘MacBook Pro’ Revealed: Huang Built Its Own CPU!

< img id="wx_img" src="https://www.qbitai.com/wp-content/uploads/imgs/qbitai-logo-1.png" width="400" height="400">

2026-05-31 22:17:43 Source: QbitAI

We Need a “Printer”

Written by Jay from Ao Fei Temple

QbitAI | Official Account QbitAI

The NVIDIA版 MacBook Pro is really coming.

After waking up one day, I noticed that NVIDIA’s official account posted this message yesterday:

A New Era of PC.

25.0528, 121.5990

I know most people’s first reaction to this post would be:

Huh? That’s it? What’s there to write about?

But actually, this post hides two very critical pieces of information.

1. 25.0528, 121.5990.

If you paste these coordinates into a map app, you’ll see this location—

Taipei, Taiwan — Taipei Music Center.

Next week’s Computex GTC main venue in Taipei. Huang will deliver his keynote there.

2. A new era of PC.

This is most likely the long-rumored NVIDIA laptop powered by N1X, running Windows on Arm.

I’m not making this up — Microsoft and ARM released nearly identical posts on the same day, hinting at a joint effort to “dominate” Apple (jokingly).

After making a fortune selling shovels in the AI boom, Huang is now bringing his self-developed CPU to enter the PC market.

N1X

This name has been circulating in tech circles for at least half a year.

People have been talking about it endlessly, but no concrete proof existed—until today, when all three parties simultaneously revealed their cards.

So what exactly is N1X?

Simply put, it’s NVIDIA’s version of Apple’s M-series chips.

To help you understand, let me show you something similar.

NVIDIA previously released a mini PC called DGX Spark — the small cube in the bottom-right corner — designed as a local sandbox for AI developers. Its core computing power comes from GB10, essentially a super chip combining an NVIDIA GPU, a 20-core ARM CPU, and a large amount of memory.

And N1X is likely the laptop version of GB10.

Rumored specs are quite impressive:

• Co-developed with MediaTek, built on TSMC’s N3B process, featuring a 20-core CPU
• Blackwell GPU with 6,144 CUDA cores
• 128GB LPDDR5X unified memory, shared between CPU and GPU

6,144 CUDA cores — you might not grasp the number intuitively.

Let me put it this way: The desktop RTX 5070 also has 6,144 CUDA cores.

Of course, don’t get too excited yet.

Since no such product has appeared before, there must be issues.

First, forget gaming.

Bandwidth is insufficient. Since CPU and GPU share the same pool of LPDDR5X memory, the actual GPU bandwidth is likely around 273 GB/s.

For gaming… you can play, but compared to traditional dedicated GPU laptops with GDDR memory, performance will definitely fall short.

Beyond that, there’s another practical issue.

Like Qualcomm’s Snapdragon X, N1X is based on ARM architecture. To run traditional x86 applications, ARM needs to go through emulation translation.

And for decades, almost all PC games were built for x86.

So if your dream is a 6,144 CUDA-core ultra-thin gaming laptop… you should scratch that off your wishlist.

But honestly, I don’t care much — if I’m buying it, I’d prioritize AI experience.

So if it turns out to be genuinely good, we’ll need to seriously consider the price.

After all, it’s NVIDIA. You know how it goes.

Please, stop talking about the “new PC era.” We haven’t even solved the problem of rising RAM prices yet!

AI-Native PC = Printer

Lately, more and more news about consumer-grade compute devices has emerged.

NVIDIA DGX Spark, ASUS GX10, Lenovo just launched AI desktops.

From manufacturers’ perspective, they’re racing to capture a new entry point. But for consumers, this is also quite exciting.

All consumer electronics ultimately rely on compute power — but only after being turned into products can compute potential truly be unleashed.

PCs fueled the internet explosion; smartphones enabled the百花齐放 (flourishing) of mobile internet.

In the AI era, we’ve always been missing an AI-native compute device.

I’ve been pondering a historical analogy recently — I mentioned it in a previous article.

In 1450, Gutenberg scaled up metal movable type printing. Before that, the cost of a single book was roughly equivalent to a small estate. Knowledge was locked behind barriers of resources and power — only churches and nobles could own books and spread ideas.

But what happened after printing became scalable?

Infrastructure upgrades allowed commoners to become writers. No longer did only nobility publish books, or only the church disseminate knowledge…

As long as you had an idea, you could present it to the world.

Honestly, I think using AI today feels remarkably similar to writing books 500 years ago.

All cutting-edge models are outrageously expensive. Everyone talks about embracing AI, but to actually use AI for anything meaningful, you first need to afford the subscription fees…

The emergence of AI PCs may be changing that.

With sufficient compute power, many tasks no longer need to rely on cloud services — you can deploy models locally and pair them with intelligent routing agents.

In this case, you only pay once upfront for hardware, then generate infinite tokens at near-zero marginal cost to recoup your investment.

Simple automation tasks like organizing files or scheduling periodic data pulls? No more hesitation — since you’re not paying per unit, you can use freely.

It won’t be just big companies who can use AI, nor will it require burning money to enjoy its benefits. Every creator, every coder with ideas should strive to turn their concepts into reality.

We need a “printer.”

References:

[1] https://www.pcworld.com/article/3151058/nvidias-n1x-could-be-the-jolt-windows-laptops-need-with-one-big-catch.html [2] https://x.com/nvidia/status/2060390710797328574

© Copyright Reserved. Unauthorized reproduction or usage in any form is strictly prohibited. Offenders will be prosecuted.