When Nvidia Builds Faster, Google's Playing Chess

Consider this your wake-up call.

Morning Brew}} is the free daily newsletter that powers you up with business news you’ll actually enjoy reading. It’s already trusted by over 4 million people who like their news with a bit more personality, pizazz — and a few games thrown in. Some even come for the crosswords and quizzes, but leave knowing more about the business world than they expected.

Quick, witty, and delivered first thing in the morning, Morning Brew takes less time to read than brewing your coffee — and gives your business brain the boost it needs to stay sharp and in the know.

Try Morning Brew for free

Hey, josh here. check these wild stories out. A.I just keeps evolving globally.

Here's the weird thing about what's happening right now with Nvidia. The company is spending billions of dollars to build a huge new research center in a place called Kiryat Tivon in northern Israel. They're acting like they own the future. But at the exact same time, Google is working on something that could totally mess up everything Nvidia has built.

The Big Nvidia Move

Let's talk about what Nvidia is actually doing. They just announced they're building a massive campus in Kiryat Tivon. We're talking about a building that's bigger than 35 football fields, sitting on land where they're going to hire around 10,000 people. That's a lot of jobs in one place. The whole thing is going to cost billions of shekels—and for those who don't know, that's Israeli money. It'll take years to build, but it's supposed to open by 2031.​

CEO Jensen Huang basically said Israel is like Nvidia's second home. The company already has thousands of workers there, so this is like they're saying, "We're staying. We're going all-in."​

But Here's The Problem

Now here's where things get weird. While Nvidia is busy building this giant campus, Google is secretly working on something that could destroy Nvidia's biggest advantage. It's called TorchTPU.​

To understand why this matters, you need to know one thing: almost every person building AI uses something called PyTorch. It's like the common language for AI engineers. But PyTorch only really works great with Nvidia's chips right now. This is Nvidia's biggest power. If you want to use PyTorch—and most people do—you basically have to buy Nvidia stuff.​

Switching to Google's TPU chips would normally mean completely rewriting all your code. It would cost tons of money and time. So nobody does it. They're stuck with Nvidia. That's the lock-in.​

What TorchTPU Does

TorchTPU changes everything. Google figured out how to make PyTorch work perfectly with their TPU chips without anyone having to rewrite anything. This means you could switch away from Nvidia to Google, and your code would just... work.​

Meta, the company that actually made PyTorch, is helping Google with this. That's important. If Meta says it works, then people will trust it.​

Why This Matters

Right now, Nvidia is acting like they own the future because they built an ecosystem that's impossible to leave. That's a smart move—it makes money, locks in customers, and seems safe. But TorchTPU could change that in one move.​

Google isn't trying to make faster chips. They're trying to make it so you don't need Nvidia's specific stuff anymore. If they pull it off, suddenly Nvidia's most important advantage disappears.

So Nvidia is building expensive campuses and hiring thousands of people, thinking the future is theirs. Google is working quietly on software that could make all of that way less valuable than Nvidia thinks.

That's the real story. One company is building in stone and metal. The other one is trying to change the rules of the game.

Citations:
Times of Israel (December 17, 2025) - Nvidia picks Kiryat Tivon for large R&D campus, bringing tech jobs north​
Open Source For You (December 17, 2025) - Google And Meta Bet On Open Source PyTorch To Break Nvidia's CUDA Lock-In​
ByteIota (December 18, 2025) - Google TorchTPU + Meta Challenge Nvidia's $5T CUDA Lock