Lisa Su Gives Mark “The Zucc” as AMD Joins Nvidia Inside Meta’s 30-Data-Center AI Compound

It’s called diversification… look it up…

The man, the myth, the lizard (allegedly), Mark Zuckerberg wants everyone to know he’s down to Zucc… if you just so happen to sell his favorite flavor of chips.

Last week, he looked at Nvidia and said, “Don’t be shy… give me the bulk discount,” before committing to millions of GPUs for their lofty data center buildout (coming soon to a rural area near you).

And based on today’s news? Zuck seems to have gotten horny and decided to send AMD the infamous “you up?” text in the middle of the night. In other words: Meta just inked a multiyear deal with AMD to deploy up to 6 gigawatts of GPUs across its AI data centers.

And believe me, I know that sounds like just a small package of lightbulbs… but for context, 1 gigawatt of AI compute can run into the tens of billions of dollars when you factor in chips, racks, networking, cooling, land, and the electric bill that will 100% cause a national outage once a week. And Meta wants six of those from AMD alone.

Again, this is happening days after committing to millions of Nvidia processors. Translation: Meta is building the AI equivalent of a nuclear submarine fleet… and buying missiles from two different defense contractors just to keep everyone sharp.

(Source: Reuters)

According to reports, the AMD deployment kicks off with customized GPUs… meaning these are better than the “off-the-shelf” chips. AMD’s MI450 GPUs, delivered in its Helios rack-scale systems, are being tuned specifically for Meta’s workloads.  That’s a lot different from Meta’s (I’ll take what I can get) Nvidia relationship.

AMD CEO Lisa Su wasted no time running to the mic and saying her company is delivering “high-performance, energy-efficient infrastructure optimized for Meta’s workloads.”

Translation: “You better watch you a**, Jensen.”

All jokes aside, this is big time for AMD… who’s constantly viewed as Nvidia’s younger, dumber brother. Nvidia controls roughly 90% of the AI chip market and has a market cap of $4.65 TRILLION (although that number will fold like a lawn chair if Michael Burry has anything to say about it). 

On the other hand, AMD is valued around $320 billion. Respectable. But still hardly a threat to Jensen’s GPU empire. Meta, meanwhile, is throwing around up to $135 billion in capital expenditures this year and they’ve got thirty data centers planned. Twenty-six of them just so happen to be in the U.S. (if you happen to live in Louisiana or Ohio, I’m sorry).

As much as I’d love to paint the picture that Zuck is dumping leather jacket man for Lisa… that’s not the case. This move is simply your classic hedge (you know, what you were supposed to do when you went all in on that ONE company you swore was gonna moon).

If you’re building frontier AI models, recommendation engines, ad systems, generative video tools, and who-knows-what-else inside the metaverse bunker… you don’t care whose logo is on the silicon.

You care about who can ship at scale, who can customize, who can keep the lights on, and who won’t ghost you when demand goes vertical.

Sam Altman (read: OpenAI) already struck a similar deal with AMD last year. Now Meta’s doing the same.

Needless to say, right now, it feels like everyone’s winning… except the power grid.

Elon needs to get a head start on those space data centers and fast… before America turns into one giant data center.

At the time of publishing this article, Stocks.News holds positions in Meta as mentioned in the article.