AMD stock exploded again on Friday, jumping +7% intraday and smashing past $245 before settling for a new all-time high of $252.
That rally puts the chipmaker up an insane +60% in October alone, officially making it the 25th most valuable company on the planet.
The trigger was a surprise announcement by IBM, which said it’s now running quantum error-correction algorithms on AMD silicon, specifically the field programmable gate array (FPGA) chip.
This puts AMD right in the middle of the global quantum arms race, as IBM is battling it out with Microsoft and Google, which just announced their own quantum breakthrough earlier this week.
IBM tests AMD chips for real-time error correction
Quantum computers use qubits, which can do things normal chips can’t, like simulate how trillions of atoms behave over time. But qubits screw up a lot. One tiny error can derail the whole computation. That’s why error correction is a big deal.
And IBM says it’s cracked it. Even better? Their solution runs on AMD’s chips, and it’s “10 times faster than what is needed,” according to Jay Gambetta, IBM’s director of research.
“It’s not ridiculously expensive.”
The error-correction algorithm was originally announced back in June, but IBM is publishing new results Monday showing the thing works live on AMD hardware. This development also arrives one year ahead of IBM’s own schedule, as they plan to launch their Starling quantum system by 2029.
But this isn’t the only link between IBM and AMD right now. On October 1, the two companies launched a joint effort to power AI research firm Zyphra, based in San Francisco.
Zyphra just raised a $1 billion Series A round to build what it calls a superintelligence lab, focusing on stuff like new neural networks, memory systems, and continuous learning.
Under the deal, IBM will deploy a massive fleet of AMD Instinct MI300X GPUs on IBM Cloud, giving Zyphra access to some of the most powerful AI compute on Earth.
Zyphra deploys AMD full-stack platform on IBM Cloud
Zyphra’s goal is to train frontier multimodal foundation models, basically AI that can handle language, images, and audio all at once. They’ll use it to power Maia, an “AI superagent” meant to boost productivity for enterprise users.
The infrastructure behind this isn’t half-baked. IBM’s cloud cluster is the first large-scale deployment to use a full AMD stack, including Pensando Pollara 400 AI NICs and Ortano DPUs. The initial cluster was delivered in September, with plans to scale up further in 2026.
“This collaboration marks the first time AMD’s full-stack training platform—spanning compute through networking—has been successfully integrated and scaled on IBM Cloud,” said Krithik Puthalath, Zyphra CEO. He said Zyphra is “honored” to lead the development of frontier models using AMD tech.
The backstory here runs deeper. IBM and AMD actually announced a similar cloud accelerator service last year. That platform, which runs MI300X GPUs on IBM Cloud, was built to handle Gen AI models and high-performance computing with better speed and lower power costs.
IBM Cloud boss Alan Peacock called the Zyphra deal part of a strategy to build scalable AI infrastructure that keeps costs in check. Meanwhile, Philip Guido, AMD’s EVP, said the collab is about enabling “real-world outcomes” for businesses building smarter AI.
If you’re reading this, you’re already ahead. Stay there with our newsletter.
