SK Hynix secured a multiyear partnership with Nvidia to develop next-generation hardware for so-called AI factories, a deal that comes just days after securing a supplier deal for Nvidia’s upcoming Vera Rubin platform.

Having long enjoyed a co-engineering partnership, the pair will now work on next-generation memory systems for AI infrastructure, with specific efforts covering Nvidia Vera central processing units (CPUs), its Spark line of desktop supercomputers, and Jetson Thor robotic computing platforms.

The multiyear deal will see Nvidia and SK Hynix work on addressing extended development cycles and advanced fabrication, while ultimately ensuring memory supplies can keep pace with the former’s infrastructure roadmap and sustain the buildout of global AI infrastructure.

“AI factories are the engines of the next industrial revolution, and advanced memory is essential to their performance,” Nvidia CEO Jensen Huang noted on the deal. “Together, we will codevelop the next generation of memory for AI factories and support the accelerating global expansion of AI infrastructure – from frontier model training to agentic and physical AI.”

One of the big three memory makers, SK Hynix has seen its market cap surpass the $1 trillion mark for the first time amid unprecedented demand for its hardware. That demand, though, is vastly outstripping supply, with the South Korean giant forced to push lead times as far back as 2028.

It joined its big three rivals in having been certified to supply Nvidia with fourth-generation high-bandwidth memory (HBM4) products, with Huang telling press at the Computex event last week that SK Hynix and its rivals were “in production, and they’re all racing to support Vera Rubin.”

SK Hynix has had to push back production of its HBM4 lines to the third quarter of 2026, instead of the second quarter as originally planned, with prior reports suggesting such delays would impact Nvidia’s Vera Rubin rollout – rumors the chip giant has since denied.

To bolster its stacked workflows, SK Hynix is making use of Nvidia’s CUDA-X software libraries and AI to speed semiconductor simulations for computational lithography workflows, while also employing its PhysicsNeMo framework to accelerate AI physics workflows.

SK Hynix is also using Nvidia platforms like Omniverse to create digital twins of its fabs to optimize manufacturing environments.

The latest multiyear agreement sees Nvidia keep faith with SK Hynix, with SK Hynix Chairman Chey Tae-won saying the pair have “been building toward this for years,” adding: “This partnership reflects the depth of that collaboration.”

“Together, we are codeveloping the next generation of memory for AI factories and applying AI to how we design and manufacture semiconductors – work that will shape the future of AI infrastructure,” Chey added.