On June 8th, NVIDIA and South Korea’s SK Group officially rolled out their fresh partnership roadmap.
NVIDIA and SK Hynix have officially locked in a long-term tech alliance. They’re teaming up to co-develop next-gen memory specifically built for global AI data centers, while also integrating AI-driven workflows right into semiconductor design and fabrication.
According to SK Hynix, this partnership kicks off a deep dive into crafting next-gen storage alongside NVIDIA for AI factories worldwide. They’ll keep a steady pipeline flowing with chips that meet every requirement on NVIDIA’s infrastructure roadmap. Since bringing high-end memory to market takes time, this long-term tie-up is essentially a lock on stable supply. On top of that, it positions SK Hynix to diversify and dive headfirst into the brand-new markets NVIDIA is currently paving the way for.
Under the agreement, SK Hynix will co-engineer custom memory for NVIDIA’s Vera Rubin AI supercomputers, Vera CPUs, RTX Spark PCs, and the Jetson Thor robotics compute platform. This strategically places them squarely inside NVIDIA’s booming new frontiers, including core AI infrastructure, consumer AI devices, and physical AI systems.
When it comes to semiconductor manufacturing, SK Hynix will tap into NVIDIA’s CUDA-X libraries and PhysicsNeMo framework to dramatically speed up chip simulations and lithography workflows. At the same time, they’re leveraging Omniverse and cuOpt to spin up digital twins of their fabs, driving the plants straight into autonomous, self-optimizing operations.
Before the headline broke, an SK Hynix spokesperson confirmed that SK Group Chairman Chey Tae-won and NVIDIA CEO Jensen Huang had been deep in talks over the partnership details. Jensen didn’t pull any punches when speaking to the press:“We’re covering a ton of ground here—from AI supercomputers and CPUs to next-gen personal computers and robotics. His visit was all about finalizing the roadmap, and we’re dropping the official news on Monday.”Jensen also gave us a straight answer on the chip shortage: there’s zero sign of it easing up. The current crunch is going to stick around for years.“The entire supply chain is stretched thin. From wafers and packaging to silicon photonics, everything is in high demand because buyer appetite is just massive. This supply squeeze is going to last for several more years.”
In a refreshing break from stuffy corporate meetings, Jensen, Chey, SK Hynix President Kweon Eun-Soo, and other SK Group executives actually got together at a classic Seoul fried chicken joint. While digging into Korea’s famous chicken-and-beer combo, they hashed out the nitty-gritty of the partnership.
Going back to June 2nd, during a media roundtable at GTC Taipei 2026, Jensen publicly shared his thoughts on the industry chatter surrounding the pay dispute between SK Hynix and Samsung Electronics. In his view, companies “should reward employees as generously as possible”. While he stressed it’s not his place to critique another company’s compensation model, he did open up:“I absolutely try to pay my team as much as I possibly can, but that doesn’t mean my approach is necessarily the right one.”
Lately, the Korean memory chip sector has been riding a massive wave thanks to soaring global AI compute demand. Corporate profits are looking great, and the way the big players split the pot has sparked serious industry debate. Notably, SK Hynix and its labor union struck a landmark deal back last September. They completely scrapped the old rule capping profit-sharing bonuses at “10 times base salary”, replacing it with a fresh formula that funnels 10% of annual operating profits directly into the employee bonus pool.
According to SK Hynix’s 2025 fiscal report, the company pulled in 97.15 trillion KRW in total revenue, with operating profit hitting 47.21 trillion KRW (a 49% margin) and net profit landing at 42.95 trillion KRW (a 44% margin). These numbers completely eclipse the record-breaking highs SK Hynix set back in 2024. Revenue surged past 30 trillion KRW year-over-year, while operating profits literally doubled, smashing previous annual records. When you run the math, that works out to an average bonus of 140 million KRW (roughly 650,000 RMB) for each of the 30,000-plus Hynix employees.
And here’s where it gets even juicier. Based on projections from Macquarie Securities, if SK Hynix pushes its 2027 operating profit to 447 trillion KRW, we’re looking at per-person bonuses nearing 610,000 RMB, calculated against that 10% payout rule and a workforce of 35,000.
In the midst of this AI super-cycle, global tech giants aren’t showing any signs of slowing their hardware buying spree. The chip shortage isn’t getting solved anytime soon. Earlier this March, SK Group Chairman Chey laid out a sobering timeline, noting that systemic bottlenecks in chip manufacturing mean global memory shortages could easily drag on for four to five years.
During that exact same GTC Taipei 2026 media roundtable, Jensen didn’t sugarcoat the global supply chain situation. He was upfront about how upstream bottlenecks in wafer foundries and component manufacturing are still causing periodic supply snags for NVIDIA’s product lineup. However, he pointed out that current production scheduling is already prioritizing initial shipments for newly launched PC chips and AI agent processors.
UBS Analyst Timothy Arcuri recently highlighted in a report that hyperscalers like Microsoft, Google, Amazon, and Meta are already securing about 60% to 70% of their future server memory chip allocations.“Hyperscale cloud providers are increasingly willing to trade premium pricing for multi-year supply visibility and much tighter predictability on future deployment economics,” UBS noted in their latest report.
On May 26th, SK Hynix officially debuted its “iHBM” technology. By embedding an integrated cooling component directly into the HBM package, “ICE* drastically cuts down on heat buildup during operation. SK Hynix plans to roll this iHBM tech into next-generation products like HBM5. It’s built to tackle the brutal thermal demands of ultra-high-density, high-bandwidth setups in HPC and AI data centers, ultimately boosting overall system stability and operational efficiency.
