THE APEX TIMES
Nvidia Signs Multiyear Technology Partnership With SK hynix to Co-Develop AI Memory and Fab Digital Twins
The companies said their collaboration targets next-generation memory for Nvidia’s Vera Rubin AI supercomputers, Vera CPUs, RTX Spark PCs, and Jetson Thor robotics, while using AI tools to speed semiconductor design and autonomous manufacturing.
Nvidia and SK hynix announced a multiyear technology partnership aimed at advancing next-generation memory for the buildout of global “AI factories,” a term Nvidia uses for the end-to-end compute, networking, storage, and manufacturing ecosystem needed to deploy AI at scale. Nvidia said the agreement is also meant to accelerate semiconductor design and manufacturing by pairing memory co-development with AI-assisted chip simulation and fab automation.
Under the deal, SK hynix will “diversify into new markets” Nvidia is creating, with codevelopment and supply support tied to Nvidia’s Vera Rubin AI supercomputers, Vera CPUs, RTX Spark-powered PCs, and Jetson Thor robotic computing platforms. Nvidia framed the move as a way to keep advanced memory supply aligned with its infrastructure roadmap as AI factories expand from frontier training to agentic and physical AI use cases.
Nvidia said the partnership leverages its CUDA-X libraries and PhysicsNeMo framework to speed semiconductor simulation workloads at SK hynix. In the company’s description, that includes technology computer-aided design, computational lithography workflows, and acceleration of in-house simulation code and AI physics workflows used during chip development. Nvidia also suggested the approach could extend into broader electronic design automation and simulation ecosystems.
A second pillar of the announcement centers on fab “digital twins,” or virtual 3D replicas of semiconductor manufacturing environments used to visualize and test changes before they reach production. Nvidia said SK hynix will build these digital twins using Omniverse libraries and OpenUSD scene optimization pipelines, then apply operational optimization via cuOpt, and Metropolis, which Nvidia describes as a platform for AI-driven decision-making tied to physical systems. Nvidia added that the digital twins are intended to support the movement of autonomous mobile robots and other fab assets, and that the companies are exploring ways to connect the twins with legacy software and agentic AI workflows.
Industry analysts and component makers have argued that memory is one of the hardest parts to scale in advanced AI systems because it requires long lead times and heavy capital investment. In coverage of the announcement, Tom’s Hardware said the deal is designed to address extended development cycles and the large capital expenditures required for advanced memory, while improving planning visibility for Nvidia and more predictable demand for SK hynix.
Beyond Nvidia’s high-level description, Tom’s Hardware reported specific memory families and storage included in the “initial part” of the cooperation. It said the early scope covers HBM4, LPDDR5X, and 3D NAND for the Vera Rubin systems, LPDDR5X for standalone Vera processors, LPDDR5X and 3D NAND for RTX Spark-powered PCs, and LPDDR5X and 3D NAND for Jetson Thor. Tom’s Hardware also noted that the joint press release did not spell out whether memory would be customized beyond maintaining compatibility with Nvidia processors, suggesting the collaboration may focus on industry-standard solutions designed for Nvidia platforms.
The companies did not disclose financial terms, the exact length of supply commitments, or detailed performance targets, such as bandwidth, capacity, or timing, in Nvidia’s posted release. The SK hynix newsroom page was not accessible during this drafting session, and the Yahoo Finance item referenced in the initial report could not be retrieved due to access limits, so some market headlines may reflect additional details not confirmed directly from the companies’ own language.
What to watch next is whether the partnership translates into measurable improvements in memory availability and manufacturing throughput as Nvidia’s Vera Rubin platform ramps and as RTX Spark PCs and Jetson Thor robotics expand beyond early deployments. The most immediate indicates are likely to come from subsequent product announcements and any follow-on disclosures from either company about digital twin rollouts and how AI-driven optimization is applied inside real fabs.
Why It Matters
- Securing and co-developing advanced memory can reduce risk that AI system rollouts stall due to memory bottlenecks.
- By tying memory collaboration to digital twins and autonomous fab optimization, the deal aims to shorten time from design intent to production execution.
- The announcement extends Nvidia’s relationship with a key memory supplier beyond chips and into manufacturing software and decision systems.
- If the approach works, it could become a template for how other platform makers and memory suppliers coordinate roadmaps, tooling, and capacity planning.
Sources
Key Facts
- Nvidia and SK hynix said they signed a multiyear technology partnership focused on next-generation memory for AI factory buildouts.
- The announced scope spans Nvidia’s Vera Rubin AI supercomputers, Vera CPUs, RTX Spark PCs, and Jetson Thor robotics platforms.
- Nvidia said SK hynix will use CUDA-X libraries and PhysicsNeMo to speed semiconductor simulation and AI physics workflows.
- The companies said SK hynix will develop fab digital twins using Omniverse and OpenUSD, and use cuOpt and Metropolis to support autonomous manufacturing optimization.
- Nvidia said the partnership is intended to address extended development cycles for advanced memory supply.
- Nvidia’s release did not provide disclosed financial terms or detailed memory specifications; Tom’s Hardware reported initial memory families including HBM4, LPDDR5X, and 3D NAND based on the joint announcement.
Technology Related
U.S. stocks slipped before the bell as traders weighed Meta and Tesla earnings
Major indexes moved lower ahead of the open Thursday, with investors turning their attention to the latest earnings outlines from Meta (META) and Tesla, two of the most watched technology names in the market.
AMD share swing prompts traders to ask whether anything fundamental changed
A market recap focused on how a recent move in AMD’s stock price should be interpreted, while offering limited new detail about the company’s near-term operating outlook.
Anthropic rolls out “Claude Duet” phone agent that proactively handles email and shopping, upping pressure on rivals like Meta and OpenAI
The new assistant is positioned as a more proactive consumer “agent,” moving from chat toward tasks such as reading messages and making purchases. The move lands in a category Meta has also been working, including efforts the company has described publicly as reaching millions of users.
SEEQC says NVIDIA-linked decoding approach improves quantum error correction performance
In a new set of results, the chip-based quantum computing company SEEQC reported performance gains tied to a digital interface-optimized decoder architecture built around NVIDIA Ising and NVQLink, aiming to make quantum error correction more practical.
Nvidia’s DSX POD enters the conversation as executives weigh how fast AI gets deployed
In a Wall Street panel discussion tied to Nvidia’s latest DSX POD announcement, investors and analysts focused on one question: what happens when AI moves beyond pilots and into day-to-day enterprise workflows.
Rockwell Automation adds NVIDIA Nemotron support to FactoryTalk Conductor
The industrial automation company says its FactoryTalk Conductor platform will recognize NVIDIA’s Nemotron open models, aiming to bring more configurable AI capabilities into industrial operations workflows.
Apple pushes iPhone Duo developer updates live this Friday, expanding apps across fold positions
Apple says apps optimized for its first foldable iPhone will roll out starting Friday, highlighted by seamless transitions between the outer and inner displays, new iOS tooling for folding layouts, and early feature work from Netflix, Microsoft and major productivity, media and games partners.
QumulusAI boosts closed-lease funding tied to NVIDIA Blackwell GPUs, adding $123.6 million via Tech Finance
The neocloud infrastructure provider says the expanded equipment lease financing with Tech Finance Co., LLC will support additional NVIDIA Blackwell graphics processing units without issuing new equity.
Yahoo Finance asks investors whether to trim or add Intel amid ongoing market debate
A new Yahoo Finance piece frames the latest question facing Intel shareholders, weighing the case for adding versus trimming, but does not offer new Intel disclosures in the post itself.
Oracle’s “wildcard” funding mix targets Microsoft’s cloud war, but the plan’s staying power is unclear
A market report argues Oracle is using a high-variance strategy to bankroll its push against Microsoft in cloud and AI spending, mixing stock sales, new debt, and customer prepayments. Whether that financing holds together could shape who gains control of the next wave of enterprise computing budgets.
