Nvidia and SK Hynix Strike $500 Billion Deal to Lock Down AI Memory Supply

A $500 Billion Deal Signed in San Francisco

SK Group and Nvidia announced on July 24, 2026, at an AI summit in San Francisco, plans for a partnership worth more than $500 billion. The deal spans both the construction of AI factories and the supply of next generation memory for Nvidia's chips. According to CNBC and Nvidia Newsroom, the two sides signed letters of intent to formalize the agreement, meaning the final contract is not yet fully binding even though the strategic direction has already been made public.

A 2 Gigawatt AI Factory

At the center of the partnership is a 2 gigawatt AI factory that SK Telecom, an SK Group affiliate, will build using Nvidia's next generation computing platform, Vera Rubin DSX. Per SK hynix Newsroom and GlobeNewswire, the facility is expected to come online in 2027 and will serve as the foundation for a new cloud business at SK Telecom, extending the carrier's reach well beyond its traditional telecom operations in South Korea.

HBM4: The Memory Behind Vera Rubin

In parallel, Nvidia and SK hynix, one of the world's largest memory chipmakers alongside Samsung and Micron, established a long-term partnership to secure and co-develop next generation AI memory, including High Bandwidth Memory, HBM, and specifically HBM4. That memory will run inside Nvidia's Vera Rubin platforms, making SK hynix a strategic supplier for a component now as fiercely contested as the graphics processor itself.

Memory Becomes AI's New Bottleneck

The announcement came just two days after Nvidia detailed, according to HotHardware and EasternHerald, a similar buildout of Vera Rubin racks at partners CoreWeave, Google Cloud, Microsoft Azure and Oracle Cloud Infrastructure. Taken together, the announcements make clear that memory supply, not just graphics compute capacity, has become one of the central battlegrounds in the race to build AI infrastructure.

Memory Scarcity as Economic Security

The deal also reinforces a narrative SK Group's own leadership has pushed throughout the year: that AI memory scarcity amounts to a matter of economic security, not just a one off operational problem. By locking in supply contracts at this scale, Nvidia and SK hynix are signaling that they expect demand for HBM to keep outpacing available supply for an extended period.

What This Means for Brazilian Agencies and SMBs

For marketing and customer service agencies in Brazil running AI workloads, the takeaway is direct: memory chip scarcity, not just GPU scarcity, has been a hidden driver of AI infrastructure cost and availability. A $500 billion supply lock deal between Nvidia and one of the world's top memory makers signals the industry expects this bottleneck to persist for years. In practice, that likely keeps cloud AI compute prices from falling as fast as raw model capability improves, an important expectation to set for any business planning multi year AI infrastructure budgets. Anyone counting on predictable cloud costs should treat this scarcity as structural, not temporary.