Nvidia Secures $500B SK Hynix Memory Deal for AI Chips

Nvidia locks in critical memory supply with massive SK Hynix partnership

Nvidia has announced a multi-year agreement with South Korean memory manufacturer SK Hynix potentially worth $500 billion, securing supply of high-bandwidth memory essential for its AI processors while funding large-scale data center construction expected to begin operations in 2027.

The deal, revealed late Friday in San Francisco, addresses a critical bottleneck in AI infrastructure: the global shortage of high-bandwidth memory (HBM) that powers advanced graphics processing units. SK Hynix currently leads the HBM production market, making it a strategic partner as Nvidia races to meet surging demand for AI computing capacity.

Massive infrastructure buildout planned

The agreement includes plans for data centers requiring 2 gigawatts of power capacity—a scale that suggests deployment of hundreds of thousands of GPUs. SK Telecom, an SK Hynix affiliate, will construct a cloud business built around Nvidia's Vera Rubin systems.

"The expansion will include a co-develop opportunity for us on the next-generation SK Hynix AI memory, and this will help us secure a stable supply of HBM memory," said Raj Mirpuri, Nvidia's enterprise vice president, during a call with reporters.

The joint development component signals Nvidia's intent to influence future memory specifications rather than simply purchasing existing products.

Why it matters

This deal represents a fundamental shift in AI infrastructure investment beyond U.S. tech giants. Foreign governments and major conglomerates are now committing hundreds of billions to secure computing capacity, indicating that AI buildouts have entered a new phase of global competition. For Nvidia, locking in memory supply addresses its most significant production constraint while the HBM shortage persists industry-wide.

Broader Korean AI investments

Nvidia separately announced a $1 billion investment in Naver, a Korean cloud company constructing data centers around Nvidia GPUs. That project will deliver 200 megawatts of capacity and provide customers in South Korea and internationally the ability to reserve AI computing resources before construction completes.

The announcements came during an AI summit in San Francisco attended by South Korean President Lee Jae Myung and other government officials, underscoring the strategic importance both nations place on AI infrastructure.

SK Hynix, South Korea's second most valuable company, listed on the Nasdaq earlier this month specifically to finance infrastructure developments like this partnership.

CNBC first reported the details of the agreement.