Nvidia announces $500 billion Korean memory supply alliance, locking in DRAM/HBM allocation ahead of August 26 product launches.
Nvidia and South Korea's SK Group unveiled a partnership on Monday valued at more than $500 billion, spanning AI factory construction and next-generation memory chip supply. The memorandums of understanding cover a two-gigawatt facility to be built by SK Telecom on Nvidia's forthcoming Vera Rubin DSX platform, with the first AI factories slated to come online in 2027.
The deal also establishes a long-term development pact between Nvidia and SK Hynix focused on next-generation AI memory, including High-Bandwidth Memory chips. That component category has emerged as one of the tightest bottlenecks in the current AI buildout, making early capacity commitments strategically critical for the chip designer.
The Korean agreement extends a pattern of ecosystem-deepening moves that has defined Nvidia's recent months. In late July, Nvidia announced a long-term partnership with Safe Superintelligence, the AI startup founded by Ilya Sutskever, alongside an equity investment valued at approximately $5 billion. Nvidia is providing SSI access to the upcoming Vera Rubin platform, a step expected to substantially increase the startup's compute capacity.
SpaceX has also entered the picture. Elon Musk announced that the company's new Starmind AI1 orbital constellation will run exclusively on Nvidia hardware, with reports indicating that the aerospace firm has secured a substantial allocation of the next graphics chip generation for 2027. This arrangement gives Nvidia a significant customer base beyond traditional data center operations.
Through these partnerships and supply agreements, Nvidia is locking in demand for its next chip generation while binding partners more tightly to its platform. However, not all signals are uniformly positive. Media reports indicate Nvidia is currently testing variants of its upcoming Rubin Ultra graphics chips with reduced memory capacity, a direct response to tight High-Bandwidth Memory supply constraints.
The company has pursued additional initiatives as well. Alongside 120 other technology organizations, Nvidia introduced the Shared AI Findings Exchange initiative to standardize information sharing around security risks in agent-based AI. At QuakeCon, the company launched a Verified Priority Access program allowing attendees to purchase GeForce RTX 5090, 5080, and 5070 models directly at suggested retail prices.
Regulatory filings reveal insider sales totaling approximately $1.82 billion through July 2026, with no insider purchases registered during that period. In the past 90 days, Nvidia insiders have executed 13 transactions—all sales—totaling roughly $767.2 million, with board member Mark A. Stevens among the most prominent sellers. Investment firm Blue Grotto Capital also trimmed its position by 199,358 shares.
Market attention now focuses on August 26, when Nvidia reports results for its second fiscal quarter ending in July. Consensus estimates call for revenue around $91.8 billion with earnings per share of $2.08, with analyst projections ranging from $91.0 billion to $93.6 billion in revenue and $2.01 to $2.13 in EPS. The company's first fiscal quarter delivered record revenue of $81.6 billion—an 85 percent year-over-year increase—powered by $75.2 billion in data center sales. Management simultaneously authorized an additional $80 billion share buyback program and raised the quarterly dividend from $0.01 to $0.25 per share.
Wall Street's view remains predominantly constructive, with an average price target of $304.26. Recent analyst assessments include Cantor Fitzgerald at $240 with an Overweight rating, BofA Securities at $235 with a Buy, and KeyBanc, Evercore ISI, Wells Fargo, J.P. Morgan, and Citi Research ranging between $210 and $230, all with buy recommendations. The stock closed Friday at €193.68, roughly 4 percent below its 52-week high of €202.50 reached in May.