SK Hynix breaks ground on a new facility critical to its Nvidia supply chain, expanding advanced packaging and memory capacity.
SK Hynix is expanding its footprint in the U.S. AI supply chain, breaking ground Thursday on an approximately $4 billion high-bandwidth memory (HBM) packaging and research complex in West Lafayette, Indiana. The facility addresses a critical bottleneck in Nvidia-powered AI systems, with SK Hynix projecting sustained demand strong enough to justify significantly expanded capacity through the end of the decade.
As one of the world’s largest memory-chip manufacturers and a leading HBM supplier, SK Hynix produces memory that operates differently from conventional DRAM. By stacking multiple memory chips vertically, HBM delivers massive quantities of data to AI accelerators at extremely high speeds, making it indispensable for modern AI workloads.
Spanning 133.5 acres, the Indiana campus will house advanced HBM packaging production and R&D facilities, with operations slated to begin in the second half of 2028. SK Hynix estimates the project and its surrounding supply chain will generate roughly 7,000 direct and indirect jobs.
The initiative received substantial federal backing, including up to $458 million awarded under the CHIPS Act and eligibility for as much as $500 million in loans. The Commerce Department has explicitly identified HBM as a key constraint in the broader AI supply chain, underscoring the strategic importance of domestic production.
The project’s timeline aligns with accelerating infrastructure demand, recently validated by Nvidia’s latest earnings report. For fiscal second quarter, Nvidia posted revenue of $96.2 billion, a 106% year-over-year increase, while Data Center revenue surged 117% to $89 billion. The company expects approximately $108 billion in revenue for the next quarter alone.
For SK Hynix investors, the Indiana development prioritizes long-term capacity positioning over near-term earnings impact. Key indicators to monitor include HBM shipment growth, pricing trends, customer commitments, and whether the facility commences operations on schedule in 2028. Nvidia’s ongoing Data Center expansion reinforces the demand outlook, particularly as next-generation platforms require increasingly sophisticated memory architectures.
Nevertheless, the extended construction horizon introduces notable risks. A deceleration in AI infrastructure spending prior to 2028, accelerated capacity ramp-ups from competitors such as Samsung or Micron, or softer HBM pricing could compress projected returns.
Should AI compute demand remain elevated, SK Hynix’s first U.S.-based advanced HBM packaging hub would enhance both customer proximity and its competitive standing within one of the AI sector’s most constrained supply chains.