The race to build AI infrastructure is moving beyond GPUs and data centers to the memory systems that feed them. SK hynix has broken ground on a more than $4 billion advanced packaging facility in Indiana that will become the company’s first U.S. production base for High Bandwidth Memory (HBM), with mass production of next-generation HBM planned for the second half of 2029.
The artificial intelligence infrastructure boom is creating a new bottleneck: getting enough high-performance memory close enough to AI processors to keep them supplied with data.
SK hynix is responding with a major expansion of its U.S. manufacturing footprint.
The South Korean memory semiconductor company has broken ground on an advanced packaging facility in West Lafayette, Indiana, designed to become its first U.S. production base for High Bandwidth Memory (HBM). SK hynix expects to invest more than $4 billion in the facility, open its cleanroom by October 2028 and begin mass production of next-generation HBM in the second half of 2029.
The project is significant for the AI hardware market because HBM has become a critical component of accelerated computing systems used to train and run increasingly large AI models.
Unlike conventional memory, HBM uses vertically stacked memory dies and a high-bandwidth interface to move large volumes of data between memory and processors. That architecture makes it particularly important for AI accelerators, where processor performance can be constrained if data cannot reach the compute engine quickly enough.
SK hynix’s Indiana operation is therefore about more than adding another semiconductor factory. It is an attempt to bring part of the AI memory supply chain closer to the U.S. customers building the infrastructure behind generative AI.
A new link in the AI hardware supply chain
The Indiana facility will operate as part of a Korea-U.S. manufacturing network rather than as an isolated production site.
SK hynix plans to manufacture wafers in South Korea before sending them to Indiana for advanced packaging and testing. Finished products will then be supplied to U.S. customers as Made-in-USA HBM.
That distinction matters because advanced packaging has become increasingly important as semiconductor manufacturers push beyond simply shrinking transistor dimensions.
Modern AI accelerators combine enormous computing capability with complex memory architectures. Packaging determines how effectively those components communicate, and HBM is typically integrated closely with processors to provide the bandwidth required by demanding workloads.
The Indiana site is expected to employ approximately 1,000 people once commercial operations are established.
SK hynix also plans to build an Advanced Packaging R&D Testbed alongside its production lines. The facility is intended to give customers, universities and industry partners a location to develop packaging technologies, build prototypes and validate performance.
That could make the site relevant not only to manufacturing but also to the next generation of AI hardware research.
HBM has become strategic AI infrastructure
The importance of HBM has grown alongside the adoption of GPUs and other AI accelerators.
Companies including NVIDIA, AMD and Google rely on advanced memory technologies to support increasingly demanding AI workloads. NVIDIA’s data-center accelerators, for example, pair high-performance compute architectures with HBM to provide the memory bandwidth needed for large-scale AI training and inference.
The resulting demand has turned HBM into one of the most strategically important segments of the memory semiconductor market.
SK hynix competes in that market with Samsung Electronics and Micron Technology, both of which are investing heavily in high-bandwidth memory capacity and next-generation packaging.
For AI infrastructure buyers, that competition is important because accelerator availability is no longer determined solely by GPU manufacturing capacity. Memory supply, advanced packaging capacity and the ability to integrate those components have become part of the same equation.
Indiana becomes an AI semiconductor hub
SK hynix’s investment also fits into a broader U.S. effort to rebuild domestic semiconductor manufacturing.
The CHIPS and Science Act has encouraged semiconductor companies to establish or expand manufacturing and research operations in the United States, with policymakers increasingly viewing advanced chips as both an economic and national-security priority.
Indiana is positioning itself as part of that semiconductor ecosystem, with Purdue University playing a prominent role in engineering research and talent development.
SK hynix has signed a memorandum of understanding with Purdue for joint research into advanced packaging and system integration, including HBM.
The partnership could help address another constraint facing the semiconductor industry: specialized engineering talent.
The company says more than 100 companies are being considered as potential suppliers of materials, components and equipment for the Indiana operation. SK hynix also estimates that approximately 7,000 direct and indirect jobs could be created across construction and subsequent operations.
That creates the possibility of an ecosystem effect, where a major semiconductor investment attracts suppliers, engineering expertise and additional technology companies.
The supply-chain implications extend beyond memory
For U.S. AI companies, the project could eventually provide another domestic source of advanced memory packaging.
That does not mean the United States will become independent of Asian semiconductor supply chains. SK hynix’s own manufacturing model demonstrates the opposite: wafers will continue to originate in South Korea before moving to Indiana for packaging and testing.
Instead, the project creates a distributed supply chain connecting Korean semiconductor manufacturing expertise with U.S. packaging capacity and AI customers.
That model could become increasingly important as governments and technology companies seek supply-chain resilience without completely abandoning the highly specialized global semiconductor ecosystem.
It also illustrates how AI infrastructure is becoming a geopolitical industry.
The competition between the U.S. and China over advanced computing has placed GPUs, memory, semiconductor manufacturing equipment and packaging technologies under increasing strategic scrutiny. South Korea, meanwhile, remains one of the world’s most important semiconductor manufacturing centers.
SK hynix’s Indiana investment sits directly at that intersection.
Advanced packaging could become the next AI hardware battleground
The semiconductor industry has traditionally focused heavily on transistor density. AI is changing the equation.
As accelerator architectures become larger and more sophisticated, simply producing smaller transistors is no longer enough. Memory bandwidth, chip-to-chip communication, power efficiency and packaging increasingly determine overall system performance.
That makes advanced packaging a potential competitive differentiator.
SK hynix’s planned R&D testbed suggests the company expects packaging innovation to remain important as AI accelerators evolve.
The company is also developing its U.S. operation at a time when AI infrastructure spending continues to expand. Data centers are being built around increasingly dense accelerator clusters, while enterprises and cloud providers are deploying larger models and more inference capacity.
Every additional accelerator increases demand for high-performance memory.
A long-term bet on AI infrastructure
The Indiana facility will not begin HBM mass production until 2029, making it a long-term investment rather than an immediate response to today’s memory shortage.
By then, the AI hardware landscape is likely to look substantially different.
Next-generation accelerators will demand greater memory bandwidth, advanced packaging technologies will become more sophisticated, and AI inference is expected to account for a growing share of computing demand.
For SK hynix, the project provides a U.S.-based manufacturing and R&D presence close to major technology customers.
For the broader AI industry, it highlights a less visible part of the AI race.
The future of artificial intelligence will not be determined only by who builds the most powerful model or fastest GPU. It will also depend on the semiconductor supply chains, memory architectures and packaging technologies capable of supporting those systems at scale.
SK hynix is betting that some of that infrastructure will increasingly be built in America.
Market Landscape
The HBM market has become a strategic component of the AI semiconductor ecosystem, with SK hynix, Samsung and Micron competing to expand capacity and develop faster generations of memory.
The market is closely tied to demand for AI accelerators from NVIDIA, AMD and hyperscale cloud providers. As model sizes and inference workloads grow, memory bandwidth becomes increasingly important to overall accelerator performance.
At the same time, advanced packaging is becoming a strategic manufacturing capability. The ability to combine compute and memory efficiently can affect performance, power consumption, yield and ultimately the economics of AI infrastructure.
SK hynix’s Indiana project therefore represents three overlapping trends:
- AI infrastructure localization in the United States
- Growing strategic importance of HBM
- The rise of advanced packaging as a semiconductor differentiator
The facility also demonstrates that semiconductor localization does not necessarily mean fully domestic supply chains. Instead, future production may increasingly involve geographically distributed manufacturing networks connecting Asian wafer fabrication with U.S. packaging, testing and customer operations.
Top Insights
- SK hynix is investing more than $4 billion in Indiana, establishing its first U.S. HBM production hub as AI accelerator demand drives memory requirements.
- The facility will package and test Korean-made wafers, creating a Korea-U.S. manufacturing chain designed to deliver Made-in-USA HBM to American customers.
- Advanced packaging is becoming critical AI infrastructure, connecting high-performance processors with HBM while addressing bandwidth, power and system integration requirements.
- Purdue University will collaborate with SK hynix on R&D, potentially strengthening advanced packaging research and semiconductor talent development across the U.S. Midwest.
- The project intensifies competition with Samsung and Micron, as AI infrastructure spending makes HBM capacity increasingly strategic for semiconductor manufacturers and accelerator suppliers.
Power Tomorrow’s Intelligence — Build It with TechEdgeAI
