On August 27, 2026, SK hynix broke ground on an advanced HBM packaging plant in West Lafayette, Indiana. Total investment exceeds $4 billion, with mass production of next-generation HBM now targeted for late 2029. In an April 2024 announcement, the company had planned to invest $3.87 billion and begin mass production in the second half of 2028. The groundbreaking moves the project from concept to construction, but the mass production timeline has slipped by roughly a year.
SK hynix calls the new facility its "first HBM production base in the United States." However, the plan does not move front-end processes—where DRAM wafers are made—to the US. Wafers that have completed front-end processing in Korea will be shipped to Indiana, where the facility will handle stacking, encapsulation, and testing. It's important to distinguish between what "US-made HBM" actually covers and what remains in Korea.
What "US-made HBM" Actually Covers
What the Indiana plant receives are HBM wafers that have already completed circuit formation in Korea. According to SK hynix's explanation, the US facility will perform advanced packaging and testing, then supply the finished HBM products to US customers. Because front-end DRAM production will continue to depend on Korea, this groundbreaking does not bring the entire HBM process chain onto US soil.
HBM works by vertically stacking multiple DRAM dies and connecting them through Through-Silicon Vias (TSVs). SK hynix has also used its MR-MUF technology, which bonds bumps between dies in a single step and fills the gaps with a protective material. As capacity and bandwidth increase, the post-wafer processes of stacking, encapsulation, and inspection become critical to product performance and yield while managing heat dissipation and durability. This back-end work is what the Indiana facility will handle.
One more stage remains further down the supply chain. According to SK hynix, after HBM is delivered to system companies, it gets integrated into the same package as logic chips such as GPUs. So while the Indiana plant will complete HBM products within the US, it is not a facility that assembles entire AI accelerators. What the US is newly securing is capacity in the memory back-end process—a segment that has often constrained AI computing infrastructure supply.
Two distinct packaging processes overlap here. SK hynix stacks DRAM dies and ships them as HBM with established electrical connections and heat dissipation. Placing that HBM near a GPU and finishing it into an AI-oriented package with wide interconnect bandwidth is a separate step performed later by system companies. The Indiana investment adds new production capacity for the former process in the US; it has not been announced as expanding capacity for the latter process or for overall GPU supply.
Still, the effect of localizing this back-end process is significant. The US Department of Commerce has repeatedly identified HBM production capacity as one of the major supply constraints for the AI industry. By maintaining wafer fabrication in Korea while placing stacking and inspection closer to US customers, the supply chain now spans both Korea and the United States. This is better understood as an investment that splits the process across two regions rather than a complete domestication of production.
October 2028 Is Not the Start of Mass Production
Under the latest plan, the cleanroom will open by October 2028, with mass production of next-generation HBM beginning in the second half of 2029. Since the April 2024 announcement had targeted the second half of 2028 for mass production, the schedule has been pushed back by roughly a year. The investment amount has also grown from $3.87 billion to over $4 billion, a difference of at least $130 million. SK hynix has not disclosed the reasons for the increased investment or the delay.
Completing a cleanroom is not the same as shipping products. Manufacturing equipment must be installed, processes brought online, yields improved, and quality requirements met before mass production can begin. This time, SK hynix did not disclose the HBM generation, annual production capacity, or customer names. The roughly one-year gap between the October 2028 cleanroom opening and late 2029 mass production is the period needed to convert the building into functioning production capacity.
The change in mass production timing was disclosed alongside the groundbreaking, itself a step forward. Looking only at the fact that construction has begun, the plan appears more concrete—but the timing for when this new supply capacity will actually reach the market has moved further away than originally planned. When assessing AI memory supply and demand, the Indiana facility's output cannot be counted as adding to 2028 supply. The company's current stated start date for supply is the second half of 2029.
The plant is expected to employ about 1,000 people once commercial operations begin. The approximately 7,000 figure mentioned in the announcement is not the plant's direct employment count, but the combined total of direct and indirect jobs generated from construction through post-operation phases. More than 100 companies supplying materials, parts, and equipment are also described as being "under consideration"—not as companies with signed contracts. The larger the number, the more important it becomes to separate what it actually covers and over what time period.
Up to $958 Million in Federal Support Comes with Conditions
In December 2024, the US Department of Commerce finalized support for SK hynix under the CHIPS Act: up to $458 million in direct funding and up to $500 million in loans, for a combined ceiling of $958 million. However, the direct funding is disbursed based on achieving construction and operational milestones. A formal support decision does not mean the full amount has already been transferred to SK hynix.
As of August of that year, this had been a non-binding preliminary agreement anticipating up to $450 million in direct funding. After completing its review in December, the Commerce Department advanced to a final support decision of up to $458 million. Even so, disbursement conditions remain. The policy decision, the finalized funding ceiling, and the actual transfer of funds cannot be treated as a single event.
As of 2024, the state of Indiana had also offered up to $80 million in conditional payments, up to $554.7 million in tax refunds, and up to $45 million in surrounding infrastructure support. These, too, are performance-based incentives where eligibility to claim funds arises only after the investment is executed. Because this project involves corporate investment and public funding accumulating simultaneously, progress needs to be verified milestone by milestone beyond the groundbreaking ceremony.
The state's offer also includes training grants and manufacturing readiness grants of up to $3 million each. Purdue University and the Purdue Research Foundation have offered support valued at approximately $60 million, including discounted land of 90 acres plus an option for an additional 30 acres. Because the federal, state, and university support serves different purposes, simply adding up the figures does not yield the actual cash amount that will flow into the plant.
The US aim here is to bring onshore the HBM back-end process that has been a bottleneck in the AI semiconductor supply chain. To gauge whether federal support has translated into actual production capacity, it is more reliable to track the sequence of achievements—cleanroom completion, equipment operation, and mass production start—than to focus on the ceiling amount of support. The new late-2029 deadline will also serve as the deadline against which the subsidy's results are measured.
An R&D Hub with Purdue Joins the Production Plant
Adjacent to the production line will be an "Advanced Packaging R&D Testbed," a facility where customers, universities, and business partners will jointly develop next-generation packaging technologies, along with prototyping and performance verification. On the day of the groundbreaking, SK hynix also signed an MOU with Purdue University to jointly research system integration—including HBM—and advanced packaging technologies.
The 2024 plan had also outlined research into heterogeneous integration with Purdue's Birck Nanotechnology Center and others, with an intention to expand cooperation into memory-centric system design and in-memory/near-memory computing. On the manufacturing workforce front, Purdue University and Ivy Tech Community College are considering training programs and interdisciplinary degree programs. The goal is to connect research and education to the production plan in order to secure the roughly 1,000 employees the plant will need locally.
The significance of placing production and research in the same region lies in the ability to verify next-generation HBM design challenges using prototyping facilities and feed that knowledge back into manufacturing. But the MOU does not guarantee research outcomes or mass-production yields. Which companies will participate, which technologies will be prototyped, and how many months customer qualification will take have not been disclosed.
Whether the Indiana plant actually strengthens the US AI supply chain will depend on whether it can reliably process wafers arriving from Korea and ship them by late 2029 in the quality and quantity customers need. Once equipment utilization, yields, and annual production capacity are disclosed after the cleanroom opens in October 2028, it will become possible to judge how much additional supply the more-than-$4-billion investment actually delivers.