SK Hynix Inc. is preparing a major capacity push as the global race to build AI-ready data centres strains the memory supply chain.
The South Korean chipmaker plans to invest 19 trillion won ($12.9 billion) to build a new advanced chip packaging facility, targeting the fast-growing market for high-bandwidth memory (HBM) and other cutting-edge components used in AI hardware.
The company said it will begin construction in April in Cheongju, a city in South Korea’s southern region, and aims to complete the complex by the end of 2027.
SK Hynix is currently the world’s leading supplier of HBM for Nvidia Corp.’s AI accelerators, placing it at the centre of one of the most critical hardware segments powering today’s AI boom.
Cheongju project targets packaging bottlenecks
The new Cheongju site signals a strategic shift toward expanding advanced packaging capability, a part of chipmaking that has become increasingly important for AI-focused products.
Packaging is no longer a minor final step. For high-performance memory like HBM, it can determine how quickly manufacturers can scale supply and deliver chips that meet demanding AI workloads.
The facility is expected to strengthen SK Hynix’s ability to support customers as data centre buildouts accelerate, particularly where high-end memory is needed to pair with AI accelerators.
Global memory supply tightens as AI buildouts grow
The move comes at a time when global memory supply is tightening, raising concerns about potential constraints on AI investment.
Demand for HBM and other advanced memory chips has surged faster than many in the industry anticipated, driven by the pace at which companies are building and upgrading AI data centres.
In the current cycle, memory has shifted from being treated as a commoditised component to becoming a limiting factor.
When supply is tight, it directly affects how quickly data centres can bring new AI accelerator systems online, even if the accelerator chips themselves are available.
Why HBM shortages may persist
The imbalance between supply and demand has been reinforced by practical constraints across the industry.
Even as suppliers work to lift output, long qualification cycles, complex packaging requirements, and limited fabrication capacity slow the pace of expansion.
These dynamics suggest shortages may persist for some time, keeping prices firm and giving memory manufacturers more leverage over customers than in past cycles.
For AI infrastructure builders, that changes procurement priorities, with memory increasingly treated as a key resource rather than an interchangeable input.
SK Hynix’s plan is part of a wider rethink among top memory makers, as the sector responds to demand signals from the AI economy.
Major producers, including Samsung Electronics Co. and Micron Technology Inc., are also reassessing capital expansion strategies, with more focus on faster investment in advanced packaging lines.
SK Hynix expects the HBM market to grow at an average annual rate of 33% from 2025 to 2030, underlining why producers are moving quickly to secure capacity.
The company said the “importance of proactively responding to rising HBM demand is becoming increasingly critical,” highlighting how supply planning has become central to AI hardware growth.
Chey Tae-won, chairman of SK Hynix parent SK Group, also raised supply concerns in November, warning that the industry has entered an era where supply constraints are creating bottlenecks as more companies request access to memory chips.
The post SK Hynix to ramp up advanced chip packaging for AI memory demand appeared first on Invezz
