Nvidia and SK Hynix partnership positioned as mutual win: Nvidia gains stable HBM supply, SK Hynix gains AI accelerator volume ramp.

Just before last weekend, Nvidia and SK Hynix completed the largest memory deal in history. Under the partnership, SK Hynix will supply Nvidia with high bandwidth memory (HBM) going forward, and the two companies will co-develop future generations of AI memory. Additionally, SK Telecom, SK Hynix's subsidiary, will build a 2-gigawatt AI cloud data center in Korea using Nvidia's Vera Rubin Platform.

Memory represents one of the largest bottlenecks in AI infrastructure build-out today, particularly HBM. To reduce latency and improve power efficiency, GPUs and other AI chips must be paired with HBM. As inference increasingly prioritizes fast memory access over raw compute power, demand for HBM continues to grow.

Expanding HBM capacity faces significant constraints. Manufacturing requires access to extreme ultraviolet lithography (EUV) machines, which are solely produced by ASML and also used to make GPUs and advanced logic chips—creating intense supply competition. Additionally, HBM requires roughly three times the wafer capacity of conventional DRAM. Manufacturing must also occur in massive clean rooms, which require years to construct, compounding the supply challenge.

For Nvidia, this partnership strengthens its already dominant position in AI infrastructure. The company controls the training market, having optimized its CUDA software platform and GPUs that became the foundation for foundational AI models. By securing substantial HBM supply, Nvidia secures its leadership as the inference market expands. Its recent "acquisition" of Groq and its language processing units further advantages Nvidia: its GPUs handle the computationally intensive prefill phase, while Groq's SRAM-based LPUs, with memory built directly onto the chips, handle the decode phase, reducing latency. In a severely capacity-constrained HBM market, combining large HBM access with SRAM-based alternatives provides a decisive supply-chain advantage. Nvidia trades at a forward P/E of just over 15 times analyst estimates for fiscal 2028.

For SK Hynix, the deal represents a major windfall. As Nvidia's largest HBM supplier, the partnership cements its status and provides substantial revenue visibility. The agreement supports SK Hynix's multiyear capital expenditure expansion while tying the company to one of the largest and fastest-growing technology companies, reducing execution risk. With other chipmakers scrambling to secure HBM supply—Broadcom has already secured commitments from Samsung—overall DRAM capacity will remain tight and prices elevated for years. SK Hynix trades at a forward P/E under 6, suggesting meaningful room for multiple expansion as deal visibility increases.