Samsung, Broadcom ink $200‑billion AI chip pact while world faces memory crunch
Samsung Electronics’ latest tie‑up with US chip giant Broadcom underlines how the global memory crunch is reshaping corporate alliances as much as it is squeezing device makers and data‑centre budgets, including in India
This strategic partnership, valued at over $200 billion by 2030, will see Samsung supply industry-leading memory like HBM for Broadcom's AI accelerators and manufacture Broadcom chips on 2-nanometre and smaller nodes. The alliance also encompasses advanced packaging techniques such as 2.3D and 2.5D integration to boost performance and energy efficiency in AI and networking silicon.
This strategic partnership, valued at over $200 billion by 2030, will see Samsung supply industry-leading memory like HBM for Broadcom's AI accelerators and manufacture Broadcom chips on 2-nanometre and smaller nodes. The alliance also encompasses advanced packaging techniques such as 2.3D and 2.5D integration to boost performance and energy efficiency in AI and networking silicon.
This strategic partnership, valued at over $200 billion by 2030, will see Samsung supply industry-leading memory like HBM for Broadcom's AI accelerators and manufacture Broadcom chips on 2-nanometre and smaller nodes. The alliance also encompasses advanced packaging techniques such as 2.3D and 2.5D integration to boost performance and energy efficiency in AI and networking silicon.
Samsung and Broadcom have signed a memorandum of understanding in San Francisco to deepen their partnership across high‑bandwidth memory (HBM), leading‑edge foundry processes and advanced packaging for next‑generation AI infrastructure.
The Korean government has framed the deal as part of a broader, roughly $950‑billion semiconductor cooperation push with global tech firms, with the Samsung–Broadcom collaboration alone estimated at more than $200 billion over five years through 2030.
Under the MOU, Samsung said it would supply “industry‑leading” memory solutions such as HBM for Broadcom’s future AI accelerators, while its foundry arm would manufacture Broadcom chips on 2‑nanometre and smaller nodes, including wireless broadband communication products.
The partnership also covers advanced 2nm‑class packaging, including 2.3D and 2.5D integration, aimed at squeezing more performance and energy efficiency out of AI and networking silicon by stacking memory closer to logic.
The announcement comes during a severe global memory shortage, driven by AI data‑centre demand and a shift in wafer capacity towards premium products like HBM.
Industry trackers such as TrendForce report that contract DRAM prices have risen 90–95 per cent quarter‑on‑quarter in early 2026, with NAND flash up more than 50 per cent, and warn that supply will remain tight for at least the next 12–18 months.
Samsung and SK hynix themselves have cautioned that AI‑driven shortages across DRAM and HBM could persist into the next decade.
For India, which is trying to build out AI data centres and expand domestic electronics manufacturing, the move cuts both ways.
On one hand, long‑term HBM and foundry commitments between Samsung and major AI chip designers help stabilise global supply chains, reducing the risk of abrupt shortages that can derail cloud rollouts and server procurement.
On the other, analysis by research outfits covering India’s datacentre and smartphone markets suggests that memory now accounts for 10–15 per cent of server costs, with DRAM and advanced memory price spikes already forcing 10–20 per cent hikes in device prices and pushing up capex for AI infrastructure.
This could mean that hyperscale buyers tied into multi‑year supply contracts are likely to be served first, while smaller Indian cloud providers, OEMs and start‑ups might still face longer lead times and higher bills even as headline deals like Samsung–Broadcom expand global capacity.