The news
On July 25, 2026, Samsung Electronics (005930.KS) and Broadcom (AVGO) announced a memorandum of understanding (MOU) on memory, 2-nanometer chipmaking and packaging. The Samsung Broadcom deal is an MOU, not a purchase order; the companies estimated the collaboration at more than $200 billion through 2030.
Samsung said the MOU was announced at the AI Summit at The Midway in San Francisco. Attendees included Jinman Han, head of Samsung's foundry business, Broadcom president and CEO Hock Tan, and representatives of the Korean government. Fortune, publishing a Bloomberg report, said South Korean President Lee Jae Myung was in Silicon Valley for an AI summit, with several tech deals announced during the trip.
On memory, the companies plan to collaborate on supplying high-bandwidth memory (HBM), the stacked DRAM that sits beside AI processors, for Broadcom's next-generation AI accelerators. On foundry, meaning contract chip manufacturing, the work centers on Samsung's 2nm and smaller process technologies for Broadcom products, including its wireless broadband communications chips. Samsung said the work is expected to extend to 2.3D and 2.5D advanced packaging built on its 2nm process, aimed at faster, more power-efficient AI and networking silicon.
Young Hyun Jun, vice chairman and CEO of Samsung's Device Solutions division, pointed to "unprecedented demand for tightly integrated semiconductor technologies" driven by AI. Charlie Kawwas, president of Broadcom's Semiconductor Solutions Group, framed the pact as a way to keep pace as AI infrastructure scales.
Fortune's report described it as a contract worth more than $200 billion. Samsung's own statement described an MOU whose value the companies estimate, and it disclosed no committed volumes or pricing.
CNBC reported that long-term commitments from Broadcom, one of the leading custom AI chip designers, could lift utilization at Samsung's advanced factories as it tries to narrow the gap with industry leader Taiwan Semiconductor Manufacturing Co. (TSM). CNBC also noted that Samsung won a $16.5 billion contract in 2025 to make logic chips for Tesla (TSLA).
The numbers
- Estimated value of collaboration (companies' estimate)
- More than $200 billion
- Term
- Five years, through 2030
- Foundry process nodes covered
- 2nm and below
- Advanced packaging types
- 2.3D and 2.5D
- Samsung's 2025 Tesla chip contract (CNBC)
- $16.5 billion
- Broadcom fiscal Q3 AI semiconductor revenue (reported Sept. 2)
- $16.7 billion
Why CEOs should care
For buyers of AI infrastructure, memory supply is a practical limit on how many accelerators get built. A memory supply arrangement between Samsung and one of the leading custom AI chip designers bears directly on how quickly Broadcom-designed accelerators reach customers. If your cloud or AI provider's roadmap relies on custom silicon, ask about delivery timelines and whether memory for 2027 capacity is already secured. When reserving AI compute, favor contracts with capacity guarantees over best-effort availability.
CFOs and investors should read the $200 billion as an estimate of intent, not booked revenue. The figures that matter will show up in definitive supply agreements and in each company's quarterly results, and neither company disclosed volumes or prices. Budget owners should also note that Broadcom chose to line up memory and manufacturing partners through 2030, a sign of how tight leading-edge supply is expected to stay.
For boards overseeing supply-chain risk, the pact matters because it strengthens a second source at the leading edge. Samsung is trying to narrow the gap with TSMC, the industry leader, CNBC noted. Ask your hardware vendors which foundry and memory suppliers their products depend on, whether any are single-sourced, and what switching would take if a supplier or region were disrupted.
The bigger picture
The pact reflects a shift CNBC described: large technology companies increasingly design their own AI accelerators instead of relying only on general-purpose graphics processors. That raises demand for design partners such as Broadcom and for manufacturers that can build, package and supply memory for those chips. Samsung's pitch is breadth, since it makes memory, runs a foundry and offers packaging, which it said positions it to provide integrated solutions.
The deal also carries national weight. Fortune, citing Bloomberg, reported that South Korea aims to double its memory production capacity within five years, as Samsung and SK Hynix push for AI chip orders against rivals including TSMC.
What happened next
On July 30, 2026, Samsung reported that its foundry business's earnings improved significantly before incentive-related provisions, driven by HBM base-die demand and strong orders from US customers, with 2nm high-performance computing engagements among its design wins. It targeted double-digit foundry revenue growth in the second half. On August 3, TrendForce, citing Korean outlet Chosun Biz, reported that Samsung's 2nm yields were estimated at about 60%, against roughly 70% needed to win large-scale production contracts.
On September 2, 2026, Broadcom reported fiscal third-quarter AI semiconductor revenue of $16.7 billion, up 221% from a year earlier, and forecast $21.7 billion for the fourth quarter; its release did not mention Samsung. The next milestone to watch is whether the MOU turns into definitive agreements with committed volumes.
