Samsung Electronics and Broadcom sign $200bn memory and foundry MOU through 2030

Samsung Electronics Co., Ltd. (KRX: 005930; OTCMKTS: SSNLF) and Broadcom Inc. (NASDAQ: AVGO) have signed a memorandum of understanding to expand their collaboration across memory, foundry and advanced packaging technologies, with the two companies estimating the value of the combined activity at more than $200 billion over the five years to 2030. The MOU was announced at the AI Summit at The Midway in San Francisco on 25 July 2026, with Korean government representatives in attendance alongside Jinman Han, President and Head of Samsung Electronics’ Foundry Business, and Hock Tan, President and Chief Executive Officer of Broadcom. The pact commits the two companies to work together on High Bandwidth Memory supply for Broadcom’s next-generation AI accelerators, on Samsung’s 2-nanometer and sub-2nm foundry process for Broadcom’s Wireless Broadband Communications silicon, and on 2.3D and 2.5D advanced packaging integration. The immediate significance lies in the second front it opens against Taiwan Semiconductor Manufacturing Company Limited on leading-edge foundry and against SK Hynix Inc. on HBM supply to hyperscaler-facing accelerators. The unresolved question is how quickly the MOU converts into binding tape-out and volume-production commitments with dated revenue attached.

Why does the Samsung Electronics and Broadcom memorandum of understanding matter for the global AI infrastructure supply chain right now?

The pact is a memorandum of understanding, not a definitive supply agreement, and the $200 billion figure is described by both companies as an estimate of the collaboration’s combined scope through 2030 rather than a contracted purchase commitment. That distinction matters for how the market prices the announcement. What has changed is the direction of travel: Broadcom has publicly named Samsung Electronics as a strategic multi-technology partner for its custom accelerator roadmap, and Samsung has publicly named Broadcom as an anchor customer for its 2nm and sub-2nm foundry business alongside its HBM and advanced packaging offer. For an industry in which fab utilisation, HBM allocation and CoWoS-class packaging capacity have all functioned as pricing chokepoints through 2025 and into 2026, a second qualified two-way supplier-customer relationship at the leading edge is a structural signal, not a marketing one.

What exactly did Samsung Electronics and Broadcom agree to pursue across memory, foundry and advanced packaging?

The MOU covers three distinct workstreams. On memory, Samsung Electronics plans to supply Broadcom with High Bandwidth Memory, including current-generation HBM and the roadmap towards HBM4 and HBM4E, to support Broadcom’s next-generation AI accelerators used by hyperscaler custom silicon customers. On foundry, the collaboration focuses on Samsung’s 2nm and below process technologies for Broadcom products, with Wireless Broadband Communications solutions identified as the first named workload category. On advanced packaging, Samsung will make available its 2.3D and 2.5D integration options built on the 2nm process node, allowing Broadcom to co-locate logic, memory and interconnect within a single package for higher-performance and more power-efficient AI and networking silicon. The announcement did not disclose which specific Broadcom accelerator programme is being ported to Samsung 2nm, the volume commitments attached to any tape-out, or the pricing framework governing HBM allocation, all of which will need to be surfaced in subsequent filings or product launches for the MOU to be tested commercially.

How does the Broadcom deal change Samsung Foundry’s competitive position against Taiwan Semiconductor Manufacturing Company on 2nm?

Taiwan Semiconductor Manufacturing Company Limited has been the near-exclusive foundry partner for Broadcom’s leading-edge custom accelerator work, and the Broadcom account is one of the largest external revenue streams for TSMC’s advanced-node business alongside NVIDIA Corporation, Apple Inc. and Advanced Micro Devices, Inc. Samsung Foundry, under Jinman Han, has spent the past two years attempting to convert its 2nm gate-all-around process from a technology demonstration into an anchor-customer platform. Landing Broadcom on 2nm, even for a subset of workloads such as Wireless Broadband Communications silicon initially, gives Samsung Foundry the reference customer it has lacked at the node. It also gives Broadcom an insurance policy against single-source foundry risk at a moment when its custom accelerator revenue with hyperscaler clients is growing at pace and when geopolitical exposure around Taiwan remains an unresolved input into supply-chain planning. The commercial test will be tape-out cadence and yield, neither of which the MOU discloses.

What does the memorandum mean for Samsung’s High Bandwidth Memory business relative to SK Hynix and Micron Technology?

Samsung Electronics has historically been the largest DRAM producer globally, but on High Bandwidth Memory it has trailed SK Hynix Inc. through the HBM3 and HBM3E generations, particularly in supply to NVIDIA’s Hopper and Blackwell accelerator lines. Micron Technology, Inc. has taken share in the same window. Samsung began shipment of industry-first HBM4E samples on 29 May 2026 and unveiled its comprehensive HBM roadmap at NVIDIA GTC in March 2026, both of which reset the narrative on Samsung’s HBM4 and HBM4E qualification. Naming Broadcom, whose custom accelerator business is the second-largest AI silicon franchise outside NVIDIA by revenue disclosed in AI networking and XPU commentary, gives Samsung a second flagship HBM customer alongside its existing hyperscaler engagements. The pact does not disclose the volume split between Samsung, SK Hynix and Micron Technology within Broadcom’s next-generation accelerator platforms, and Broadcom has typically qualified multiple HBM sources to protect against allocation risk, so this is best read as Samsung securing a place at the table rather than displacing incumbents outright.

How should investors interpret the $200 billion five-year estimate given that the deal is a non-binding memorandum of understanding?

The $200 billion figure is presented by Samsung Electronics and Broadcom as an estimate of combined collaboration value across memory and foundry over the five years to 2030. It is not a purchase order, not a firm commitment and not a contracted minimum. In the context of Broadcom’s fiscal 2025 semiconductor solutions revenue of roughly $30 billion and Samsung’s semiconductor division revenue running above KRW 100 trillion on an annualised basis at recent run-rate, the figure implies a directional target rather than a booked backlog. Investors will need three subsequent proof points to validate the estimate: a first named tape-out on Samsung 2nm with a Broadcom product code attached, a disclosed HBM supply agreement with volume and pricing, and a joint packaging engagement on 2.3D or 2.5D integration that ships to a hyperscaler customer. Absent those, the MOU functions as strategic positioning rather than as revenue. Samsung’s full second-quarter 2026 earnings on 30 July 2026 provide the first opportunity for management to add colour on foundry customer engagements and HBM allocation, though no binding disclosure is expected on the Broadcom MOU itself at that stage.

What is the timing pressure on Samsung Foundry to convert the Broadcom relationship into commercial revenue before 2027?

Samsung Foundry’s 2nm process is targeted for high-volume manufacturing over the 2026 to 2027 window, with sub-2nm nodes further out. Broadcom’s next-generation AI accelerator platforms for the largest hyperscaler custom silicon customers are on roughly a 12- to 18-month refresh cadence, and design-in decisions for accelerators shipping in 2028 are being made in 2026. That means the Broadcom MOU has a narrow window in which to convert into tape-out and mask commitments if Samsung is to appear in the 2028 accelerator generation rather than the 2029 one. The strategic case for Broadcom in signing the MOU is precisely to preserve that optionality. The strategic case for Samsung is that missing the 2028 design window would push the first material Broadcom-attributable foundry revenue into fiscal 2029, which is the tail of the five-year estimate rather than the middle. Investors reading the announcement should track subsequent product-level disclosures more closely than headline figures.

How does the Samsung Electronics and Broadcom pact fit into the wider United States and South Korea semiconductor policy backdrop?

The presence of Korean government representatives at the signing, and the timing of the announcement during President Lee Jae Myung’s US visit and the parallel expectation of supply-deal announcements between Korean chipmakers and US technology companies, position the MOU within a broader bilateral semiconductor framework rather than as a purely commercial transaction. Samsung Electronics and SK Hynix Inc. together account for a materially larger share of the Korea Composite Stock Price Index than they did five years ago, and their strategic alignment with United States hyperscaler and fabless customers has become a policy variable for both governments. For Broadcom, the pact aligns with its parallel expansion of United States semiconductor facility investment announced earlier in July 2026, including the $30 billion chip deal referenced in the Apple Inc. context. For Samsung Electronics, the pact reinforces the case for its United States foundry footprint, including the Taylor, Texas facility, as a manufacturing option for Broadcom-designed silicon.

What does the current market reaction reveal about how investors are pricing the Samsung and Broadcom collaboration?

Samsung Electronics shares closed at KRW 249,500 on 24 July 2026 on the Korea Exchange, down 7.6% on the session, with a 52-week range of KRW 65,500 to KRW 374,500 and a market capitalisation of approximately KRW 1,600 trillion. The stock has fallen roughly 21.5% since its 6 July 2026 pre-guidance close, reflecting profit-taking after a near-150% year-to-date rally and concerns about the sustainability of AI infrastructure capital expenditure at hyperscaler customers. Preliminary second-quarter revenue of KRW 171 trillion and operating margin of about 52% for the April to June quarter were flagged by analysts including those at Deutsche Bank AG as only 6% ahead of estimates, insufficient to sustain further re-rating in isolation. Broadcom shares closed at $381.92 on 24 July 2026 on the Nasdaq, down 2.69%, with a 52-week range of $281.61 to $495.00 and a market capitalisation of approximately $1.92 trillion. The 26-analyst consensus price target on Broadcom sits near $501.58 with a Buy skew. The MOU was announced after Broadcom’s 24 July session close and before Samsung’s next Seoul session, so the initial price impact will be visible from the 28 July trading windows in Seoul and New York rather than in the closing prices above.

What are the principal execution risks that could weaken the Samsung Electronics and Broadcom collaboration between now and 2030?

The pact carries several execution-dependent uncertainties. Samsung Foundry’s 2nm yield ramp remains to be demonstrated at commercial volumes with an external anchor customer of Broadcom’s complexity. HBM4 and HBM4E qualification at NVIDIA and AMD sets the pace at which Samsung’s memory division can absorb incremental Broadcom volumes without disrupting existing hyperscaler commitments. Advanced packaging capacity in 2.3D and 2.5D configurations is a global bottleneck, and Samsung’s ability to scale that capacity to Broadcom’s requirements alongside its own memory and logic customers is not yet publicly disclosed. Broadcom retains dual-sourcing flexibility and could reduce the pace at which Samsung-manufactured content ramps if TSMC yields hold or if hyperscaler customer preferences remain conservative. The $200 billion estimate assumes a smooth build-out through 2030, which is a strong assumption given the sensitivity of the AI infrastructure spend cycle to hyperscaler capital expenditure discipline, an area in which Morgan Stanley and other sell-side analysts have flagged tightening controls.

What are the key takeaways for investors weighing the Samsung Electronics and Broadcom collaboration against the AI silicon supply chain through 2030?

• Samsung Electronics Co., Ltd. and Broadcom Inc. signed a memorandum of understanding on 25 July 2026 in San Francisco covering HBM supply, 2nm and sub-2nm foundry, and 2.3D and 2.5D advanced packaging, with a combined collaboration value estimated by the two companies at more than $200 billion over the five years to 2030.
• The signatories present at the AI Summit were Jinman Han, President and Head of Samsung Electronics’ Foundry Business, and Hock Tan, President and CEO of Broadcom, with Korean government representatives in attendance during President Lee Jae Myung’s United States visit.
• The MOU names Broadcom as a strategic customer for Samsung Foundry’s 2nm process starting with Wireless Broadband Communications silicon, and names Samsung Electronics as a strategic supplier of HBM for Broadcom’s next-generation AI accelerators.
• The pact opens a second front against Taiwan Semiconductor Manufacturing Company Limited at the 2nm node and against SK Hynix Inc. on HBM allocation, giving both companies an insurance policy against single-source risk and diversifying customer and supplier concentration in the AI silicon supply chain.
• The $200 billion figure is a non-binding estimate of five-year collaboration scope, not a booked purchase commitment; investors should treat it as directional positioning until validated by tape-out disclosures, HBM volume commitments and shipped packaging engagements.
• The commercial test window is narrow: for Samsung Foundry to appear in the 2028 accelerator generation, tape-out commitments will need to land inside 2026 and 2027, ahead of the tail-end of the five-year MOU horizon.
• Samsung Electronics closed at KRW 249,500 on 24 July 2026, down 7.6% and down 21.5% from its 6 July pre-guidance close after a near-150% year-to-date rally; Broadcom closed at $381.92, down 2.69%, with consensus analyst target near $501.58.
• Full disclosure will require the first Broadcom product tape-out on Samsung 2nm, a named HBM supply agreement with volume and pricing terms, and a shipped 2.3D or 2.5D packaging engagement at a hyperscaler customer.
• Downside catalysts include Samsung 2nm yield disappointment, HBM allocation conflicts, hyperscaler capital expenditure discipline flagged by Morgan Stanley and others, and Broadcom retaining dual-sourcing optionality with Taiwan Semiconductor Manufacturing Company.
• Near-term investor checkpoints are Samsung Electronics’ full second-quarter 2026 earnings on 30 July 2026, Broadcom’s next quarterly update, and any subsequent joint product announcement naming a Broadcom silicon SKU on Samsung 2nm or in Samsung HBM configurations.