Samsung and Broadcom Sign a $200 Billion AI Chip Cooperation Memorandum

On July 25, Samsung Electronics and Broadcom signed a memorandum of cooperation during the AI Summit in San Francisco. The two companies expect the collaboration to exceed $200 billion over the next five years through 2030, covering high-bandwidth memory, foundry manufacturing, and advanced packaging. For the expanding AI industry, this is not just a supplier switch, but an infrastructure deal that puts multiple stages of chip manufacturing under a long-term plan at the same time.

What happened

According to a Samsung Electronics announcement, the two sides will advance next-generation memory solutions, including HBM, to support Broadcom's next-generation AI accelerators. In foundry manufacturing, the collaboration will involve Samsung's 2-nanometer and more advanced process nodes to provide manufacturing capacity for Broadcom's related products. The two companies also plan to advance advanced packaging based on 2-nanometer processes through 2.3D and 2.5D integration to improve the performance and energy efficiency of AI and networking chips.

It should be noted that the legal nature of the publicly disclosed information is still a memorandum of understanding, not a final procurement contract with all commercial terms disclosed. Samsung used the wording that it expects the scale to exceed $200 billion, while the specific order pace, capacity allocation, product mix, and revenue recognition methods have not yet been fully disclosed.

Why it matters

The bottleneck for AI chips is no longer just the GPU itself. Accelerators need high-bandwidth memory to provide data throughput, advanced process nodes determine power consumption and integration density, and packaging affects communication efficiency between different chips. By putting these three parts into the same long-term cooperation framework, Samsung and Broadcom show that supply chain competition is shifting from "who can produce a faster chip" to "who can reliably deliver a complete computing system."

For Samsung, the collaboration serves two strategic lines at once: on the one hand, expanding HBM's share in AI accelerators; on the other, winning more advanced foundry orders. Reuters reported that the deal is seen as an important opportunity for Samsung to improve foundry utilization and catch up with TSMC. For Broadcom, stable memory, manufacturing, and packaging resources will help support its expansion in the custom AI accelerator and networking chip markets.

This will also affect public market valuation logic. In the past, investors mainly priced companies based on a single category such as chip design, memory, or foundry. Now, long-term cooperation may make order visibility, capacity lock-in, and packaging capability equally important valuation variables. But a memorandum does not equal realized sales, and the market will ultimately still need to wait for verification from mass production, yields, and customer delivery data.

What still needs watching

First, when the two sides will sign a binding final agreement, and how much of the "more than $200 billion" is already committed orders. Second, whether Samsung's 2-nanometer process mass production progress, yields, and advanced packaging capacity can meet Broadcom's delivery requirements. Third, whether HBM supply will continue to be the main constraint on AI accelerator expansion. Fourth, whether Broadcom's custom chip business can turn manufacturing cooperation into sustained actual revenue rather than remaining at the level of capex and supply chain expectations.

The true significance of this deal may only be fully priced by the market once the supply chain moves from the announcement to the fab, and then into customers' data centers.

Sources

• Samsung Electronics: Expanding strategic collaboration with Broadcom across memory and foundry technologies
• Reuters: Samsung reaches $200 billion AI chip partnership with Broadcom, boosting foundry push
• Reuters: South Korea announces $950 billion AI plans involving Samsung, SK Group and U.S. tech companies