Samsung Electronics won a contract worth more than $200 billion to supply chips to Broadcom in the largest publicly known foundry supply agreement of the AI buildout era.
The Broadcom Chip Deal covers 2-nanometer AI accelerator chips and advanced packaging through 2030, and it arrives as both companies compete for a larger share of AI semiconductor revenue.
Samsung Broadcom Chip Deal Locks In A Five-Year AI Supply Chain
The Broadcom Chip Deal spans multiple product categories.
Samsung will manufacture 2nm logic chips, supply high-bandwidth memory, and provide advanced packaging services. The contract runs through 2030 and is structured to scale with Broadcom's AI custom silicon business.
Fortune reported on July 25 that the agreement represents the largest publicly known foundry supply contract of the current AI era.
Broadcom's core AI product is a custom accelerator chip, called an XPU, designed for hyperscale customers who want purpose-built silicon rather than general-purpose GPUs.
Hyperscalers including Google and Meta have used Broadcom to develop proprietary AI chips tailored to their own workloads.
The 2nm process node Samsung will use is among the most advanced in commercial production, placing transistors closer together to pack more computing power into a smaller die while lowering energy consumption per operation.
The contract size, $200 billion over five years, implies roughly $40 billion per year. That figure would represent a substantial share of Samsung's semiconductor revenue and signals that Broadcom's custom AI chip business is scaling faster than most public estimates had suggested.
The sheer scale of the Broadcom Chip Deal underscores how aggressively hyperscale customers are investing in purpose-built silicon.
Why The Broadcom Chip Deal Changes AI Supply Chain Dynamics
The agreement matters for a structural reason that goes beyond the two companies involved. Nvidia has dominated the GPU market for AI training, but custom silicon from Broadcom and rivals has grown sharply as hyperscalers seek to reduce their dependence on any single vendor and optimize chips for inference rather than training.
Inference is the process of running a trained AI model to generate outputs, the step that occurs billions of times per day across search, chatbots, and recommendation engines. Inference workloads favor chips tuned to specific architectures, which is where custom XPUs compete directly with Nvidia's off-the-shelf products.
Also Read: Korea’s 200MW AI Factory Hides a Stunning $10 Billion Financing Play
Samsung, for its part, has trailed Taiwan Semiconductor Manufacturing Company in the race for advanced foundry contracts.
TSMC manufactures the bulk of Nvidia's GPUs and handles most of Apple's chip production. The Broadcom Chip Deal gives Samsung a major anchor customer for its 2nm node and validates its process technology at a moment when it has faced yield challenges on previous advanced nodes.
How Samsung Staged Its Comeback In Advanced Chip Manufacturing
Samsung's foundry business lost ground to TSMC through the early 2020s, as customers prioritizing yield and performance concentrated orders with the Taiwanese manufacturer.
TSMC's 2026 US investment commitment of $265 billion underlined its dominance and its confidence in sustained AI chip demand.
Samsung responded by investing heavily in its 3nm and 2nm gate-all-around transistor process, a design that wraps the gate electrode on four sides of the transistor channel rather than three, improving current control and reducing leakage. The technology is central to achieving competitive performance at 2nm and is the same architecture Samsung will use to fulfill the Broadcom Chip Deal.
Winning a five-year agreement worth $200 billion gives Samsung's foundry division the kind of committed volume that justifies further capital expenditure on 2nm capacity.
It also reduces the business risk of ramping an expensive new node without guaranteed customers.
What Comes Next For AI Chip Supply
The Broadcom Chip Deal does not threaten Nvidia's near-term position. Nvidia's GB200 and next-generation Rubin GPUs are manufactured at TSMC and remain the default choice for AI training at scale.
However, the custom silicon market, where Broadcom competes, is growing as a share of total AI semiconductor spending.
If Samsung can demonstrate consistent yield at 2nm over the first production runs, it creates competitive pressure on TSMC's pricing and could attract additional hyperscale customers seeking supply chain diversification. The next disclosure to watch is Samsung's foundry utilization rate in its quarterly results, which will show how quickly the Broadcom volume ramps.
Read Next: Can Bitcoin ETFs Reclaim The Flow Crown? Ethereum Has Won Three Weeks Running