Broadcom’s AI Semiconductor Revenue Surge Reshapes Foundry Landscape
Broadcom’s AI semiconductor business is exploding, a surge that is fundamentally reshaping the global foundry landscape. The numbers speak for themselves. In the fourth quarter of fiscal year 2025, Broadcom’s AI chip revenue rocketed 74% year-over-year to $6.5 billion. Projections for the first quarter of fiscal year 2026 show this momentum accelerating, with revenue expected to double year-over-year to $8.2 billion. This puts Broadcom on a clear path to hit $20 billion in AI revenue for fiscal year 2025, with an ambitious target of over $100 billion annually by 2027. A staggering $73 billion AI backlog underpins this growth, which the company aims to convert into revenue at a rate exceeding $12 billion per quarter over the next 18 months. To fuel this expansion and secure chip supply for key clients like Anthropic and OpenAI, Broadcom is preparing a massive debt raise, potentially between $60 billion and $100 billion.
TSMC CoWoS Bottleneck Creates Strategic Opportunity for Samsung Foundry
At the heart of the current AI chip supply chain crunch is a critical bottleneck: TSMC’s advanced CoWoS (Chip-on-Wafer-on-Substrate) packaging technology. This process is indispensable for fusing high-performance AI processors with High Bandwidth Memory (HBM). While TSMC is aggressively expanding its CoWoS capacity—projected to nearly quadruple from approximately 35,000 wafers per month at the end of 2024 to 125,000-130,000 by late 2026—it’s simply not enough. TSMC’s CEO, C.C. Wei, has been blunt, stating capacity remains “extremely tight and sold out through 2026.” The global demand picture confirms this strain, with CoWoS demand expected to double from 1.3-1.4 million wafers in 2026 to 2.5-2.7 million in 2027, creating an anticipated supply shortfall of 10-20%. With Nvidia already locking up an estimated 60% of TSMC’s 2026 CoWoS allocation, other heavyweights like Broadcom have no choice but to diversify their supply chains.
Samsung Foundry Secures $200 Billion Collaboration with 2nm and Advanced Packaging
TSMC’s production ceiling has opened a golden door for Samsung Foundry. Samsung Electronics and Broadcom have now cemented a strategic partnership valued at over $200 billion through 2030. This sweeping collaboration covers AI chips, 2nm foundry manufacturing, advanced packaging, and HBM memory. Specifically, the deal targets Samsung’s 2-nanometer and more advanced process nodes for Broadcom’s next-generation products, incorporating sophisticated packaging solutions like 2.3D and 2.5D integration. Broadcom will also tap Samsung for HBM in future AI accelerators. To meet this demand, Samsung Foundry is fast-tracking the mass production of its 2nm process, which leverages its proprietary Gate-All-Around (GAA) technology, and is also ramping up production of 4nm Language Processing Unit chips and HBM base dies for external clients like Qualcomm and its own HBM. Samsung’s integrated model—combining memory, logic, foundry, and advanced packaging under one roof—gives it a powerful, unique advantage in delivering the turnkey semiconductor solutions required for the AI era. As a result, AI and high-performance computing are poised to account for over 30% of its total foundry revenue.
Outlook and Investment Implications
The combination of Broadcom’s insatiable demand for AI chips and TSMC’s packaging constraints is redrawing the map of the foundry industry. Samsung Foundry’s landmark, long-term deal with Broadcom is a strategic masterstroke, cementing its role in the critical 2nm-and-beyond process and packaging ecosystem. For investors, the key will be to watch Samsung’s execution. Its ability to stabilize yields for the new GAA-based 2nm processes and rapidly scale its own advanced packaging capacity, akin to CoWoS, will be paramount. The exponential growth of the AI semiconductor market isn’t slowing down, creating ample room for multiple dominant foundries. Samsung’s technology roadmap and its aggressive capacity expansion are the crucial factors that will determine its standing as a pillar of the AI supply chain.
References & Sources
