Broadcom Inc. has set an ambitious target to sell at least 1 million chips utilizing its 3D stacking technology by 2027, a move that could generate billions in revenue and significantly impact energy efficiency in compute-intensive sectors like cryptocurrency mining and AI. The forecast was shared exclusively with Reuters by Harish Bharadwaj, Broadcom’s vice president of product marketing, on February 26, 2026.
The core innovation lies in stacking two pieces of silicon directly on top of each other. This design allows for faster data transfer between chips while reducing energy consumption by up to 10 times compared to traditional layouts. "Now, pretty much all of our customers are adopting this technology," Bharadwaj stated, highlighting the demand for energy-efficient performance as AI and other high-performance computing workloads expand.
Fujitsu has been named as the first customer, currently producing engineering samples with plans for full production later in 2026. The fabrication is being handled by TSMC, which is using its advanced 2-nanometer process fused with a 5-nanometer chip. Broadcom's roadmap includes shipping two additional stacking-based products in the second half of 2026 and sampling three more designs in 2027. Engineers are even exploring stacks with up to 16 layers of silicon in a single package.
This development occurs against a backdrop of intense competition in the semiconductor space, as highlighted by a separate analysis of Nvidia's dominant Q4 2026 earnings. While Nvidia's strength in AI inference is pressuring rivals like AMD, Broadcom's unique position is more nuanced. The company reported a projected AI chip revenue of $8.2 billion for its fiscal Q1 2026, doubling year-over-year, largely driven by custom chip deals with hyperscalers like Google and OpenAI.
Analysts note that while Broadcom stock (AVGO) dipped following Nvidia's report, its business model is defensible. Unlike AMD's direct GPU competition, Broadcom focuses on the "plumbing" of AI—networking switches and custom accelerators (XPUs)—with a reported AI backlog of $162 billion. Oppenheimer recently reiterated an "Outperform" rating, citing Broadcom's crucial role in the AI infrastructure build-out.
