Key Takeaways
- Broadcom has announced plans to deliver no fewer than 1 million units of 3D stacked chips by 2027, a milestone expected to drive substantial revenue growth.
- The innovative design layers two silicon components vertically, accelerating data transmission while reducing power consumption by as much as 10 times.
- Fujitsu has become the inaugural customer, currently developing engineering samples with full-scale manufacturing scheduled for late 2026.
- TSMC handles the production using its advanced 2-nanometer technology combined with 5-nanometer architecture.
- The company intends to launch two additional stacking-enabled products during the latter half of 2026, with three more prototypes coming in 2027.
Broadcom ($AVGO) has established an ambitious objective to deliver no less than 1 million units of its advanced 3D stacking chip technology by 2027, as revealed in an exclusive Reuters report released on February 26, 2026.
Harish Bharadwaj, who serves as Broadcom’s vice president of product marketing, disclosed these figures during direct conversations with Reuters.
This announcement represents a tangible commercial milestone for a technology platform that required approximately five years of dedicated research and development.
The innovative architecture operates by positioning two semiconductor chips in a vertical stack. This configuration enables significantly faster data exchange between components while dramatically reducing power requirements.
According to Bharadwaj, the technology achieves approximately 10 times superior energy efficiency versus conventional chip arrangements — a critical advantage as artificial intelligence computing demands continue expanding.
“Now, pretty much all of our customers are adopting this technology,” Bharadwaj told Reuters.
Fujitsu has emerged as the inaugural partner deploying this architecture. The Japanese technology giant is presently manufacturing engineering prototypes and anticipates transitioning to volume production before 2026 concludes.
TSMC serves as the manufacturing partner for Fujitsu’s chip, leveraging its cutting-edge 2-nanometer fabrication process integrated with 5-nanometer technology. Partners retain flexibility to select different TSMC process nodes based on specific application requirements.
Expanded Product Portfolio Coming
Broadcom’s strategy extends well beyond the Fujitsu collaboration. The semiconductor company anticipates delivering two more products utilizing the stacking architecture during the second half of 2026.
An additional three designs are slated for prototype sampling throughout 2027. Development teams are simultaneously advancing toward configurations featuring up to eight stacked pairs — potentially creating packages with 16 distinct silicon layers.
The 1 million unit forecast encompasses the complete portfolio of these designs, rather than representing a single product.
Broadcom’s approach within the AI semiconductor sector centers on collaborative custom chip development with major technology corporations. Google relies on Broadcom for designing its Tensor Processing Units (TPUs). OpenAI has similarly established a partnership with Broadcom for proprietary processor development.
Artificial Intelligence Revenue Acceleration
Broadcom forecasted its AI semiconductor revenue would hit $8.2 billion during its initial fiscal quarter — representing approximately 100% growth compared to the equivalent period in the previous year.
This expansion stems predominantly from bespoke chip agreements with hyperscale cloud providers, where Broadcom converts conceptual designs into production-ready chip architectures that TSMC subsequently manufactures.
The 3D stacking capability introduces both a technical and commercial dimension to this established business model.
Broadcom’s AI chip revenue was projected at $8.2 billion for Q1 fiscal 2026, doubling year-over-year.
