Broadcom targets 1 million 3D stacked chips by 2027 as it challenges Nvidia's AI dominance

3 Sources

Share

Broadcom is shipping a new custom AI chip design to Fujitsu featuring innovative 3.5D stacking technology that promises better power efficiency. The company expects to sell at least 1 million 3D stacked chips by 2027, potentially generating billions in revenue as it positions itself as a formidable competitor to Nvidia in the AI chip market.

Broadcom Ships Innovative AI Chip to Fujitsu

Broadcom has begun shipping a groundbreaking custom AI chip design to Fujitsu that uses a novel stacking approach to deliver superior power efficiency and performance

1

. The company's 3.5D eXtreme Dimension System in Package takes two chip dies and stacks them top to top, rather than the traditional top-to-bottom method pioneered by Advanced Micro Devices

1

. This configuration allows more data to be transferred with better energy use, addressing critical challenges in meeting escalating AI software demands. Fujitsu is currently making engineering samples to test the design and plans to produce the 3D stacked chips later this year for data centers and potentially supercomputers in the future

2

.

Source: Seeking Alpha

Source: Seeking Alpha

Ambitious Sales Target Signals Major Revenue Stream

Broadcom expects to sell at least 1 million chips by 2027 based on its chip stacking technology, according to Harish Bharadwaj, vice president of product marketing

2

. This forecast marks a significant product and sales target that could represent a revenue stream potentially worth billions of dollars

2

. The company's AI chip revenue is already projected to double year-over-year to $8.2 billion in its first fiscal quarter

2

. The million-chip figure includes several designs beyond the Fujitsu chip, with Broadcom having half a dozen of these designs in development for various customers

1

. The company expects to ship two more products based on the stacking technology in the second half of this year and to sample an additional three in 2027

2

.

Five-Year Development Tackles Complex Technical Challenges

Broadcom has worked for approximately five years with Taiwan Semiconductor Manufacturing Co to refine this technology to commercial viability

1

. The development process required engineers to overcome significant challenges related to delivering power to all parts of the chip and preventing overheating

1

. The stacking approach gives customers the ability to build chips with more horsepower while using less energy

2

. For the Fujitsu chip, Taiwan Semiconductor Manufacturing Co is fabricating the chip using its cutting-edge 2-nanometer process and fusing it with a 5-nanometer chip

2

. Companies can mix and match which fabrication process TSMC uses with the Broadcom technology, with TSMC fusing the top and bottom chip during the manufacturing process

2

. Engineers are now working to create chips with as many as eight stacks of two chips each

2

.

Source: ET

Source: ET

Hyperscalers Adopt Technology as Broadcom Challenges Nvidia

The next phase will involve hyperscalers—the largest data center operators—adopting the technology, with Bharadwaj stating that "across all our existing hyperscaler customers -- every one of them has adopted this for their next generation"

1

. Nearly all of Broadcom's customers are now adopting the stacking technology

3

. Broadcom works with major players including Google for its tensor processing units and OpenAI for its in-house custom processors, helping translate early designs into physical chip layouts that can be fabricated

2

. The company has emerged as a key competitor to Nvidia and a significant rival in the race to produce silicon that meets the rapidly growing computing requirements AI presents

1

2

. A range of customers—from hardware sellers to AI labs to chip startups—are seeking designs that can deliver greater performance and efficiency, which helped lift Broadcom shares 49% last year

1

.

Today's Top Stories

TheOutpost.ai

Your Daily Dose of Curated AI News

Don’t drown in AI news. We cut through the noise - filtering, ranking and summarizing the most important AI news, breakthroughs and research daily. Spend less time searching for the latest in AI and get straight to action.

© 2026 Triveous Technologies Private Limited
Instagram logo
LinkedIn logo