The AI revolution just entered a new dimension—literally.
We are excited to announce that Broadcom has begun shipping the industry’s first 2nm custom compute SoC (system on a chip), built on our groundbreaking 3.5D eXtreme Dimension System in Package (XDSiP) platform.
As AI clusters scale toward gigawatt-level demands, the “old way” of building chips isn’t enough. Our 3.5D F2F (Face-to-Face) technology changes the game by combining 2.5D techniques with 3D-IC integration. This modular approach allows compute, memory, and network I/O to scale independently, delivering a trifecta of performance:
đź”· Unparalleled Signal Density: Maximizing throughput in a compact form factor.
đź”· Superior Power Efficiency: Unlocking high-efficiency computing to meet massive AI demands.
đź”· Ultra-Low Latency: Essential for the next generation of XPUs.
The AI revolution requires more than just innovation; it requires execution at scale that Broadcom consistently delivers.
Read the full announcement here:
Broadcom Ships 3.5D Face-to-Face Compute SoC…
Industry’s first 3.5D Face-to-Face (F2F) technology enables large-scale integration of XPUs delivering breakthrough performance, efficiency and scalability for AI clusters