Samsung's Hybrid Bonding technology is expected to be used in NVIDIA's next-generation AI accelerator, Feynman, with the chip featuring custom HBM.

BlockBeats News, July 21st - Samsung Electronics has started construction of a Die-to-Wafer (D2W) hybrid bonding production line at its Pyeongtaek P5 plant, aiming for advanced packaging of the next-generation HBM and logic semiconductor. The core equipment supplier of choice is Dutch company Besi. Samsung plans to purchase approximately 50 hybrid bonding machines, with each machine priced at around 60 billion Korean Won (approximately $43 million), twice the price of competing products. However, due to Samsung's request for customized modifications to the equipment architecture, the negotiation process has been slow. Moreover, Besi supplies equipment to both TSMC and Micron, so they are cautious about customizing the equipment solely for Samsung.

Samsung is also considering local alternatives simultaneously: its subsidiary SEMES has completed the development of the D2W hybrid bonding machine and conducted sample verification earlier this year; Hanwha Semitech completed the development of the second-generation "SHB2 Nano" in February, sent samples to SK hynix in April, and its cluster system integrating EFEM, plasma activation, and cleaning modules has formed a differentiating competitive edge.

This technology uses hybrid copper bonding to replace traditional thermocompression bonding and can directly bond copper to dielectric material, significantly reducing interconnect length. The core application scenario is for custom HBM—stacking the HBM DRAM layer directly on top of a logic die containing customer computing circuits and IPs to form a 3D system-level package.

Analyst Jukan from Citrini stated that hybrid bonding is expected to be used in NVIDIA's next-generation AI accelerator, Feynman. This chip will utilize custom HBM, and the technology deployment timeline has been "pushed back from HBM4E." Internally, Samsung anticipates achieving widespread application around 2029 to 2030, aligning with the expected window for NVIDIA's Feynman.