Samsung combines two 450-layer chips into 900-layer V-NAND

Samsung is using advanced wafer bonding to combine two 450-layer V-NAND modules into a single 900-layer chip. The approach is designed to push the company toward 1000-layer NAND Flash by 2030.

Samsung is advancing its NAND Flash roadmap by stacking two 450-layer V-NAND modules into a single 900-layer V-NAND chip. The company is pursuing this approach because manufacturing complexity makes it impractical to reach that density with a single chip. The design marks a packaging-led step toward Samsung’s goal of creating 1000-layer NAND Flash storage modules.

Samsung is using Cell Multi-Bonding (CMB) technology, a form of wafer stacking based on hybrid bonding. The process creates a permanent bond between two silicon chips through embedded metal bumps that are fused together to form a single chip. Multiple dies, or in Samsung’s implementation, multiple whole silicon wafers, can be connected by stacking two chip back ends together. Samsung’s proprietary CMB process paves the way for 1000-layer V-NAND chips by 2030.

Early last year, Samsung introduced its 10th-generation V-NAND technology, enabling over 400 layers on a single module. It also brought in hybrid bonding for the first time, suggesting the method has since matured. Moving to bonded 450-layer designs created manufacturing challenges because thick wafers are prone to warping.

To address those issues, Samsung developed microscopic chucks and new bonding techniques aimed at correcting overlay problems and misalignment during wafer bonding. Inside the NAND structure, the company also created new bitline and wordline designs to manage power consumption and keep individual chip sizes within a reasonable range. Together, those changes support Samsung’s effort to scale V-NAND beyond conventional single-die limits.

52

Impact Score

NVIDIA Vera CPU leads early server benchmarks

NVIDIA’s pre-release Vera server CPU posted strong early results against Intel Xeon and AMD EPYC in a limited set of data center tests. The Arm-based design pairs custom cores, high memory bandwidth, and native FP8 support aimed at demanding compute and Artificial Intelligence workloads.

NVIDIA Vera CPU targets agentic Artificial Intelligence workloads

NVIDIA positions Vera as a CPU built for agentic Artificial Intelligence factories, with custom Olympus cores, high memory bandwidth and strong performance under sustained load. Early benchmark results from Phoronix and additional testing highlighted gains in throughput, memory efficiency and competitiveness against Intel and AMD processors.

Entry-level work faces an Artificial Intelligence squeeze

Employment has remained broadly stable, but early-career hiring in Artificial Intelligence-exposed occupations is showing signs of strain. The pressure is raising concerns about how young workers will gain skills, experience, and a path into long-term careers.

Labor data challenges the Artificial Intelligence jobs panic

Fears of an immediate white-collar jobs collapse from Artificial Intelligence are not yet reflected in broad US labor data. Early evidence points instead to a more uneven transition, with younger workers in some exposed fields facing the greatest pressure.

Contact Us

Got questions? Use the form to contact us.

Contact Form

Clicking next sends a verification code to your email. After verifying, you can enter your message.