Memory Solutions for Gen AI Part 8: Custom HBM
Overview: When it comes to generative AI, disruption is not just inevitable, it is the engine of progress. Amid this dramatic shift, memory solutions are racing to catch up, spurred by the demand for enhanced bandwidth, capacity, area efficiency, power management, heat dissipation, and cost-effectiveness. But there is no one-size-fits-all here, and the answer is custom memory, a space set to expand significantly by 2026 when HBM4 is poised to debut. Currently, seven to eight major IT players such as NVIDIA, Amazon, Microsoft, Broadcom, and Marvell are driving custom HBMs.
Table of Contents:
- The architectural Crossroad
- Competitive landscape with IPs and Designs
- Can custom memory structurally change the volatile dynamics of memory?
Existing Subscriber?
Sign in to continue reading
New to Counterpoint?
Register to our service
Category
Industry
AI , Semiconductors
Report Type
Report
Time period
Other
Contact us
Author
MS Hwang
MS Hwang is a research director at Counterpoint, specializing in memory semiconductor research. MS Hwang brings over 30 years of experience from Samsung Electronics and sell-side brokerage research roles including ABN AMRO, Goldman Sachs, Credit Suisse and Samsung Securities.
Related Research
Mar 24, 2025
Mar 20, 2025
Mar 18, 2025