Rambus Launches HBM4E Memory Controller for AI and GPUs, CFO Resigns Amid Strong Q4 Results
March 4, 2026
Rambus unveils the HBM4E Memory Controller IP tailored for AI accelerators and GPUs, enabling high memory bandwidth and compatibility with 2.5D or 3D memory subsystems.
Industry stakeholders, including Samsung Electronics, praise the technology for AI workloads and the broader adoption of HBM4E.
The controller supports operation up to 16 Gbps per pin and delivers up to 32 TB/s total bandwidth for an eight-device HBM4E configuration.
In Rambus’ Q4 2025 results, EPS met expectations and revenue slightly beat forecasts, while CFO Desmond Lynch resigned effective February 27, 2026, with John Allen serving as interim CFO.
The press release carries standard forward-looking statements and outlines various risks and governing factors.
Press contact is Cori Pasinetti; the release was distributed by Business Wire, with Rambus listed as the source.
Licensing for the HBM4E Controller IP is available immediately, with early access programs for design customers.
Rambus notes strong financial context, citing about 80% gross margin and roughly 27% revenue growth over the past year to support market potential.
The article frames Rambus’s broader focus on data center and AI infrastructure silicon IP and notes the leadership transition as a near-term consideration for investors.
Rambus stock moved lower in after-hours trading following the updates, despite a history of solid annual returns.
The announcement includes links to Rambus product pages and corporate information, reinforcing Rambus’s focus on faster, safer data through memory-processor bottleneck solutions.
Rambus cites its track record of over 100 HBM design wins to bolster confidence in first-time silicon success for the new HBM4E IP.
Summary based on 9 sources
Get a daily email with more AI stories
Sources

Business Wire • Mar 4, 2026
Rambus Sets New Benchmark for AI Memory Performance with Industry-Leading HBM4E Controller IP
SiliconANGLE • Mar 3, 2026
Rambus targets AI memory bottlenecks with new HBM4E controller IP - SiliconANGLE
