Nvidia custom 'NVHBM' promises 30% higher bandwidth, 15% lower power than commodity HBM4e — custom base die and PHY will be available to NVLink Fusion partners
Nvidia introduced NVHBM, a custom high-bandwidth memory for AI accelerators, promising 30% higher bandwidth and 15% lower power than HBM4e. It's available to NVLink Fusion partners, with Amazon's Annapurna Labs as the first adopter. The technology aims to improve performance and efficiency for AI workloads, though benefits won't appear in current systems.
How this was made

The 30-second read
Why it matters
The announcement introduces a new memory architecture that could improve AI accelerator efficiency and create a competitive moat for Nvidia.
Market read
The news may influence Nvidia's stock and the broader AI hardware sector as investors gauge the commercial rollout of NVHBM.
What to watch
Potential supply‑chain constraints for the custom base die and the need for partner redesigns may delay revenue impact.
Background
Nvidia's NVLink Fusion program enables multi‑chip AI systems; NVHBM is the latest addition aimed at custom silicon partners.
Ticker impact
Nvidia announced its new NVHBM custom high‑bandwidth memory offering, promising up to 30% higher bandwidth and 15% lower power, and named Amazon's Annapurna Labs as the first partner.
Short‑term neutral to slightly bullish for NVDA as investors assess the technology's market potential.
The product is a first‑time disclosure with clear technical advantages, but commercial adoption timelines are uncertain.
Market effects
Strengthens Nvidia's lead in AI hardware and may pressure competing memory providers and custom silicon vendors.
U.S. semiconductor and AI hardware markets see incremental upside; limited immediate effect on other regions.
Relevant to global AI accelerator designers and cloud providers evaluating next‑gen memory solutions.
Counterpoint
Adoption could be slower than expected if partners favor existing commodity HBM solutions or if cost differentials outweigh performance gains.
Key entities
- CompanyNvidia
Developer of the NVHBM custom memory solution.
- CompanyAmazon Annapurna Labs
First partner to integrate NVHBM into its upcoming Trainium 4 AI chips.




