NVIDIA has announced an expansion of its NVLink Fusion platform with NVHBM, a custom high-bandwidth memory offering. According to the company, growing AI workloads — including AI agents and trillion-parameter models — are creating new demands on infrastructure that go beyond raw compute. NVIDIA frames performance as dependent on how compute, memory, storage, networking and software are designed together as a unified system, and positions the NVHBM expansion as a way to help hyperscalers and AI innovators build the next generation of infrastructure.
Why it matters
The announcement reflects a shift in how large-scale AI systems are architected, with memory bandwidth cited as a key factor alongside compute. Custom high-bandwidth memory integrated into the NVLink Fusion platform is presented as addressing bottlenecks tied to increasingly large model and agent workloads.
Who should care
Hyperscalers and organizations building large-scale AI infrastructure are the stated audience for this expansion.