Tag: ai infrastructure

Researchers from Princeton University and the University of Washington propose SPAD, a novel hardware design that tailors specialized chips for the distinct prefill and decode phases of LLM inference. This approach aims to overcome the inefficiencies of general-purpose hardware, leading to significant cost and power savings.

Brookfield and Bloom Energy have announced a significant $5 billion strategic partnership to develop AI infrastructure, marking a pivotal moment for onsite power solutions. This collaboration aims to build AI factories by integrating Bloom Energy's fuel cell technology with Brookfield's infrastructure development and financing expertise, addressing the immense power demands of artificial intelligence and reshaping the future of data center energy.

NVIDIA, alongside over 50 partners, is pioneering the development of efficient gigawatt AI factories by embracing 800 VDC data center infrastructure and introducing the Vera Rubin NVL144 and NVIDIA Kyber architectures. This strategic move aims to meet the escalating demands of advanced AI workloads, enhance energy efficiency, and enable unprecedented scalability in AI computing.

Intel and Exostellar announce a strategic partnership to enhance enterprise AI infrastructure, combining Intel Gaudi accelerators with Exostellar's orchestration platform. This collaboration aims to deliver cost-effective, cloud-like agility for on-premises and hybrid AI environments, enabling faster scaling of AI initiatives with an open, multi-vendor ecosystem.