Nvidia announces next-generation datacenter processors as part of its Age of AI initiative.

At the GPU Technology Conference in San Francisco on Monday, October 6, Nvidia CEO Jensen Huang unveiled next-generation chips for networking, storage and security, along with major announcements spanning data centres, edge AI, collaboration tools and healthcare. Huang emphasized that "AI requires a whole reinvention of computing – full-stack rethinking – from chips, to systems, algorithms, tools and the ecosystem," describing Nvidia as "a full stack computing company" dedicated to solving "extremely hard computing problems that have a great impact on the world."

The centrepiece of Nvidia's announcements was the introduction of the data processing unit (DPU), a new kind of processor required to run modern data centres. The BlueField 2 DPU features programmable Arm cores and acceleration engines for at-line-speed processing of networking, storage and security operations. This technology emerged from Nvidia's acquisition of Mellanox Technologies, which closed in April. Huang announced that BlueField-2 is sampling now, BlueField-3 is finishing, and BlueField-4 is in high gear, stating that "in just a couple of years, we'll span nearly 1,000 times in compute throughput on the DPU." Nvidia also announced DOCA, a programmable data-centre-infrastructure-on-a-chip architecture, whose SDKs "let developers write infrastructure apps for software-defined networking, software-defined storage, cybersecurity, telemetry and in-network computing applications yet to be invented."

Nvidia announced partnerships with major technology companies to accelerate AI adoption. Nvidia and VMware announced two collaborations: one to port VMware onto BlueField and another to create a data centre platform supporting GPU acceleration across virtualised, distributed scale-out and composable microservices domains. Huang noted that VMware "runs the world's enterprises — they are the OS platform in 70 per cent of the world's companies." Additionally, Microsoft is adopting NVIDIA AI on Azure to power smart experiences on Microsoft Office, including smart grammar correction and text prediction, joining Square, Twitter, eBay, GE Healthcare and Zoox in using Nvidia GPUs for inference. Cloudera announced it would accelerate the Cloudera Data Platform with Nvidia RAPIDS, Nvidia AI and NVIDIA-accelerated Spark.

Huang introduced several new services and tools addressing emerging computing needs. Fleet Command, a new application in early access, enables easy deployment and management of updates across IoT devices, combining edge computing security with the remote management of software-as-a-service. Nvidia Maxine, a cloud-native streaming video AI platform, was announced for applications like video calls. To address drug discovery during the global Covid-19 pandemic, Nvidia unveiled NVIDIA Clara Discovery, a state-of-the-art suite of tools for scientists to discover life-saving drugs. The GPU Technology Conference itself, one of the biggest yet, features more than 1,000 sessions — 400 more than the last GTC — across 40 topic areas.