Nvidia Doubles Down on AI with Game-Changing Blackwell Chips

Meet Blackwell: Nvidia’s AI Powerhouse

At the 2024 GTC keynote, Nvidia CEO Jensen Huang revealed the Blackwell platform, a major leap forward in GPU architecture designed to supercharge AI workloads. Expected to roll out in late 2024, Blackwell promises six times the performance of its predecessor, Hopper, using half the energy. The new chip is targeted at training and deploying large language models (LLMs) with tens of trillions of parameters — a move aimed directly at AI leaders like OpenAI, Google, and Microsoft. Nvidia already has commitments from giants such as Amazon Web Services, Meta, and Tesla, reinforcing its position as the primary hardware supplier for the AI arms race.

An Expanding AI Ecosystem

In addition to hardware, Nvidia introduced several innovations to support and monetize AI development. A new software stack—supporting everything from inference to training—and services like Nvidia NIM (Nvidia Inference Microservices) aim to make deploying AI as seamless as using APIs on the cloud. The company is also deepening its partnerships: OpenAI will run Blackwell-powered clusters, and Microsoft will integrate Nvidia’s AI stack into Azure. These announcements signal Nvidia’s strategy to not only power AI infrastructure but also serve as the platform everyone builds on top of.