NVIDIA Just Showed Everyone What the Next 5 Years of AI Will Look Like

NVIDIA's annual GTC conference kicked off this week, and CEO Jensen Huang came with receipts. The company unveiled new chips and software designed for what it calls the next big phase of AI: inference.

Here's the simple version. Training an AI model is like teaching a student everything they need to know. Inference is when that student actually goes out and does the job. Until now, most of the hype (and money) has been about training bigger and bigger models. But NVIDIA is betting that the real money is in making those models actually work in the real world, faster and cheaper.

The new chips are built specifically for this. They're designed to run AI models inside apps, enterprise software, and autonomous systems at a fraction of the current cost. That's huge because right now, running AI is expensive. Really expensive. If NVIDIA can bring those costs down, it opens the floodgates for AI to show up in basically everything.

NVIDIA also showed off advances in its Rubin platform, which delivers 5x better performance per watt compared to previous generations. That means more AI power using less electricity, which is a big deal when data centers are already straining power grids worldwide.

As reported by Reuters.

Source: Reuters