After the dot-com bubble, people are a little more careful of tech bubbles—even after Steve Jobs’s touch phone thing paid off. But Sharma's throwing down the gauntlet to tech heads to ensure the AI dream doesn't become a nightmare.

Sharma said she had seen it all before. When everyone was gaga over cloud computing, the bills piled up as fast as the prototypes. And now, with AI, it's déjà vu all over again.

She said history's got a lesson here: just like when the techies and suits had to buddy up to create DevOps, it's time for the AI boffins to get cosy with the infrastructure crowd—Sharma's betting on a mash-up between cloud natives and AI whizz kids.

Kubernetes is the new black for launching big-time apps, and it's no shocker that it's the go-to for AI models. It's versatile.

Sharma has started calling Kubernetes the "Linux of AI" – it's the powerhouse behind the scenes and the foundation of AI fantasies.

CNCF's AI division dropped a white paper to help clear up the confusion at the conference.

One big headache? Those pesky "walled gardens" in the cloud. Developers trying to hit the big leagues are hitting walls with proprietary tech that's about as flexible as a brick.

Sharma said the only way around all this is open-source, which would break down those walls and let the AI freedom flag fly. She's all for licenses like Apache 2 or MIT that let companies tailor AI to their hearts' content.

The CNCF's AI team devised a blueprint with open-source goodness to help projects scale up, just like Kubernetes does for apps.

Sharma says it's more complicated if AI should be open source. She thinks that if you trust your data with an AI, you should know what it's up to. But if you spill all the beans on your AI model, you might as well hand your rivals the keys to the kingdom.