As generative artificial intelligence continues to disrupt everything, a big conversation taking place revolves around how much a company is going to spend and whether the juice is worth the squeeze. Everyone knows they want generative AI, but how best to handle the gen AI costs?

One of the key thoughts that is taking place in the market right now is that people are starting to realize the benefits of gen AI, according to Aran Khanna (pictured), co-founder and chief executive officer of Archera, a cloud optimization software company. That comes from a cost perspective and a revenue perspective.

“What customers are starting to realize is that, you know, the science, the best practices, they’re only about a year old, and the hardware space is constantly changing,” Khanna said.

Khanna spoke with theCUBE Research executive analyst John Furrier at the “Supercloud 6: AI Innovators” event, during an exclusive broadcast on theCUBE, SiliconANGLE Media’s livestreaming studio. They discussed impact and opportunities of generative AI in cost optimization, alongside the evolving landscape of cloud computing and AI technology.

Uncertainty reigns in gen AI costs

When it comes to the hardware space, it’s unknown what the next-generation, state-of-the-art model is going to require. The actual cost to serve these models once they get product market fit and reach production is also up in the air, Khanna noted.

“I think that uncertainty is something that’s universal across the customers that we talk to,” he said. “When they’re looking to do their one-year, three-year-plus commitments to their cloud provider and figure out how much of this capacity — from a GPU standpoint or from a high-end ASIC standpoint like Inferentia or TPU — should I plan to actually be spending on?”

At Supercloud 6, two things have been emerging, including the organic growth of developer appetite for building on top of infrastructure for gen AI. Also, enterprises are bringing their corporate data to the table, which isn’t easy to integrate into a bottom-up, according to Furrier.

It’s a multi-tiered problem because generative AI — the fancy stuff with the GPUs — requires organizations to get their data strategy right to make it useful in the context of the specific business, according to Khanna. In some cases, it requires just getting data to the cloud in the first place.

“You can use those platforms like Amazon EMR or Databricks or Snowflake to start extracting value out of those data sets that are proprietary to your business and really are the moat to the generative AI that you build that powers your business on a top-line or bottom-line basis,” Khanna said.

But there are risks at every piece of the stack that needs to be built to support gen AI, according to Khanna. CIOs are starting to realize this, as are their counterparts on the finance side.

“Which is where a product like ours, that’s really blending that insurance aspect with deep technical and cost optimization knowledge, can really make an impact on the customers we work with,” he said.

Here’s the complete video interview, part of SiliconANGLE’s and theCUBE Research’s coverage of the “Supercloud 6: AI Innovators” event:

Photo: SiliconANGLE