Prakash Arunkundrum, HP’s first-ever chief strategy and transformation officer, bets edge AI will ‘bring the token cost down’

Chat GPT transformed AI from a niche, behind-the-scenes tool into one of the fastest-growing consumer products in history. That’s only accelerated this year, as use of tokens—the basic building blocks of any AI query—has exploded, thanks to widespread enterprise adoption and the rise of agentic AI platforms like Open Claw. That’s caused surging AI costs in enterprises, as their systems start to consume more and more tokens. “Every chat generates a token, and when you scale AI use across organizations, costs start adding up,” Prakash Arunkundrum, HP’s chief strategy and transformation officer (CSTO), tells Fortune. “This is what all the data center guys are trying to solve: How to bring the token cost down.” HP’s solution to the token problem is to sidestep it entirely, by shifting AI from the cloud to people’s devices, or the “edge.” HP is betting on “AI PCs,” personal devices capable of running AI models locally, and has also debuted its own local-first AI model, HP IQ. “At the core of our strategy, we are building AI-powered devices, think AI PCs, Copilot-enabled printers, and meeting rooms that have smart intelligence peripherals which understand you,” Arunkundrum says. “We’re then stitching them together with software, so that we can deliver experiences in an office across the entirety of your device fleet.” HP reported better-than-expected earnings in the first quarter of 2026. The U.S. company generated $14.4 billion in revenue, a nearly 7% jump. Personal systems revenue came to $10.3 billion, increasing by 11%. Interim CEO Bruce Broussard credited the jump to “continued momentum in AI PCs,” which now make up 35% of devices sold. The company will report second quarter earnings on May 27. “Edge AI is local, private, and secure, and it drives down the total cost of ownership by slashing token use,” Arunkundrum says. “Imagine you’re running a quality inspection in a manufacturing line—you don’t want that to go somewhere else.” Running AI locally also helps en