Tech companies are racing to expand their infrastructure as their increasingly resource-heavy AI products consume capacity, deplete chipmakers’ supply, and demand more power. Google, once known as the “King of the Web, ” is among these companies. A senior executive from Google reportedly informed staff that the company must exponentially scale its serving capabilities to meet the growing demand for its AI services. CNBC obtained a recent presentation by Amin Vahdat, VP of Machine Learning, Systems, and Cloud AI at Google. The presentation contained a slide on “AI compute demand” stating that Google “must double every 6 months…aiming for the next 1000x increase in 4-5 years. ” According to reports from the all-hands meeting where the presentation was delivered, Vahdat said, “The competition in AI infrastructure is the most critical and the most expensive part of the AI race. ” He added, “Our job is to build this infrastructure, but not necessarily to outspend competitors. Still, we will spend a lot, ” aiming to create AI infrastructure that is “more reliable, more performant, and more scalable than what’s available anywhere else. ” After CNBC’s story came out, Google disputed aspects of the report.

While CNBC initially quoted Vahdat as saying the company would need to “double” its compute capacity every six months, a Google spokesperson told Gizmodo that his words were taken out of context. The spokesperson clarified that Vahdat “was not referring to a capital buildout of the magnitude suggested. He simply stated that demand for AI services requires significantly more computing capacity, which Google is addressing through efficiency improvements across hardware, software, and model optimization, as well as new investments. ” CNBC has since updated its report to distinguish between “compute” and “serving” capacity. Serving capacity refers to Google’s ability to manage the rising volume of user requests, whereas compute capacity encompasses the overall AI infrastructure, including resources for training new models and other costs. When asked for clarification on these terms, the spokesperson explained that the original headline “read as if he implied we are doubling the amount of compute we have—whether measured by the number of chips or megawatts of electricity. ” Instead, the capacity growth Vahdat described will come from multiple factors such as more capable chips and efficiencies in models and optimizations. Regardless of these nuances, it appears Google—like its rivals—must scale operations significantly to support its emerging AI infrastructure business. Vahdat’s remarks came shortly after Google reported strong profits from its Cloud business and stated plans to increase spending in the coming year. During the presentation, Vahdat also reportedly stated that Google needs to “deliver 1, 000 times more capability, compute, storage, and networking [than its competitors] at essentially the same cost, and increasingly, the same energy consumption. ” He acknowledged the challenge by saying, “It won’t be easy, but through collaboration and co-design, we’re going to get there. ”