Copyright SiliconANGLE News
Cloud companies are spending billions on artificial intelligence infrastructure while AI agents battle it out. In the past few days, Amazon Web Services Inc. made a $38 billion deal with OpenAI Group PBC to provide compute, in addition to a $5.5 billion deal with Cipher Mining, and Microsoft inked a $9.7 billion deal for AI cloud capacity with IREN Ltd., an Australian data center business. Amid the spending frenzy, there are concerns about powering all these data centers, and the cloud giants are taking stronger steps to defend their turf from rival AI agents. “The skirmishes going on between who’s got the best agent technology is a signal that people are preparing,” said John Furrier (left), executive analyst for theCUBE Research. “The picks and shovels are getting ready for the gold rush. The gold is the energy.” On the latest episode of theCUBE Pod, Furrier and Dave Vellante (right), chief analyst for theCUBE Research, discussed the agentic AI battlefield and why compute is the bottleneck on AI infrastructure. They also discussed how AI platforms are changing the browser experience and who is going to own the intelligence layer in the AI stack. The gold rush for AI infrastructure The tech industry is seeing a shift from traditional data centers to AI factories, purpose-built systems for AI production. TheCUBE Research predicts that this new kind of infrastructure will evolve into agentic-powered services, where the consumer pays for the outcome, rather than the software itself. There are two different economic models emerging from the growth of AI infrastructure: capital expenditure and the pay per graphic processing unit model. The biggest demand right now is energy, which companies need to power the entire agentic enterprise. “It’s not just about the money on the CapEx,” Furrier said. “It’s the shape of the next computing cycle, because the power envelope, the policies that are very crypto and AI friendly right now in the U.S., is spawning a massive gold rush for AI infrastructure.” Do all of those billion-dollar investments mean we are in a bubble? Furrier thinks agentic AI represents a deeper shift in the industry, while Vellante sees the bubble popping, but not just yet. AI will significantly increase in value around 2030, Vellante predicts, so it may be a long road until companies see a return on their investments. “It’s just going to take a while,” Vellante said. “Part of the reason why it’s going to take so long is … the denominator is so big, the CapEx going in is so large. We got to create $4 trillion in value. That just doesn’t happen overnight. There’s not enough energy.” The power problem Compute is quickly becoming the new oil in enterprise technology — a finite, fiercely contested resource powering the AI boom. Microsoft Corp. CEO Satya Nadella underscored the urgency this week, noting that the company doesn’t have enough electricity to deploy all the GPUs it already owns — a bottleneck that’s only expected to intensify as demand for AI compute continues to soar. In addition to power, AI factories require land and clean water, and that means their builders have to work with the local community, who may or may not be receptive to the intrusion. The centers also need skilled laborers, who might demand large incentives. “These are not just everyday skills,” Vellante said. “[Companies need] to attract them to go live in these places that aren’t necessarily these big metropolises. They’re sort of out in the middle of nowhere. So, they’re going to have to pay some serious doubt to lure these guys.” The data centers have the potential to create jobs for their communities, but they’ll have to ingratiate themselves with the locals first, according to Furrier. He compares it to the resentment big-box companies such as Walmart face when they displace local businesses. “When you start to look at these factories … they’re bringing a lot of money to the table,” Furrier said. “They want to have balance. The big-box companies would come in, take over small town America. The data center could bring that same mojo, and we’ll see. I think Abilene was a good example.” Agentic turf wars ignite Last Tuesday, Amazon.com Inc. fired what could be the opening salvo in a battle between AI agents. The company blocked AI startup Perplexity AI Inc. from using its Comet browser’s agents to make purchases on Amazon’s Marketplace, presumably because it infringes on Amazon’s ad revenue and its own AI shopping agent. The intelligent browser is part of the new “system of intelligence,” according to Furrier and Vellante. Whoever can harmonize all of the data and context needed for enterprise agents will be a winner in this ecosystem, and Palantir Technologies Inc., Celonis SE, Salesforce Inc. and ServiceNow Inc. are all going after that crown, with Snowflake Inc. and Databricks Inc. as the data infrastructure layer. “Snowflake and Databricks … they’re in good shape because they have the data that the customers that use them have — good data,” Vellante said. “You can bring agents to them now and actually be productive. But the question is, as they go up the stack and into the world of process … Salesforce, SAP, ServiceNow, Workday, etc., how do they play there?” Furrier predicts that platform companies will essentially become the applications, with AI utilities as the new “apps.” The future of AI workloads will be a big topic at KubeCon this week, and Microsoft Ignite later this month. “The platforms have to let agents connect and work on behalf autonomously for users,” Furrier said. “The app is the platform, and it’s not so much vertical integration. It’s just more of whoever builds the end-to-end substrate, if you will. When that happens, the agent tsunami comes.” Watch the full podcast below to find out why these industry pros were mentioned: Sam Altman, co-founder and CEO of OpenAI Brad Gerstner, founder and CEO of Altimeter Capital Management Satya Nadella, chairman and CEO of Microsoft Erik Brynjolfsson, American academic and author Jackie McGuire, principal analyst at theCUBE Research Andy Jassy, president and CEO of Amazon James Hamilton, SVP and distinguished engineer at Amazon Jensen Huang, founder and CEO of NVIDIA Steve Jobs, co-founder and former CEO and chairman of Apple Rob Strechay, principal analyst for theCUBE Research Savannah Peterson, principal analyst at theCUBE Research George Gilbert, principal analyst at theCUBE Research Michael Saylor, CEO of MicroStrategy Tom Lee, cryptocurrency analyst Here’s the full episode of this week’s theCUBE Pod: Don’t miss out on the latest episodes of “theCUBE Pod.” Join us by subscribing to our RSS feed. You can also listen to us on Apple Podcasts or on Spotify. And for those who prefer to watch, check out our YouTube playlist. Tune in now, and be part of the ongoing conversation. Image: SiliconANGLE
