Welcome to Eye on AI, with AI reporter Sharon Goldman. In this edition…Meta’s giant Hyperion data center gets even bigger…OpenAI and Amazon talk about an alliance…Is the AI bubble popping because AI is actually working?…AI makes inroads at the Super Bowl (at least in the ads). I’m in a bit of recovery mode today, having just returned from several days in northeast Louisiana visiting Meta’s massive AI data center site, known as Hyperion, for a feature story I’m reporting.
I’ve been scouring the thesaurus, trying to land on the right word to describe just how large, loud, and chaotic this construction site is. Colossal? Mammoth? Sprawling? Let’s put it this way: it takes a while just to drive the length of the site—it stretches roughly five miles from top to bottom.
And during that drive, I discovered that Hyperion is getting even bigger. While the expansion had long been suspected—and was something of an open secret among some locals—I confirmed that
Meta has quietly purchased roughly 1,400 additional acres, an area nearly twice the size of Manhattan’s Central Park, adjacent to its already-mega 2,250-acre campus. I also observed active construction underway on the newly acquired land—when I wasn’t worrying about getting mowed down by the endless parade of 18-wheelers hauling materials in and around the site.
A potential Amazon-OpenAI deal to power AlexaNow that I’m back in one piece, I want to turn to a different news item that caught my eye this week. As
Amazon weighs an equity investment of tens of billions of dollars in OpenAI, it is reportedly in talks to use OpenAI models to power some of its internal AI products, including the Alexa voice assistant. The deal, first
reported by
The Information, would involve OpenAI employees helping to customize models for Amazon’s needs.
The news comes just a day after Amazon finally made its Alexa+ AI assistant available to everyone in the U.S., nearly a year after its initial launch. I attended the splashy unveiling in New York City last February, when the company pitched the service as a souped-up version of the original 11-year-old Alexa—one that could handle multiple queries at once and act as an “agent,” taking actions on your behalf like booking a repairman or ordering an Uber.
But even recently, beta testers were
voicing plenty of frustrations. “When I ask Alexa to turn off the light, it should turn off the light—not everything on the strip,” one software engineer wrote last October in an internal channel for feedback on unreleased Alexa+ features. “It turned off the power strip that my aquarium filter is on and killed my fish.”
Other testers complained that the assistant talked nonstop, ignored repeated commands to be quiet, or blasted music at full volume when no one was home.
Alexa’s AI has been a long, slow journeyI’ve been following the journey of Amazon’s Alexa closely since the post-ChatGPT dawn of generative AI in 2023. That September, Amazon held another glitzy event—this time at its second headquarters in Washington, D.C.—where David Limp, then the company’s head of devices and services, demonstrated a new generative-AI-powered Alexa by saying, “Alexa, let’s chat.”
But nearly a year later, I
reported—based on interviews with more than a dozen former employees who worked on Alexa’s AI—that the organization was beset by structural dysfunction and technological challenges that repeatedly delayed shipment of the new generative AI Alexa. Those former employees painted a picture of a company lagging behind Big Tech rivals Google, Microsoft, and Meta in the race to deploy AI chatbots and agents, and struggling to catch up.
The September 2023 demo, they emphasized, was just that—a demo. The large language model at the heart of the new Alexa, which Amazon positioned as a rival to OpenAI’s ChatGPT, was, according to former employees, far from state-of-the-art. Research scientists said Amazon lacked both the data and the specialized computing infrastructure needed to train and run leading-edge LLMs at scale.
Since that report, Amazon
has shifted to a more hybrid approach. Alexa+ has now been powered by a mix of Amazon’s own Nova models and models from Anthropic, the AI startup in which Amazon has invested $8 billion.
And yet, problems persist. Can OpenAI really solve Alexa’s long-standing problems? Or would an Amazon–OpenAI deal divert OpenAI’s attention at a moment when it is locked in fierce competition with Google and Anthropic? And looming over it all is Apple, whose own deal with Google to power Siri complicates an already crowded AI landscape.
More than anything, the reported talks reveal how desperate even the biggest players have become to stay ahead in a no-brakes AI race.
With that, here’s more AI news.
Sharon Goldmansharon.goldman@fortune.com
@sharongoldman
