The world has yet to turn any of today’s AI hype and spending into a meaningful lift in the actual economy, says Microsoft CEO Satya Nadella. The comments were made during a podcast interview in which he put a measuring stick on AI’s true impact: Success will be measured through global economic growth rather than arbitrary benchmarks of how well AI programs can complete challenges like obscure math puzzles. Those are interesting in isolation but do not have practical utility.
The comments seem to be targeted directly at the tech industry, imploring insiders to stop spreading hype about a so-called “artificial general intelligence” that could replace humans in most tasks. Nadella said essentially that it will not happen, and either way is an unnecessary distraction when the industry needs to get practical and just try and make money before investors get impatient. Microsoft is the primary backer of OpenAI, whose CEO Sam Altman has long stoked fears of AI taking over the world. Critics often say Altman’s hype is primarily about placing himself at the centers of power and maintaining control through regulatory capture.
“If you’re going to have this explosion, abundance, whatever, commodity of intelligence available, the first thing we have to observe is GDP growth,” Nadella said. “This is where we get a little bit ahead of ourselves with all this AGI hype. When we say this is like the Industrial Revolution, let’s have that Industrial Revolution type of growth.”
He went on to say that 10% of inflation-adjusted growth attributed to AI would prove that the technology is actually as meaningful as the Industrial Revolution. “It can’t just be supply side,” meaning the hundreds of billions being invested into AI companies today does not count as real GDP growth, because eventually there needs to be actual demand on the other side for the products being built or else these companies will crash and burn.
.@satyanadella on:
– why he doesn’t believe in AGI but does believe in 10% economic growth
– Microsoft’s new topological qubit breakthrough and gaming world models
– whether Office commoditizes LLMs or the other way around
Links below. Enjoy!
Timestamps
0:00:00 – Intro… pic.twitter.com/ywwCsPn1xd— Dwarkesh Patel (@dwarkesh_sp) February 19, 2025
Nadella’s comments appear to be intended to lower expectations while the AI industry remains white-hot. Still, Microsoft itself continues to plow billions of dollars into competing in the AI race, having been an early backer of OpenAI and putting more than $12 billion into the startup. Nadella justified massive infrastructure projects like OpenAI’s Stargate as necessary to bring the costs of AI down and support that massive GDP growth (Meta CEO Mark Zuckerberg has made similar comments), but Microsoft has also let OpenAI take infusions of cash from others and has been unwilling to supply all the compute resources the startup wants, suggesting it is being cautious. It has been developing its own lower-cost models to supplement OpenAI’s ChatGPT.
Across numerous industries, companies are struggling to implement the AI tools that have been built today, perhaps because they are not comfortable using AI for critical functions, or they do not know how best to use it. The early reception of Microsoft’s Copilot, an AI agent for enterprises, has not been good, with companies and even insiders at Microsoft complaining it is lackluster and not worth the cost. Microsoft and Google have begun incorporating AI chatbots into their standard enterprise subscriptions rather than trying to charge extra, essentially forcing them on users following low adoption.
Klarna, the buy-now-pay-later company, garnered a lot of attention after its CEO said last year it would replace most of its customer support representatives with AI chatbots. The CEO recently walked that claim back, as reports found that Klarna had essentially just replaced a basic phone tree system with AI, which is not revolutionary and likely resulted in many customers quitting chats out of frustration. Many banks tried replacing their L1 customer support with bots in the past to bad results.
We just had an epiphany: in a world of AI nothing will be as valuable as humans!
Ok you can laugh at us for realizing it so late, but we are going to kick off work to allow Klarna to become the best at offering a human to speak to!!!
So excited about this, more to come!
— Sebastian Siemiatkowski (@klarnaseb) February 14, 2025
The AI industry continues to look for the ultimate unlock that will make AI truly “intelligent” and useful for a broad range of tasks, like adding bots that can navigate a computer. But language models can fundamentally be described as supercharged autocomplete tools, prone to returning incorrect information because they are skilled at creating a facsimile of a human-written sentence—something that looks good—but chabots are not doing any actual thinking. Various methods like test-time thinking have been deployed to try and improve the accuracy and performance of chatbots and make them more closely mimic humans. But the nature of chatbots to answer questions confidently even when they do not know the truth remains a problem that results in users spending significant time just fact-checking the results. Or worse, simply accepting what they return as fact and making decisions based on it.
In that sense, Nadella is trying to slap tech executives awake and tell them to cut out the hype. AI safety is somewhat of a concern—the models can be abused to create deepfakes or mass spam—but it exaggerates how powerful these systems are. Eventually, push will come to shove and the tech industry will have to show that the world is willing to put down real money to use all these tools they are building. Right now, the use cases, like feeding product manuals into models to help customers search them faster, are marginal. Bank of America issued a report in late 2024 warning that AI remains in its early stages and is following the trajectory of the internet in the 1990s, implying there will be a correction similar to what occurred in the dot-com boom. Nadella echoed similar concerns of a washout coming.
A lot of tech-savvy people on X are using chatbots all the time, but X is an echo chamber, and that usage is not reflected in the real world. ChatGPT claims more than 400 million active users, but that includes consumers on the free tier, who do not have the same mission-critical needs of businesses. Eventually, the check will come due.
