AI is crashing now.

SAN FRANCISCO — In June, Inflection AI was riding high. The artificial intelligence start-up founded by veterans of Google’s famous DeepMind AI lab had just raised $1.3 billion from Microsoft and tech billionaires Bill Gates, Reid Hoffman and Eric Schmidt to build out its chatbot business.

But less than a year later, the tide has turned. Co-founders Mustafa Suleyman and Karén Simonyan have left Inflection AI for jobs at Microsoft, which now has the rights to use its technology. Inflection AI is now focused on helping other companies improve their AI tools.

It’s not the only case of AI hype coming back down to earth. After 11 months of public testing, Google’s AI search tool still constantly makes mistakes and hasn’t been released to most people. New scientific papers are undermining some of the flashier claims about the tech’s capabilities. The AI industry is also facing a growing wave of regulatory and legal challenges.

A year and a half into the AI boom, there is growing evidence that the hype machine is slowing down. Drastic warnings about AI posing an existential threat to humanity or taking everyone’s jobs have mostly disappeared, replaced by technical conversations about how to cajole chatbots into helping summarize insurance policies or handle customer service calls. Some once-promising start-ups have cratered, and the suite of flashy products launched by the biggest players in the AI race — OpenAI, Microsoft, Google and Meta — have yet to upend the way people work and communicate with one another. While money keeps pouring into AI, very few companies are turning a profit on the tech, which remains hugely expensive to build and run.

“If you compare a mature market to a mature tree, we’re just at the trunk,” said Ali Golshan, founder of Gretel AI, a start-up that helps other companies create data sets for training their own AI. “We’re at the genesis stage of AI.”

The tech industry is not slowing down its charge into AI. Globally, venture-capital funding in AI companies grew 25 percent to $25.87 billion in the first three months of 2024, compared with the last three months of 2023, according to research firm PitchBook. Microsoft, Meta, Apple and Amazon are investing billions into AI, hiring PhDs and building data centers. Earlier this month, Amazon poured $2.5 billion into Anthropic AI, bringing its total investment in the OpenAI competitor to $4 billion. This week, Microsoft, Google and Meta all confirmed they would spend billions of dollars on new data centers for AI this year alone.

The immense cost of training AI algorithms — which requires running mind-boggling amounts of data through warehouses of expensive and energy-hungry computer chips — means that even as companies like Microsoft, OpenAI and Google slowly begin charging for AI tools, they are still spending billions to develop and run those tools.

Google Cloud CEO Thomas Kurian said this month that Deutsche Bank, the Mayo Clinic, McDonald’s and other companies were using Google’s tools to build AI applications. And Google CEO Sundar Pichai said on the company’s most recent conference call that interest in AI helped contribute to an increase in cloud revenue. But on the same call, Chief Financial Officer Ruth Porat said the company’s investment in data centers and computer chips to run AI would mean Google’s expenses would be “notably larger” this year compared with last year.

“We are helping our customers move from AI proof-of-concepts to large-scale rollouts,” said Oliver Parker, an executive with Google Cloud’s AI team. He pointed to Discover Financial Services as one example — the company used Google’s AI to build a tool for its call centers that reduced the time it took to do each call.

Microsoft has also been trumpeting the interest in its AI tools and says 1.3 million people use its GitHub Copilot AI code-writing assistant. It’s also pushing a $30-a-month AI assistant to the millions of Microsoft Office users around the world. But the company has been mum on whether any of the tools are profitable when compared with the cost of running them. Like Google, it has focused much of its energy on getting customers to use its cloud services to run their AI apps.

“We’re finding that AI requires a paradigm shift,” said Jared Spataro, Microsoft’s corporate vice president of AI at Work. “It’s not like a traditional technology deployment where IT flips a switch. Businesses need to identify areas where AI can make a real impact and strategically deploy AI there.”

In October, OpenAI announced its own version of Apple’s app store, where people could make custom versions of the popular ChatGPT chatbot, upload them to a public marketplace, and get paid by OpenAI if many people use them. Three million custom GPTs have been created, but the company hasn’t said whether it has paid out any money yet.

Meta, which is integrating chatbots into its social networks and messaging platform, said this week that it would spend as much as $40 billion this year on new infrastructure to support its AI build-out.

“Building the leading AI will also be a larger undertaking than the other experiences we’ve added to our apps, and this is likely going to take several years,” CEO Mark Zuckerberg said on a conference call Wednesday.

“These tools are not yet pervasive, not even close,” said Radu Miclaus, an AI analyst at tech research firm Gartner. He expects that to begin changing soon. “My expectation is that this year the applications will start taking off.”

The money game

Beyond the Big Tech companies, a legion of start-ups is trying to find ways to make money with such “generative” AI tech as image generators and chatbots. They include trying to replace customer service agents, writing advertising copy, summarizing doctors’ notes and even trying to detect deepfake AI images made by other AI tools. By replacing workers or helping employees become more productive, they hope to sell subscriptions to their AI tools.

Many computer programmers say they already use chatbots to help write routine code. Usage across the rest of the workforce is also slowly increasing, according to the Pew Research Center. Around 23 percent of U.S. adults said they had used ChatGPT at work in February, compared with 18 percent in July, the research organization found.

“We’re at the very, very beginning,” said John Yue, founder of Inference.AI, a start-up that helps other tech companies find the computer chips they need to train AI programs. AI will work its way into every single industry, he said, but it might take at least three to five years before people really see those changes in their own lives. “We have to take a longer look.”

There are big roadblocks that could also slow down the industry. Governments have bought into AI hype, and politicians in the United States and abroad are busy debating how to regulate the tech. In the United States, smaller AI companies have expressed worries that AI leaders such as Google and OpenAI will lobby the government to make it harder for new entrants to compete.

In the European Union, final details of the bloc’s AI Act are being hammered out, and they promise to be more restrictive than companies are used to. One of the biggest concerns that Golshan, the Gretel AI CEO, hears from business clients is that new AI laws will render their investments a waste. Lawsuits also have been launched against OpenAI and other AI companies for using people’s work and data to train their AI without payment or permission.