Some things are going up, most noticeably Nvidia’s stock price, and some aren’t. Most surprisingly, workers and companies aren’t adopting AI.

The market capitalization of the magnificent seven (Apple, Microsoft, Google parent Alphabet, Amazon, Nvidia, Meta Platforms, and Tesla) has risen by more than $10 trillion since January 2023, driven by the increasing expectations for generative AI. 

Even some startups are enjoying rising valuations, such as OpenAI, Anthropic, and Hugging Face. The most recent entrant to this club is Elon Musk’s xAI, which is now valued at $50 billion. It plans to spend most of its recently acquired funding, $6 billion, to purchase 100,000 Nvidia chips for one of its data centers, just as capital expenditures at Microsoft, Meta, Amazon and Google are set to exceed $200 billion this year and top $300 billion next year for chips and other hardware for their data centers.

Training large models still remains the largest application for Nvidia chips, outpacing their use in actual problem solving, sometimes called inference, whose growing demand is being met by Nvidia’s rivals. The two years of explosive growth in Nvidia’s market cap and other members of the magnificent seven has become the “biggest gamble in business history.” 

Few Companies Use AI

As the market capitalizations have risen, so too have claims that these share price increases are one big bubble. One reason is that few companies are actually using AI. The Economist claims that “only 5% of American businesses say they are using AI in their products and services.”

Individual workers have been a bit quicker to adopt AI often on the sly. “One-third of employees in America say they are using it for work once a week,” with 78% of software engineers “using AI at least weekly, up from 40% in 2023, as are 75% of human-resources staff, up from 35%. And OpenAI says 75% of its revenue comes, tellingly, from consumers rather than from corporate subscriptions.” 

Having used computers almost every working hour for the last 25 years and to a slightly lesser extent for phones, I am not overwhelmed by these figures.  Once a week is nothing. Even once a day is not that big a deal. How many workers are using AI every hour? 

Another study finds that adoption by individuals is plateauing. “Between September 2023 and March 2024, AI adoption had steadily increased globally, growing from 20% of the global desk worker population to 32% in March 2024 — or about a third of all desk workers. But over the past three months, AI adoption rates have stalled in some countries; France saw just two percentage point growth, from 31% to 33% of desk workers trying AI, and the U.S. saw just a single percentage point growth, from 32% to 33% of desk workers.”

These adoption numbers are consistent with the small amount of AI software revenues. While the internet bubble of 2000 also required huge outlays for infrastructure, users bought computers to access the internet, paid for internet subscriptions, and also spent billions buying things on the internet, for a total of about $1.5 trillion, far more than the $10 billion or so spent on AI software in 2024. More evidence of a bubble.

Worker Excitement Declines

An even bigger problem might be that worker excitement is falling. A global survey by Slack that had 17,372 respondents found that worker excitement dropped

“six percentage points (47% to 41%) among the overall global population. This trend is being driven by the U.S., where the portion of workers who say they’re excited about AI helping them to complete tasks at work dropped nine percentage points (from 45% to 36%) over the past three months and France where excitement rates dropped 12 points (from 53% to 41%). Excitement also dropped in Japan and the UK.”

How should we interpret these data? The stock market is betting heavily that AI will increase the profits of the magnificent seven and, in the long run, productivity growth. Whether workers will share in this boom is a separate question. This article is more concerned with whether those profits will justify the $10 trillion increase in market capitalization for the magnificent seven, and that will depend on whether an acceleration in productivity growth emerges for corporate users of AI. 

Proponents will say that AI is already diffusing rapidly citing recent anecdotes such as Google’s claim that one-fourth of its code is now written by AI, Amazon CEO Andy Jassy statement that its Company’s AI Assistant Has Saved $260M and 4.5K Developer-Years Of Work, and the New York Times reporting that AI gets more medical diagnoses correct than doctors. When you look under the hood, however, the results don’t look so great. The claims of AI writing code are mostly examples of autocomplete, the same type of autocomplete used by many writers to complete their sentences. And the fact that sample size of six was used to demonstrate the purported superiority of AI doing medical diagnoses suggests that we can’t yet say if AI is as good as doctors, particularly because of its constant hallucinations.  

Increases in Productivity Require Improvements in Processes

When we look at corporate adoption, one big problem is the need for AI to improve processes. Cost reductions emerge when processes are improved, and this requires cooperation between workers, and between them and management. Unfortunately, the above mentioned surveys claim that “much adoption of AI is in secret, as workers use it without telling their bosses” because there is a struggle “under way, as individuals and companies try to work out how best to use the technology. This takes time: investments need to be made, processes rethought and workers retrained.” Even the Wall Street Journal says “[AI] clearly isn’t there yet.”

So even if companies begin to adopt AI, it might take time for the cost reductions to emerge. With investors already getting edgy, will they wait years before the promised productivity improvements and rising AI software revenues emerge. Or will investors lose their nerves long before then? 

The fall in excitement is also worrisome. Many knowledge workers are likely finding that AI doesn’t work as well as the top managers believe. One of the above surveys reports that “desk workers are most concerned that AI’s ability to deal with admin tasks could in fact result in them having more administrative workload to get through each day, not less, all the time with the boss looking on and expecting to see more of AI’s much vaunted productivity gains in evidence.”

Stepping back for a second, one of the reasons that many people like capitalism is that more decisions are made by people closest to the creation and production of products and services, and the solving of problems involved with that creation and production. In the case of AI, knowledge workers are trying to tell us something. They don’t like AI as much the top managers do, and their excitement is falling. This suggests that AI may not be worth the $10 trillion the stock market says it’s worth.