Since OpenAI launched ChatGPT in November 2022, AI has entered the public consciousness and the boardroom discussions of every company. As people realized the capabilities of this technology, the possibilities it could achieve seemed endless.
Large language models (LLMs) that underpin the latest AI models offer a range of capabilities not available in previous iterations, such as the ability to reason and take autonomous actions based on that reasoning
Modern AI takes in data in different forms, such as images, voice, and written text, and understands the meaning of their combination. Agentic AI gives AI the agency to take autonomous action based on its analysis of data, and it is now active. Agentic AI has enabled the previously impossible automation of many activities and entire processes, and is leading to unprecedented technology investments by investors and companies.
The AI investment
Cumulative investment in AI in private and public markets has surpassed a trillion dollars over the last decade. In 2024, venture investments in AI were around $250 billion dollars. That exceeds the total venture investment in all technologies and different sectors of the economy in most years on record.
Many large technology companies—such as Google, Facebook, Amazon, and Microsoft, the so-called hyperscalers—are making massive bets in AI-related areas, including data centers, compute capacity, and data scientists.
This has led to surging valuations for private-market AI companies and market-cap growth for public companies whose business models could benefit from the AI revolution.
Against this backdrop, concerns about an AI bubble entered the global dialogue. What is a bubble? Well, in this case, we are in a bubble if investment levels and the value of AI companies are out of line with their fundamentals. If AI doesn’t deliver on its lofty promises to drive growth and efficiency, current high levels of investment won’t pay off.
In that situation, company valuations will come back down to earth, investment levels will fall, and many companies that have received large investments won’t survive. The ripple effect will lead to lower economic growth or, worse, a recession.
The dotcom era’s experiences
We’ve been through such periods before. When the internet was a new phenomenon, and its promise seemed endless, the era was characterized by huge investments in many companies. When it became clear that most of those companies could not deliver on their lofty promises within a reasonable timeframe, a significant market correction ensued.
The dotcom crash led to a recession in the early 2000s, several years of lower private-market investment in technology, and muted valuations of internet companies in public markets. The fear today is that investors may be overestimating the economic benefits of AI in the short and medium term. If it becomes clear that AI will not deliver those benefits anytime soon, the markets will experience a significant correction.
Productivity accelerates
The real question, then, is whether AI will start delivering on the lofty expectations in a timeframe that avoids what happened in the early 2000s?
The most important way AI could start delivering will be through further productivity gains. If AI helps people get more done, or if agentic AI completes activities autonomously, gains in productivity and economic output may result quickly.
A close examination of the data suggests we’re already seeing some evidence of that. Productivity growth in the nonfarm business sector averaged 1.1% from 2010 to 2019 and around 2% from 2019 to 2024. In 2025, this sector’s productivity seemingly accelerated to near 5% in the third quarter.
The current AI investment boom began in 2023 after the world saw generative AI’s capabilities. The recent acceleration in productivity would be hard to explain without AI as the main driver.
This is in light of the fact that most corporations have yet to meaningfully adopt AI. In a recent Menlo Ventures report, only 9% of US businesses are estimated to have paid commercial licenses for AI applications.
This means that most of the productivity gains from AI are still ahead of us, and most companies have yet to incorporate AI into their products and operations. Also, many AI-native companies, newly formed in the last three years, are growing at unprecedented rates, as I discussed in a previous Fast Company article.
A new era dawns
While the internet proved to be every bit as transformational as people expected, the dotcom crash was caused by many companies with no real business models getting funded simply because they had .com at the end of their names.
This may sound familiar. Now, being an “AI company” gets you funded and at much higher valuations. The key difference between the two eras is that many dotcom companies showed little revenue or growth, whereas this era’s headliners are seeing explosive revenue growth.
OpenAI approached $20 billion in revenue for 2025, and we’re seeing the same dynamics in many other AI companies. History shows that venture-backed company success rates are in the single digits.
However, the productivity growth that characterizes these companies can be expected to eventually diffuse to every sector and company.
Ronald Razmi, MD is the managing director of Zoi Capital.