The AI investment fervor is not showing any signs of abating. Between 2022 and 2025, VCs have funneled almost $32B into AI startups, in spite of disappointing results with high-profile bets. Since mid-2024, there has been a combined $52.4 billion worth of deals. Andreessen Horowitz is reportedly attempting to raise a $20B mega-fund to invest even more heavily in AI. 

Some VCs have even redoubled investments in companies, deepening their exposure and potentially leading to inflated valuations. Much of the increase in investment went to companies developing industry-leading models like OpenAI and Anthropic. There is already overcrowding in some niches, like AI agents and model evaluation.

The Economist warns that AI valuations “are verging on the unhinged.” A prime example being Thinking Machines Lab, a startup led by former OpenAI CTO Mira Murati. Less than five months after its founding, the company is valued at $10 billion, and it has yet to establish exactly what it will do. Investors are also giving Murati unusual preferential terms, including a board vote that holds the same weight as all the other board directors’ votes plus one.

Foundation model developers are still struggling to turn a profit. xAI is burning through a billion dollars a month, with expected losses of $13 billion in 2025. The company is trying to raise $9.3 billion in debt and equity to cover the shortfall as it struggles to develop revenue streams at the same rate as its competitors. But even OpenAI lost $5 billion last year, despite being the clear market leader. OpenAI is now valued at $500 billion, and has increased its projected cash burn from $1 billion in 2025 to $8 billion. This comes on the heels of the release of Chat GPT-5 being met with widespread disappointment.

The willingness of investors to look past these losses signals a belief that there is a large untapped market for AI and that costs will continue to plummet, but there are reasons to question both premises. A recent MIT study revealed that 95% of generative AI pilots failed to achieve rapid revenue acceleration for businesses and delivered little to no measurable impact. Many companies are looking for ways to make AI more useful, not more powerful models. And costs are not plummeting. With models doing more ‘thinking,’ AI is getting more expensive. 

Apollo Global Management’s chief economist is warning that AI stocks are currently even more over-valued than dot-com stocks were in 1999 and there could be serious market damage if the bubble pops. He points out that the price-to-earnings ratio of the top 10 companies in the S&P 500 today are even more inflated than they were in the 90s.