The AI Valuation Paradox: Balancing Hype With Real-World Impact

Forbes

09-04-2025

Tomas Milar is the Founder and CEO of Eqvista, an equity management platform.
For the past year, soaring artificial intelligence (AI) startup valuations have been justified by rapid revenue growth, driven by various industries recognizing AI's potential to reshape operations and enhance productivity.
A prime example is OpenAI's valuation, which grew more than tenfold in just three years, from $14 billion in 2021 to $157 billion in 2024, fueled by ChatGPT's success and its impressive projected earnings. The market's confidence in AI is evident in the lofty average revenue multiple of 23.4x commanded by AI startups.
However, we may soon witness a decline in the high funding levels AI startups currently attract, driven by the rise of low-cost, asset-light alternatives. While this in itself is a strong reason for AI startup valuations to deflate, I believe the exaggeration of current AI capabilities leaves room for further corrections.
Recently, we have seen AI startups secure valuations that were thousands of times their annual revenues. For example, xAI and Infinite Reality were valued at $40 billion and $12.25 billion, respectively.
Even considering the growth potential of AI startups, such valuation multiples are excessive. Such outliers can skew data that most AI companies can achieve such heights when the reality is that many more AI startups tend to close their doors before achieving such market success.
AI startups have distinguished themselves from their predecessors with an unprecedented ability to generate revenue. According to Stripe (paywall), today's leading AI startups that have reached an annualized revenue of $30 million have done so five times faster than past SaaS companies.
At the same time, we must acknowledge that AI startups are much more capital-hungry than other tech startups.
OpenAI faces significant operational costs from its flagship product, ChatGPT, spending approximately $700,000 daily (paywall)—over $255 million annually. While these operational costs are offset comfortably by its $3.6 billion annualized revenue (paywall), OpenAI faces intense competition from tech giants such as Google as well as emerging players such as Anthropic.
To maintain its competitive edge, OpenAI must spend an additional $5 billion annually to train new models. This is an expense that will likely continue until OpenAI establishes itself as the undisputed market leader. To put things into perspective, OpenAI's total funds raised stand at $21.9 billion (registration required).
However, recent advancements by new entrants and the limitations of existing AI models cast doubt on both the funding needs and valuations of AI startups.
DeepSeek, the Chinese AI startup, has disrupted the U.S. AI startup ecosystem by demonstrating that premier AI models could be built without exorbitant capital expenditure. Although various experts are disputing this, the company claims that the total training cost was $5.6 million for DeepSeek-R1, the model that delivers performance comparable to OpenAI's ChatGPT.
When we compare the training costs for the two startups, we can see that OpenAI could train new models for less than half a day with DeepSeek's entire budget.
We are already seeing the AI leaders being challenged. After the release of DeepSeek-R1, between January 23 and 25, ChatGPT lost 41.3 million views.
Thus, some investors are questioning if high-performing AI models really cost as much as advertised.
The reasons to believe that AI startups are overvalued are plentiful. Firstly, we haven't yet achieved true artificial general intelligence (AGI), which by definition is capable of performing any intellectual task a human can. What we have right now is a very narrow version of AI that can reliably carry out certain tasks, such as natural language processing or image recognition, but has limited application elsewhere.
Secondly, AI's commercial viability remains questionable. A Boston Consulting Group (BCG) report analyzing 1,000 companies that adopted AI found that only 4% generated substantial value, while only 22% had progressed beyond the proof-of-concept stage to generate any value at all. Notably, the companies that stood out were already well-positioned for success due to strong nonfinancial factors, such as patents filed and employee satisfaction.
Thirdly, various studies note that the capabilities of AI in tasks such as logical reasoning, chemical compound discovery and code writing have been exaggerated.
Thus, only a few AI startups that achieve significant breakthroughs, such as closing the gap between advertised and actual capabilities and enhancing commercial viability, are likely to survive and justify their valuations, while the majority perish.
While some AI startups' values are astronomical multiples of their annual revenues, these cases represent a small group of outliers. Once we exclude the outliers, we can observe reasonable valuation multiples across all stages.
However, a widely recognized cause for concern for AI startups is their struggle to achieve profitability due to their asset-heavy nature and the high costs associated with operations and training.
Additionally, AI's real-world impact remains limited, with narrow applications, questionable commercial viability and sometimes-exaggerated capabilities. As low-cost alternatives emerge, investors are increasingly scrutinizing whether U.S. AI startups can maintain their competitive edge.
The information provided here is not investment, tax or financial advice. You should consult with a licensed professional for advice concerning your specific situation.
Forbes Finance Council is an invitation-only organization for executives in successful accounting, financial planning and wealth management firms. Do I qualify?