When the music stops: the unravelling of AI companies’ flawed valuations

Public and private market investors have indiscriminately been attributing huge premiums to AI companies, and the party continues — the music is still playing and people are still dancing. But when the music stops, investors will scramble for assets in a frantic game of musical chairs. 

We have seen this all before. As with the dot com bubble, the long-term potential of the technology is massive; the over-valuations are just part of the initial hype. Right now, we’re already slipping into the Trough of Disillusionment and when the AI bubble bursts, the flawed valuations will be exposed. 

AI Companies’ Unhealthy Entanglements

Among the largest AI companies, there has been a pattern of suppliers investing in customers and the issue of circular investing is widespread. With Nvidia investing in OpenAI, for example, Nvidia, directly or indirectly, supplies OpenAI with GPUs [graphics processing units] and the investment enables OpenAI to buy more. A similar pattern can be found with Nvidia investing in Coreweave, which buys GPUs from Nvidia and sells GPU capacity to Open AI and other LLM (Large Language Model) providers. 

Meanwhile major cloud providers have invested in OpenAI and Anthropic, and they sell them cloud compute for training and inference. The pattern repeats downstream where OpenAI and other LLM providers can invest in other companies that can build their applications on OpenAI’s ChatGPT. 

There are risks with these entangled partnerships, and it is unhealthy for the ecosystem at large. Although the largest companies can absorb a drop in valuations, other AI companies have been dragged along by the hype, especially in the private markets. 

The bar has already been set high, for example with Lovable’s $330m Series B at a valuation of $6.6 billion and Mistral AI’s €1.7 billion Series C at a €11.7 billion valuation. While there is nothing problematic with these individual company valuations per se, it sets unrealistic expectations for investors and entrepreneurs. Generating a 10x return on a Series B investment at a $6.6 billion valuation requires selling the company at $66 billion. To put that into context, there are less than 10 public cloud software companies with a market cap of more than $60 billion! 

Undoubtedly there will be generational businesses that are created as GenAI is widely adopted that can achieve such valuations, but it is problematic if investors value every start-up as a breakthrough company when 99% of them simply aren’t. 

Annual Recurring Revenue (ARR), But Not As We Know It

Another flaw in AI company valuations has been the reporting of ARR. This is the biggest driver of value for software companies, but it isn’t what it used to be. Previously, it would have been based on an accrual of subscriptions, but now there is a hotchpotch of other measures included, such as one-off, volume-based, performance-based, and value-based contracts, which are much less predictable.

Beyond the structure of contracts there are other qualitative aspects to look out for. Short sales cycles and short implementation times have been a boon to AI companies’ growth. The quick sales, however, may be due to hype and excitement and a target market of individuals (and not enterprise business customers) who may be less likely to renew. And the short implementation times could indicate that the technology is not making any meaningful improvement to productivity.

Gross Margins are Being Squeezed Beyond Repair

Many AI companies are operating with high costs, and revenue forecasts have been overly optimistic. Many are running at very low or even negative margins because of the high costs of developing, training, and maintaining AI models. Traditional software companies typically have gross margins in the 70%-80% range. Meanwhile, many AI companies have true gross margins in the low teens to twenties. 

One solution is to grow themselves rapidly out of this situation and hope that the model costs will drop quickly enough to overcome the poor margins. So far, however, we haven’t seen this materialize and the high costs aren’t coming down.

To Conclude: Focus on the Fundamentals, Not the Music

Amid the exuberant valuations of AI companies, there is a need to focus on fundamentals. Current expectations are excessive, and the party is getting out of hand. Away from the irrational exuberance, there are still sensible investments to be made. There is a massive opportunity to invest in B2B software companies that are targeting the replacement of incumbent enterprise solutions based on AI-native capabilities. 

There is an even bigger opportunity for agentic AI software companies to automate large parts of the existing professional services market, which is at least 10x as big as the current market for cloud software. These are the companies that will be impacting productivity and transforming business processes that are currently dominated by manual work. These solutions need the buy-in of multiple stakeholders, which means the sales process is more gruelling. But in the long run there will be stickier, more sustainable, recurring revenue. 

Just like in the cloud era, most of these companies won’t become $100 billion businesses, or even $10 billion businesses, but that is just fine as long as investments are based on realistic expectations and sensible entry valuations there is an unprecedented opportunity for investors to create long-term sustainable value here. 

The opinions expressed in Fortune.com commentary pieces are solely the views of their authors and do not necessarily reflect the opinions and beliefs of Fortune.

This story was originally featured on Fortune.com