There’s an adage when trying to unravel a corporate mystery: follow the money. And when you start following the money in connection with the AI boom, you can get a sense of the rumblings of an erupting volcano. Lenders and investors are starting to get nervous, which could jeopardize planned funding.
We have written before about the risk that current power and data center infrastructure can’t support the continued exponential development of AI tools, the fact that the cost of verifying AI outputs is exceeding the savings, and the resulting erosion of trust in historic processes and workflows.
Financial Risk
But there is another perhaps more fundamental risk that may cause the volcano eruption we have been discussing. Thus far, the focus has been on building mega data centers needed for AI to work. These data centers are expensive to build and get permitted. They take five to seven years to build before the power can even be turned on.
And the power and energy capacity needed to make these existing and contemplated centers doesn’t exist. The power plants needed to supply that capacity have to get regulatory approval and be built. That too takes time.
Not only does building the capacity and data centers take time, it takes money. It requires lots of money. That money can come from investors, or it can come from lenders. But because these projects are long-term investments and payoffs, the return won’t come for years.
So anything that happens which makes those long-term returns riskier than originally thought is a problem. If the infrastructure and power aren’t there to support the demand for AI, that long-term return is jeopardized, and they get nervous. If something happens, say flood or hurricane, that delays the completion of the project and the return, they get nervous.
And when investors and lenders get nervous, they begin looking for options like calling the loans or dumping their investments. When that happens, the stock value of the entities goes down. The investors in the entities developing the AI platforms and programs then also get nervous and begin to pull back.
The infrastructure challenges create a cascading financial risk that compounds the angst. It’s a vicious circle. At some point, the projects themselves get jeopardized. That reduces demand, reduces infrastructure building, and the whole house of cards begins to fall. It’s akin to building roads (or as discussed below, railroads) without knowing how much traffic will use them. If the traffic doesn’t materialize, you’re left with expensive and underutilized capacity, capacity that doesn’t provide the expected financial return.
And there are growing reports suggesting that this is exactly what may be beginning.
Increased Risks
The problem is compounded by the FOMO of various investors that’s been going on as the AI hype hit overdrive. Investors and lenders with little expertise and knowledge have waded in when perhaps they shouldn’t have. And if they get nervous, they may be quick to exit.
A recent article makes this very point. In 2025, credit transactions for data centers in the U. S. reached at least $178.5 billion. Major tech players have pushed the total to over $6.57 trillion. And a big piece of the future data center capacity will be built by new players with little data center experience.
We talked to an experienced infrastructure investor, Hector Fornelli, about this very problem. Fornelli’s company, AgilaInvestments, has invested in multiple data center projects, big and small. He says, “There’s too much money going into this. 40% of investments going through Wall Street are going into the AI data center space. At some point it’s going to break. And the reason why it’s going to break is because people are not taking the proper steps to get there.” He also noted that the energy capacity in the U.S. is nowhere near enough to supply the demand for power to run all the planned data centers. But entities are continuing to invest in the centers without considering where the power might come from or whether there are off-meter sources.
Moreover, the data centers generate massive amounts of heat that must be dealt with which is not being considered by the data center planners or their investors. And finally, there needs to be in place agreements ensuring that the output from centers will provide a return. As Fornelli puts it, “If it doesn’t have a clear pathway to those three, it’s not worth investing in because at some point there’s going to be a surplus of data center offers and demand is not going to be as large.” Fornelli believes that’s the kind of due diligence that should be done but is not, sowing the seeds for disaster.
History Lessons
As George Santayana famously said, “Those who cannot remember the past are condemned to repeat it.” That may be the case here as some pundits have already noted.
In the mid-1800s, there was a huge boom in railway investment as investors became enchanted with the railroads. But many of the projects in which they invested were poorly planned or never built. As investors began to see the profits had been vastly overestimated, the railroad stock prices collapsed, and the investments were lost.
A more recent example: in the 1990s, the internet and mobile phones were all the investor rage. They believed demand for data and connectivity would skyrocket, so they spent huge amounts to build the networks that would carry that traffic. But the demand didn’t play out as forecast. That led to the collapse of major players like WorldCom and Global Crossing and the Telecoms Crash in 2001.
Easing Investors’ Angst
But there are solutions that could ease investors’ angst. As we discussed in a previous article, smaller data center and energy projects that are less expensive and don’t take so long to build could be employed.
With smaller projects, the resulting risk to investors is less. Smaller projects can better match demand since if demand falls, the amount at risk for investors is smaller. They can be added incrementally and still be economically valid. If the projects are delayed, the impact is less. The window for something to go wrong is smaller. There is a faster track to the return. They allow for greater flexibility should there be unanticipated risk. Smaller projects are less sensitive to interest rate hikes.
These smaller projects can take the form of smaller data centers powered with generators that depend on existing and underutilized energy sources like combined heat and power (CHP) and behind-the-meter renewable energy sources.
Fornelli agrees that smaller projects may not only be less expensive but useful: “There are other uses for data centers that people are not really paying attention to. There’s medical services. Hospitals, medical labs, pharmaceutical labs and banks that can use AI big time, but they don’t need a 300 megawatt data center. They need a 10 megawatt data center. Or a five megawatt data center. They need to process a lot of information, but the source of the information is one, and the delivery of the information is to one single place.”
Or, for example law firms.
The Impact for Legal
Which brings us to the impact of all this on the legal community. What’s happening right now is a huge rush to adopt AI tools deep within legal workflows and processes. Often this employment is made without considering whether the tools achieve any savings or assistance.
And with that rush and lack of studied approach, that FOMO, comes the risk of overreliance and lack of contingency planning if things go south on the infrastructure or supply side. If the volcano blows, then law firms may be left with unusable and expensive technology. They will have to scramble to figure out how to get work done with which technology was a substantial contributor. In an industry that is deadline driven with little margin for error, that could be catastrophic.
It’s happened before. Perhaps most famously was the collapse of Clearspire in 2010. Clearspire promised a tech- forward virtual law firm model that would be more efficient. But when it shut down due to operational and financial issues, those relying on it had to quickly find other options. In 2010, the practice management company Aderant Expert Sierra was discontinued, forcing customers to migrate to other services. A similar thing happened when Amicus Attorney slowly unwound.
These kinds of migrations can be disruptive to say the least. Imagine the cost and disruptions should there be a significant AI infrastructure failure or if the AI services that depend on that infrastructure didn’t function properly. The Cloudflare outage could be a harbinger of things to come.
What’s a Law Firm to Do?
Warnings like these are often met with a “it can’t happen here” shrug or a “it’s too big to fail” roll of the eyes. But lawyers and legal professionals pride themselves on being skeptical and assessing risks for clients. We need to do the same for ourselves.
That means contingency planning for if (when?) the “what if” actually happens. It means hedging bets on AI to prevent overreliance. It means asking if you’re purchasing AI tools out of FOMO or for valid reasons. As the last one holding on to my BlackBerry long after the cost analysis made no sense, I know what it means to not be tech diligent.
It Means Due Diligence
Talk to any attorney today and they don’t have a clue what their contracts with AI vendors actually say or how financially stable many vendors really are. That’s not the kind of due diligence we engage in for our clients.
Due diligence means assessing vendors’ financial integrity. Can they continue supplying products and services if there is a downturn? It means looking at the debt of and investments in the vendor to determine staying power and robustness. It means looking hard at vendor agreements and indemnity and liability provisions.
It means keeping up with contractual modifications that vendors like to spew out with little notice. It means paying attention to what is going on in financial markets relating to AI and infrastructure.
It means following the money.
Stephen Embry is a lawyer, speaker, blogger, and writer. He publishes TechLaw Crossroads, a blog devoted to the examination of the tension between technology, the law, and the practice of law.
Melissa “Rogo” Rogozinski is an operations-driven executive with more than three decades of experience scaling high-growth legal-tech startups and B2B organizations. A trusted partner to CEOs and founders, Rogo aligns systems, product, marketing, sales, and client success into a unified, performance-focused engine that accelerates organizational maturity. Connect with Rogo on LinkedIn.