Good afternoon,

As always, we hope you are well.

We've written previously about institutional momentum and resistance to change, and I suspect it is a theme which will remain relevant for as long as human beings (and our organizational structures) exist. The integration of AI into everyday business life is proving to be another great example of this reality.

The Headlines vs. The Reality

A recent report out of MIT noted that as many as 95% of corporate AI pilot programs are failing to create value. Media outlets everywhere quickly distributed this news as loudly and broadly as possible because the prior sentence makes for sensational attention-grabbing headlines these days. But, if you dig a little deeper, the reason many of the pilot programs appear to be failing may have more to do with human nature than with the technology itself.

Forbes' coverage of the report does an excellent job of outlining the real takeaways beyond those headlines. It turns out businesses are still figuring out how to use this technology, and where to apply it within their organizations most effectively. The "obvious" use cases may actually be the wrong ones. At least for now. It likely won't be quick or easy, but it's reasonable to expect as leaders gain a better understanding of the strengths and weaknesses of various AI tools, those (perfectly useful and viable tools) will be more thoughtfully and deliberately applied. And the leaders and the tools themselves will both continue to improve every single day.

Market Reactions and What They Mean

After a couple years of an unyieldingly positive narrative and the accompanying investment bonanza in computer chip and equipment makers, data centers, and the energy sector, this MIT report (and the related headlines and soundbites) was enough to cause everyone to pause and take a collective breath. Relevant stocks have seemingly found some intermediate resistance, and the world is currently awaiting Nvidia's latest earnings release later today to see if the world's most valuable company can somehow push expectations even higher. Meanwhile, the U.S. Federal Reserve has recently been signaling interest rate cuts later this year are now a near certainty, but that was still not enough to offset some of the recent investor caution. Reuters' coverage of some of these recent happenings is worth a glance for those interested.

So, what does it all mean? Are we in an AI bubble which only the skeptics can see?

We Don't Think So

Sure, there could be some significant stock market volatility with valuations stretched, and not every single AI and data center company out there is doing the right thing from a business perspective. Many startups will fail. Many corporations will try to adopt AI as an enterprise-wide directive and fall short once or twice before they get it right. Others will fail and abandon the effort, only to have to come back around years later after their peers and competitors eventually prove the benefits. There will be foot-faults and setbacks, both for individual companies and for entire segments of the space.

That's what disruption and innovation look like.

But that disruption is going to bring with it massive positive change (and probably plenty of negatives too).

The 30-Year View

The evolution of the internet in the 30 years from 1995 to 2025 brought considerable turmoil, change, and opportunity. The integration of AI functionality into everyday life will likely be equally turbulent and dynamic, if not more so. But the technology is here today, it's not going away, and it works...at least for those willing to learn how to use it. It's evolving quickly to better meet our needs, but it's also up to us to evolve with it.

And underlying all of it is the physical power and data center infrastructure that needs to be planned, designed, sited, financed, manufactured, constructed, and operated.

Looking forward over a similar 30-year horizon, we think it's reasonable to assume there is going to be practically no letup in the pace of development of these long-lived physical assets.

Infrastructure Moves at a Different Pace

It takes a very long time to create these facilities and they will serve the needs of the high-tech computing equipment occupying them for a very long time to come, regardless of the quarter-to-quarter or even year-to-year ebbs and flows in the value of servers, GPUs, and compute hours. This infrastructure development process is like a massive freight train gathering speed…it's highly unlikely to slow down for short term fluctuations in perception of that 30-year trajectory.

So, if you're suddenly worried these data centers won't get built and your energy projects won't be needed because companies haven't quite yet learned how to best use their shiny new AI toys, don't be. They'll figure it out, and when they do, they won't be able to survive without them.

The Bottom Line

The physical infrastructure required to support AI computing is not a speculative bet—it's a fundamental requirement that will persist across decades. While companies learn to optimize their use of AI tools, the need for power and data center capacity will only grow.

This is why we remain focused on the intersection of power and computing infrastructure. The technology may evolve, the use cases may shift, but the underlying need for efficient, reliable, power-optimized data center infrastructure is here to stay.

That's it for this month. Thank you all once again for continuing to follow these monthly updates and we'll look forward to speaking with each of you as the year progresses. Please don't hesitate to reach out if you have questions about how to achieve your goals at the intersection of power and computing.

For those in the U.S. and Canada, an outstanding Labor Day weekend ahead!

Andrew Webber is Chairman of Greenlight Data Centers and CEO of Digital Power Optimization.