- Tech luminaries have been aggressively lobbying the U.S. Government to make massive investments in data centers
- Chris Lehane, predicted that the U.S. would need to invest $175 billion in AI infrastructure
- President Trump's executive order aims to accelerate federal permitting and provide financial incentives for large-scale data-center builds
Amid the dotcom bubble of late 1990s, a popular assumption took hold that the internet’s pipes were vastly unprepared for the massive consumption demand that was about to ensue.
Internet traffic would soon double every 100 days, a source no less credible than the U.S. Commerce Department declared. Soon, millions of miles of fiber-optic cable were dug into the streets of American cities and small towns, and sunk down onto the transoceanic paths of seabeds.
But the data was “wildly optimistic,” in the words of the Wall Street Journal. By 2002, only 2.7% of the fiber put down in the preceding few years was being used.
Corning, the world’s largest producer of optical fiber, soon saw its stock crash from nearly $100 in 2000 to about $1 by 2002. The so-called “fiber glut” led to what was, at the time, the biggest bankruptcy ever, WorldComm, the fate of which was sealed by associated accounting scandals. Plenty of other companies — Global Crossing, Winstar Communications, Metromedia Fiber Network, just to name a few — also went into Chapter 11 restructuring or liquidation.
Today, many folks in technology, media and telecom wonder if a similar dynamic is unfolding today in regard to AI and the associated data-center expansion.
Tech luminaries as powerful as OpenAI co-founder Sam Altman have been aggressively lobbying the U.S. Government, among other entities, to make massive investments in data centers they say are needed to supply a coming tsunami-wave of demand from AI companies and consumers.
OpenAI’s VP of global affairs, former Clinton administration special assistant Chris Lehane, predicted that the U.S. would need to invest $175 billion in AI infrastructure to underwrite its participation in the technology of the future.
It’s unclear as to whether the nation will be able to produce enough electrical power to based on his preferred source, fossil fuels, to support these ambitions. But on July 23, President Donald Trump issued an executive order intended accelerate federal permitting and provide financial incentives for large-scale data-center builds. It also gives these companies preferential access to limited municipal power supplies. This came soon after Trump met with executives from OpenAI, Meta and Microsoft.
Some credible glut factors
But Trump’s executive order also comes as members of the global telecom and tech community have started worrying about a data center glut. There are a number of credible factors fueling this concern.
For one, China is already facing an over-supply of AI computing power. With utilization rates reportedly at only around 20-30%, China has set up a national cloud service to sell excess compute capacity. Local governments in China have also reportedly cancelled more than 100 data-center projects in the last 18 months.
As Fierce Networks reported in July, the data center market is getting further crowded by companies which set up infrastructure to mine cryptocurrency, and which are now pivoting to supply computational resources for hyperscale and enterprise clients.
In many cases, these crypto operators have a key advantage over incumbent data centers — they have better access to electrical power.
Meanwhile, “Tariffs introduce another layer of complexity,” wrote MoffettNathanson analyst Nick Del Deo in an April report, noting that the import taxes are driving up construction costs significantly.
Then there’s a more existential demand threat — AI workloads seem to be becoming less compute-intensive than originally anticipated. The term “DeepSeek Effect” has emerged in the global tech sector after the eponymous Chinese startup demonstrated an AI model with OpenAI-comparable performance using far less resources.
Perhaps the most formidable market indicator so far was Microsoft’s decision, announced in April, to “slow or pause” a number of data-center projects, including a $1 billion project earmarked for Ohio.
OpenAI has filled the investment void left by Microsoft in a number of cases. And since OpenAI’s business is significantly intertwined with Microsoft’s, it seems that some of this shift has to do more with a reorganization of the tech giants’ relationship verses a significant no-confidence vote in the future of AI computational demand.
Recent signals indicate demand
So where will the global data center market be in, say, three years?
With the Trump administration aggressively seeking to marginalize clean energy sources, that’s not an easy question to answer.
“Nobody has any idea what AI electricity usage in data centers is going to be in three to four years,” says Jonathan Koomey, a researcher and consultant who studies the energy impact of internet and information technology, told the New Republic back in December.
Notably, despite all the aforementioned factors, MoffettNathanson remains bullish on U.S. data center and colocation providers Digital Reality and Equinix.
“Recent signals (admittedly from before the tariffs fiasco) suggest that data center demand remains robust, with key players highlighting continuing needs for capacity and/or taking on projects that Microsoft stepped away from,” Del Deo wrote. “And the secular trends that have been powering the space — AI, cloud adoptions, digitization — are long term in nature. We expect the gap between supply and demand to eventually narrow, but it doesn’t seem like that time is now.”