No One is Immune to Economic Gravity, Not Even AI
Microsoft’s reported slowdown in data center investments raises more than just eyebrows – it raises deeper questions on whether market leaders will now switch to strategic prudence in a world where everyone wants a bite of the cherry that is the AI gold rush
The artificial intelligence revolution is well and truly here. Every consultant, every tech guru, and every starry-eyed venture capitalist out there in the past few months has gone from telling us that Large Language Models (LLMs) and generative AI are not just the next big thing, but everything. People are getting seriously rich, at least on paper.
Equity prices, particularly of the hardware providers, have touched the moon (and then come back slightly on a slight slowdown earlier this year culminating into one of the biggest stock market routs of modern economic history following U.S. President Donald Trump’s Liberation Day tariffs). NVidia’s price-to-earnings ratio (a widely-used valuation metric), for example, is down from an absurd 246.8 in mid-2023 to the early 30s as of early-April, still markedly higher than the S&P500’s already-elevated mid-20s range. Yet, there seems to be a strange wrinkle in the narrative.
Microsoft – synonymous with the current AI boom due to their close association with OpenAI – seems to be tempering their optimism, widely reported to be slowing down, even backing away from a few already-greenlit data center expansion plans. Data centers are the lifeblood driving large models, and the Satya Nadella-led global tech behemoth catching caution back from the wind may be telling us something deeper about the industry as a whole.
AI Hype Train Cresting Too Soon?
Right off the bat – the AI dream doesn’t come cheap. Microsoft was reported to be looking at a nearly $80 billion worth capital expenditure for data centers in a single fiscal year – a figure that ought to sober up even the most enthusiastic techno-optimist. That kind of money can build actual cities, let alone server farms humming away in anonymous buildings.
Building these digital cathedrals takes time, concrete, steel and, crucially, skilled labor and reliable supply chains. Add on to that the likely supply chain disruptions, cost pressures and job losses the newly imposed tariffs will create, it seems the physical world insists on intruding upon the infinity of our digital fantasies.
So, is the slowdown in Microsoft’s data center investment down to some shrewd strategic forward thinking in anticipation of incoming tariffs upending the global macroeconomic backdrop and the will-they-won’t-they recession dance, or is it saying something deeper about the intrinsic value added by AI to business? It’s probably a bit of both.
While it is hardly the climate for jaw-dropping unrestrained speculative investment, Nadella’s skepticism about some of the more grandiose claims swirling around AI have raised more than just eyebrows. The CEO of one of the torchbearers of AI dismissing certain milestones as ‘nonsensical benchmark hacking’ can hardly be interpreted as cheerleading.
It suggests, perhaps, an internal awareness the hype cycle might be cresting slightly ahead of the actual, economically verifiable value creation. This isn’t to say AI isn’t transformative. It likely is, in ways businesses probably haven’t fully grasped yet. The cycle seems very much like the early internet – lots of excitement, lots of crazy valuations followed by the dot-com bubble bursting and a slower, steadier build-out of real value. Nadella’s apparent realism might simply be acknowledging that revolutionary technologies often have messy, non-linear adoption curves. Throwing more silicon at the problem doesn’t automatically conjure sustainable revenue streams.
Several Wall Street analysts, like those at TD Cowen, have also flagged a potential slowdown. Widespread investor nervousness on whether the AI-fueled stock rally has legs, coupled with a possible oversupply of data center capacity in the near-term, have tempered the outlook somewhat. After all, a frenzy of construction driven by FOMO should hardly end with vast, echoing halls of servers waiting for workloads that haven’t arrived yet. It wouldn’t be the first time irrational exuberance led to excess capacity – ask anyone who invested in fiber optic cables back in 1999.
Is Microsoft veering away from AI? No. Are they going to continue their hell-for-leather approach throwing money at anything and everything remotely AI? Almost certainly not.
Microsoft’s choice is more likely a recalibration – a well-thought out pause to bring back reason into the picture and re-align spending with a more realistic assessment of demand, costs and the generally dark macro climate. As they pointed out themselves, plans change. But this evolution is significant, suggesting even the biggest players, armed with mountains of cash and a mid-sized army of pretty smart people, are not immune to the fundamental laws of economic gravity.
Microsoft’s second-thoughts, however tiny in the grand scheme of things, will mark a valuable data point in the annals of AI history. While the relentless hype train, fueled by futuristic promises and dazzling unveilings may well resume its journey soon, it is acknowledgement of the fact that the track ahead is likely to be steeper, the construction costs higher and the ultimate destination further than most breathless predictions would have us believe. Algorithms will continue to learn exponentially, but the physical world that houses them will be built slower. A little dose of reality amidst the breathless frenzy might be exactly what the market, as a whole, needs.