The AI Buildup Is Massive. Will It Ever Pay Off? November 4, 2025 by Alex Woodie 
We’re in the middle of an AI boom that is unlike anything we’ve ever witnessed. Through the end of the decade, trillions of dollars will be spent to build and outfit thousands of new data centers to power Gen AI chatbots, co-pilots, and autonomous agents. It’s a massive technological gamble, but will it pay off in the end?
The numbers behind the AI buildup are truly astonishing. While last week’s Department of Energy announcements around the nine new supercomputers showcase HPC leadership, they’re dwarfed by the investments hyperscalers and their financial backers are making in building gigawatt data centers, or “AI factories,” as Nvidia Jensen Huang calls them.
HSBC estimated recently that $2 trillion has been planned for building AI data centers, while Citigroup says $2.8 trillion could be spent through the end of 2029. ABI Research estimates that nearly 2,300 new data centers will be built around the world by 2030. JLL, a real-estate services firm, estimates that $60 billion was spent on data center construction in 2024 and that data center construction is growing at a 15% annual rate, but that’s not enough to keep up with surging demand.
(Image courtesy Michael Thomas/Distilled)
New AI data centers are being announced all the time. Consider:
- Samsung just announced that it’s building a new AI factory that will house 50,000 Nvidia The announcement, made October 31, will be used to create a real-time digital twin of a Samsung fab. The goal is to speed up operational planning, anomaly detection, and logistics optimization in pursuit of building “a fully autonomous fab.”
- Digital Realty is partnering with Nvidia to build a 96 megawatt AI factory in northern Virgina. Dubbed the Aurora AI Factory, the data center will feature software from Emerald AI that will enable it to throttle energy usage back in a short amount of time in response to market changes. It’s currently under construction and is due to open in 2026.
- The Stargate alliance, led by Oracle and OpenAI, recently announced its latest data center, which will be built on a 250-acre site outside of Detroit, Michigan. The data center, which is being built by Related Companies, is reported to cost $7 billion to build.
- Google announced in early October that it’s spending $4 billion building a new data center in West Memphis, Arkansas, on a 1,000-acre site.
- Facebook parent Meta revealed in October that it’s actually planning to spend $27 billion building a 2-gigawatt data center dubbed “Hyperion” on a 2,250-acre site in Richland Parish, Louisiana, up from $10 billion when it was originally announced in late 2024.
- AWS last week powered up Project Ranier, the new $11-billion data center built on a 1,200-acre site in New Carlisle, Indiana, that will serve to train Anthropic AI models. AWS CEO Matt Garman said it went from “cornfields to data centers, almost overnight.”
These data centers will be filled with the latest technology, including millions of GPUs, XPUs, and other AI accelerators; tens of millions of CPUs; millions of miles of networking cables; tens of millions of SSDs.
Meta’s new Hyperion data center overlaid on a map of Manhattan, New York (Image courtesy Meta)
Nvidia has been the single largest beneficiary of the AI boom, becoming the world’s first $5 trillion company (by market capitalization), greater than the GDP of every country in the world besides the U.S. and China. Nvidia and six other tech firms–Apple, Amazon, Alphabet, Meta, Microsoft, and Tesla, or the “Magnificent Seven”–have a collective market capitalization of $22 trillion.
Powering all these new data centers is becoming a major burden, in part because of a lack of new electricity sources and grid investments by utilities. As the compute density and power consumption goes up, cooling becomes an even greater issue, which is driving more investment and innovation in the HPC sector. (Check out BigDATAwire Editor Ali Azhar’s series, “Powering Data in the Age of AI,” for more on power and cooling issues.)
The big tech firms have been among the biggest investors in AI, which can be seen with reciprocal deals involving chipmakers Nvidia, AMD, and Intel making or receiving investments in OpenAI, Anthropic, and others. How will this all play out? Nobody knows, but one goal executives have for AI is to take over repetitive white collar work currently being done by people. AWS, which has been an enthusiastic developer and adopter of AI tech, recently laid off 14,000 employees and hinted that it could eventually give 30,000 people worldwide their notices. Amazon’s stock went up 13% after announcing the layoffs, along with improved financial results.
The goal isn’t necessarily to cut jobs, but to boost productivity. OpenAI sent a letter to the White House last week claiming that a $1 trillion investment in AI could boost the country’s GDP by 5% over a three-year period. The estimate was based on an “internal analysis,” the company said.
While white collar workers will feel the pressure to retrain or lose their jobs, the AI boom is leading to an increase in demand from one unexpected segment: blue collar workers.
Artist’s rendering of Meta’s new Hyperion data center
“The country will need many more electricians, mechanics, metal and ironworkers, carpenters, plumbers and other construction trade workers than we currently have,” OpenAI stated in its letter to the White House. “Americans will have new opportunities to train into these jobs and gain valuable, portable expertise.”
This echoed recent statements by Nvidia CEO Jensen Huang. “If you’re an electrician, you’re a plumber, a carpenter–we’re going to need hundreds of thousands of them to build all of these factories,” Huang told Channel 4 News in the U.K. in September.
But is the AI buildout sustainable? That will depend on whether the companies investing the most money are getting a good return on their investment. We know that companies of all sizes are struggling with AI, as shown by the recent MIT report that found 95% of AI projects never get out of proof-of-concept. But what about the hyperscalers, who are the ones driving the big AI spending spree? Let’s consider how the market responded to the recent AI results of Google and Meta.
Google parent Alphabet, which expects to spend $93 billion on capital expenditures (largely AI) this year, saw its stock price rise 16% after it reported its third-quarter results last week. The big takeaway: it’s finding success selling basic AI services, such as AI-powered search and subscriptions to its Gemini app, which has more than 650 million monthly users. So while AI spending is high, it appears it could be contributing to sales growth.
Facebook parent Meta, on the other hand, has also made huge investments in AI, but it’s not seeing the same approval from the market. The company recently announced it would “aggressively” increase its capital expenditures–up to an estimated $97 billion this year–in order to avoid falling behind in AI.
“That way, if superintelligence arrives sooner, we will be ideally positioned for a generational paradigm shift in many large opportunities,” Meta CEO Mark Zuckerberg said on a call with investors. “If it takes longer, then we’ll use the extra compute to accelerate our core business.” Meta’s share price decreased 7% following the news.
If a broader array of companies investing in AI show small but measurable results like Google’s, then investors will likely keep the spigot turned on. However, if experiences like Meta’s “spray and pray” strategy becomes the norm, then it’s likely that investors will question the spending, and retrench.
Related
Alex Woodie has written about IT as a technology journalist for more than a decade. He brings extensive experience from the IBM midrange marketplace, including topics such as servers, ERP applications, programming, databases, security, high availability, storage, business intelligence, cloud, and mobile enablement. He resides in the San Diego area.
