Big Tech is wagering big (and getting bigger) on an artificial intelligence (AI) infrastructure bet; and the bet is worth $725 billion. How much will the infrastructure support? We do not know yet.
During last week's quarterly earnings reports, there was a notable number highlighted: According to Financial Times' research, four of America's largest tech firms (Amazon, Google, Microsoft, and Meta) are expected to commit nearly $725 billion to developing their AI capacity in 2026. This represents an increase of approximately 77% over the record amount spent by the same four companies last year. Most of this money will go toward building out data center space and designing and purchasing high-performance chips needed to run large-scale generative AI models.
This level of spending indicates a significant transition from experimenting with AI to making AI one of the top priorities for each of these companies. In addition, the sheer volume of spending has generated harsh criticism from researchers questioning whether the technological underpinnings are sufficient enough to warrant such vast amounts of capital being committed. In addition to questions surrounding technical justifications, these companies are also facing growing regulatory pressures and little clarity in terms of potential profit margins.
According to Business Insider, Microsoft currently tops all four companies in terms of projected capital commitments for 2026 at an estimated $190 billion, followed by Meta which increased its estimate to $145 billion. The company's cloud services division experienced some of the highest revenue growth during first quarter (Q1), and it should help provide justification for the substantial amount of money Google will commit to AI spending next year, according to The Irish Times.
Get the latest model rankings, product launches, and evaluation insights delivered to your inbox.
When combined with Tesla's revised budget estimate, the total amount of money expected to be committed to AI spending is expected to exceed $750 billion by the end of 2026. To put that number in perspective, the combined GDP of most countries is lower than that number.
The magnitude of the spending commitment is indicative of two things. First, hardware costs related to building out AI infrastructure continue to rise. Second, it takes a tremendous amount of computing power to operate large-scale AI systems. Jon Markman, writing for Forbes, recently noted that the efficiency with which graphics processing units (GPUs) can communicate with one another inside a data center is now the limiting factor when it comes to deploying AI compute. What used to be a problem centered around compute speed has now become a problem centered around connectivity.
Mark Zuckerberg stated during his recent presentation defending the billions being invested in AI at Meta that the company needs time to develop safe AI models including text-based models (such as Muse Spark), images, and videos before they could deploy them. However, it seems contradictory to invest so quickly at such a break-neck pace, and then claim you need time to be safe. AI researcher Gary Marcus described the companies' capital expenditure plans as "the greatest capital misinvestment in history" and warned that the massive spending levels on AI did not create sufficient barriers to entry and were not producing significant profit margins, according to Business Insider.
There exists a similarity between the defensive rhetoric used by Meta concerning safety and its rapid spending pace on AI with past tech cycles in which companies made claims about careful development while attempting to capture as much market share as possible. Furthermore, Meta faces additional challenges with regards to social media safety regulations in both the European Union and United Kingdom. Even as Meta continues to pour billions into its AI infrastructure, it faces increasing regulatory hurdles across Europe.