Big Tech plans $720 billion in AI spending despite unproven returns

The massive cost of AI

The four largest competitors in the artificial intelligence race are preparing for a massive spending spree. Amazon, Microsoft, Alphabet, and Meta told investors they plan to spend at least $720 billion on capital expenditures this year alone. The majority of this money is earmarked for building out the data center infrastructure required to power AI models.

These companies are also looking far into the future, with commitments totaling at least $1 trillion for the coming years. While this spending recently triggered a sharp selloff in tech stocks, the companies have signaled they will not slow down. They view the current pace of technological advancement as a unique window of opportunity that cannot be delayed.

The scale of the buildout is so large that it has begun to spook Wall Street. Investors are starting to question when these massive investments will finally yield significant returns. Despite the high costs, the companies remain committed to the plan, stating that the opportunity for growth is happening now rather than over several decades.

Revenue and profitability gaps

While top-line revenue is growing for these tech giants, it is not always clear if that growth justifies the AI costs. For example, Alphabet acknowledged that its new AI-enabled search queries have actually impacted the monetization of its core Search business. This creates a paradox where the company's primary source of profit is being cannibalized by the very technology it is trying to build.

Google Cloud's revenue more than doubled in a year, but the growth in AI Mode queries has come with lower monetization. Alphabet's search ad revenue grew 12% in the third quarter, which is a notable drop from the 17% growth rate seen in the same period last year. The company is still experimenting with how to effectively turn AI interactions into a profitable business model.

This trend is also visible in the startup space, where companies like Anthropic are facing high costs. Anthropic reported that its computing costs reached $17.5 billion in the third quarter, slightly exceeding its $17 billion in revenue. Even with a 1,000% increase in revenue from the previous year, the company is still struggling to cover the expenses of running its models.

Funding the infrastructure

To keep up with the demand for compute power, some of the biggest players are turning to debt. Microsoft recently broke news of a $25 billion, 10-part bond sale to fund its AI buildout. Until recently, Microsoft was the only major company that had not taken on debt for these projects, but the scale of the requirements is forcing a change in strategy.

Other companies are also finding creative ways to manage their infrastructure needs. Anthropic has committed nearly $900 billion to future compute and infrastructure costs and plans to build its own data centers. Some of that spending will go to Google, which is selling its own TPU chips to the startup to help it scale.

The logistics of these deals are complex because many of these contracts do not represent immediate cash. Google is currently fronting the costs of building the hardware it sells to other labs like Anthropic. These financial obligations will only come due once the hardware is ready for use, meaning the spending is a long-term commitment rather than an immediate expense.

What comes next for Big Tech

The future of this spending depends on both market performance and potential regulation. Some analysts suggest that new government restrictions or moratoriums on data centers could slow the pace of construction. However, the tech companies themselves have not mentioned the upcoming midterm elections as a factor in their current spending plans.

For the companies involved, the goal is to achieve better compute efficiency to make their programs profitable. Meta, for instance, expects its compute efficiency to increase tenfold next year, a goal that requires significant ongoing investment. They are betting that they can eventually drive down the costs of running these complex models.

Ultimately, the AI buildout appears to be a race that will continue until external forces intervene. Whether through new laws or a shift in investor patience, the current trajectory shows a total commitment to infrastructure. For now, the industry is moving forward with the belief that the opportunity for growth is too large to pass up.