When an executive says something like “we continue to see massive opportunity ahead” during a major tech company’s earnings call, the analysts on the other end of the line write it down, nod, and then return to their spreadsheets to try and figure out when exactly that opportunity will show up in the numbers. In a single quarter, Microsoft spent $34.9 billion on AI infrastructure and data centers. In 2025 alone, Meta has committed up to $72 billion, and its CEO has publicly acknowledged that the company is operating its core business in a “compute-starved state.” Alphabet increased its projected yearly expenditure to between $91 and $93 billion, almost twice as much as it had stated for 2024. It is truly hard to comprehend the scope of this expenditure, and it continues to rise.
The official narrative, which is presented at every conference, is that artificial intelligence (AI) will transform every aspect of life, including devices, software, advertising, healthcare, and productivity. The businesses that develop the infrastructure now will dominate the global economy for many years to come. They might be correct. It appears from history that businesses that are prepared to make significant investments during platform changes typically succeed: Before it evolved into AWS and essentially funded everything else, Amazon’s initial cloud infrastructure venture appeared reckless for years.
Key Facts & Context
| Total AI Infrastructure Spend (2026) | Big Tech on pace to spend over $600 billion on AI infrastructure in 2026 — the majority directed at training and running AI models |
|---|---|
| Meta | Capital expenditure for 2025 set at $70–$72 billion; spending growth in 2026 expected to be “notably larger”; CEO Mark Zuckerberg says the company is “compute-starved” and accelerating investment |
| Alphabet (Google) | Raised 2025 capex forecast to $91–$93 billion — nearly double its 2024 capital expenditure figure; includes data centers, chips, and AI product development |
| Microsoft | Spent $34.9 billion in a single quarter (Q1 FY2026) on data centers and AI infrastructure — above analyst expectations, up from $24 billion the prior quarter; shares lost nearly 20% of value since start of 2026 |
| Consumer Cost Shifting | Microsoft 365 Premium now $19.99/month bundling Copilot AI; Google Workspace raised prices 16–33% after adding Gemini AI; Adobe Creative Cloud raised monthly fee by $10 linked to generative AI tools |
| Investor Alarm | A net 35% of fund managers surveyed by Bank of America (February 2026) said companies are “overinvesting” in AI — a record high, up sharply from 14% in December 2025 |
| Stargate Project | Announced January 2026 by President Trump; involves OpenAI, SoftBank, Oracle and others committing up to $500 billion for U.S. AI infrastructure buildout |
| Economic Impact | Bank of America senior economist noted AI-related business investment has been one of only two forces holding up U.S. GDP growth alongside consumer spending — making the stakes of a pullback significant |
However, a growing number of professional investors are refusing to believe the long-view argument made by the tech industry. According to a Bank of America survey conducted in February 2026, a net 35% of fund managers thought businesses were overinvesting in AI, which is a record high and a significant increase from 14% just two months prior. That is no longer a minor issue. The majority of those who oversee large sums of money hold that view.
This anxiety has begun to manifest itself in tangible ways in the stock market. Even as the company continued to report robust cloud revenues, Microsoft shares lost almost a fifth of their value since the beginning of 2026. During that time, the iShares Expanded Tech-Software Sector ETF lost almost 25% of its value. The infrastructure builders are rewarded first, then questioned, and then punished if the returns don’t materialize on time. This pattern is familiar from other tech cycles. The sheer size of the numbers is what’s different this time. When the wager is valued at hundreds of billions of dollars, a calculation error not only depresses stock prices but also reroutes capital at a scale that affects other sectors of the economy.
The story becomes more intimate than most reporting indicates when one considers the question of who is actually funding all of this. Bundling is a covertly efficient way that tech companies are transferring AI costs to customers. Copilot AI features that the majority of business users didn’t specifically request and find difficult to separate from their current subscriptions are now included in Microsoft 365. Early in 2025, Google Workspace added its Gemini AI assistant to Business and Enterprise plans, with price increases ranging from about 16% to 33% depending on the tier. For a fifty-person business on Business Plus, this meant an extra $2,400 a year for AI tools that came bundled whether or not they were desired. Expanded generative AI features were cited by Adobe as the reason for the $10 monthly increase in Creative Cloud. The decision of whether or not to use AI is not being made by the customer. They only discovered it on their bill.
As this dynamic develops, there’s a sense that the tech sector has learned a valuable lesson from the streaming wars: it’s much simpler to raise prices when the new feature is integrated into the product rather than presented as an obvious add-on. When costs increased and the value was questionable, customers cancelled their Netflix subscriptions. When their entire company uses Microsoft 365 and the AI is already present, they are less likely to discontinue it. In a limited commercial sense, the bundling strategy is quite elegant, but from the consumer’s point of view, it can be challenging to oppose.
Economists are starting to pay close attention to the larger macro aspect of all of this. Consumer spending and AI-related business investment have been the two factors impeding U.S. GDP growth in recent months, according to a senior economist at Bank of America. That framing is important. The economic impact won’t be limited to tech stocks if AI investment slows down, either because returns don’t materialize, fund managers lose patience, or a DeepSeek-style development suddenly makes expensive American infrastructure seem unnecessary. Most people won’t anticipate how quickly and widely it will be felt.
This level of spending by the companies is not irrational. They sincerely think that the AI moment is real and that hesitation is more risky than overcommitment, and they have some historical support for this belief. The cost of being late is greater than the cost of being early, according to statements made by Satya Nadella, Mark Zuckerberg, and Sundar Pichai. Whether that reasoning applies at $600 billion annually is still up for debate. However, spending is still occurring, subscriptions are increasing, and returns are still primarily a promise made during quarterly earnings calls and PowerPoint presentations. Wall Street is documenting it. and holding out.
