Big Tech's AI Spending Blitz Collides With Investor Patience

Today’s Focus

Big Tech’s quarterly earnings this week put a sharper number on the AI arms race that OpenAI’s ChatGPT launch kicked off in late 2022, according to the BBC.

Alphabet, Google’s parent, reported $118 billion in revenue for the quarter but negative free cash flow, the first time the company has spent more than it brought in since going public, the BBC reported.

Meta’s free cash flow came in at just $784 million on $61 billion of revenue, meaning it spent nearly every dollar it earned. Its Reality Labs division, which houses much of its AI and headset work, lost close to $9 billion in the first half of 2026, according to the BBC’s read of the filings.

Every major U.S. tech firm now runs a consumer AI chatbot: OpenAI’s ChatGPT, Google’s Gemini, Meta AI, Amazon’s Rufus, and a rebuilt Siri from Apple. None of them, the BBC noted, generate meaningful standalone revenue.

Amazon and Apple used their calls this week to describe how they plan to route AI into products customers already pay for, rather than as a standalone subscription business, per BBC coverage.

Investors responded most sharply to Meta. Shares fell to near their lowest level in a year after CEO Mark Zuckerberg told analysts the company was still building out its next-generation AI infrastructure, the BBC reported.

Alphabet also disclosed higher capital-expenditure guidance tied to data centers and custom AI chips. The pattern across the sector is the same: capex is climbing faster than AI-attributable revenue, and executives are asking shareholders for more runway.

The Debate

Supporters argue

Executives and AI boosters say the spending is exactly what a generational technology shift requires. Zuckerberg told analysts Meta is building “industry-leading infrastructure” that will underpin its ads business, its apps, and future devices, according to the BBC.

Alphabet CEO Sundar Pichai has argued that Google’s own chips and data centers give it a durable cost advantage as inference demand scales, a point echoed on the earnings call and summarized in BBC reporting.

Amazon and Apple are making a narrower case: rather than sell AI directly, embed it in Prime, AWS, and iPhone features that already have paying customers. The BBC noted that both companies framed AI as a margin lever inside existing products, not a new P&L line.

Industry investors including Dan Ives of Wedbush Securities have argued publicly that the current build-out mirrors early cloud investment, which took years to show up in cash flow but eventually reset the economics of the whole sector.

Critics argue

Skeptics say the numbers no longer support the story. Alphabet’s negative free cash flow and Meta’s near-zero quarter are, in the BBC’s framing, the clearest signal yet that “investors are no longer placated” by promises of future returns.

MIT economist Daron Acemoglu has estimated that generative AI will add only modest gains to U.S. productivity over the next decade, a figure he laid out in a 2024 paper and has repeated in interviews, arguing that current capex assumes far more.

Short-seller Jim Chanos and hedge-fund manager David Einhorn have both said publicly this year that AI infrastructure spending is running well ahead of any demonstrated enterprise willingness to pay.

Consumer advocates add a second critique: the chatbots being funded by this spending, Rufus, Gemini, Meta AI, still hallucinate, still leak data, and still lack a clear business model beyond ad targeting, points raised repeatedly by the Mozilla Foundation and the Electronic Privacy Information Center.

What the experts say

Independent researchers have started to put ranges around the gap. A July 2025 Goldman Sachs research note estimated that hyperscalers would spend more than $1 trillion on AI capex over several years while generating only a fraction of that in direct AI revenue so far.

Stanford’s 2025 AI Index, produced by the Institute for Human-Centered AI, found that private AI investment in the United States hit $109 billion in 2024, roughly 12 times China’s figure, but that enterprise adoption rates for generative AI plateaued between 2024 and 2025.

The International Energy Agency projected in its 2024 electricity report that global data-center power demand could double by 2026, driven largely by AI, raising costs that do not appear in software margins.

Erik Brynjolfsson of Stanford has argued in published work that productivity gains from AI are real but arrive with a lag, historically a decade or more for general-purpose technologies like electricity and the internet. That timing gap, the Congressional Budget Office noted in a 2024 background paper on AI and the economy, is what makes the current investment cycle unusually risky for public shareholders.

By the Numbers

$118 billion: Alphabet’s revenue for the reported quarter, with free cash flow turning negative for the first time since its IPO, per the BBC.

$784 million: Meta’s free cash flow on $61 billion in revenue for the quarter, according to the BBC.

$9 billion: approximate losses at Meta’s Reality Labs division in the first half of 2026, per BBC reporting on Meta’s filings.

$109 billion: private AI investment in the United States in 2024, according to the Stanford AI Index 2025.

$1 trillion+: projected multi-year hyperscaler AI capex, per a 2025 Goldman Sachs research note cited across industry coverage.

2x: International Energy Agency projection for global data-center electricity demand growth by 2026, driven largely by AI workloads.

Late 2022: launch of OpenAI’s ChatGPT, the event the BBC identifies as the start of the current AI investment race.