AI's Money Problem: Google's Ballooning Costs Spook Investors Across the Whole Industry

Key points

• Google raised its 2025 capital spending estimate to between $195 billion and $205 billion, up from a previous top-end forecast of $190 billion.
• Nvidia is reportedly in deal talks across several agreements, including a $250 billion arrangement to guarantee OpenAI's debt.
• Meta, Amazon and Microsoft are all reporting earnings this week and are expected to announce similar spending increases on data centres.
• A competitive AI model from a Chinese start-up has renewed concerns that expensive chips may not be strictly necessary to build powerful AI.
• Some long-term AI investors already expect a wave of company failures once the current spending boom corrects.

Alphabet, Google's parent company, reported earnings this week and buried inside the numbers was a figure that made investors uncomfortable: the company now expects to spend between $195 billion and $205 billion this year building out its AI infrastructure. Its previous forecast had put the top of that range at $190 billion.

That's not a rounding error. The new low end exceeds the old high end by $5 billion, which means Google has essentially told investors it can't accurately forecast its own costs.

The company is currently spending more than it's bringing in from AI, while facing pressure to keep service prices low to stay competitive. Two weeks ago our story found that massive AI infrastructure spending had pushed Google into negative cash flow for the first time, even as its cloud and search businesses hit record highs. Spending more while earning the same, or less, is a straightforward problem for any business.

Is this just Google's problem?

No. The same pressure sits on the whole industry. Meta, Amazon and Microsoft are all reporting earnings this week, and analysts widely expect each to announce higher-than-expected spending on data centres, the warehouse-sized buildings packed with computers that power their AI products.

Nvidia, whose chips most AI systems depend on, is reportedly in deal talks across several agreements. One deal, worth around $250 billion, would see Nvidia guarantee debt taken on by OpenAI. Billy Leung, investment strategist at Global X Management, told Bloomberg that arrangement is "as much a reminder of funding strain in the AI build-out as it is a demand signal." It looks less like booming demand and more like a sign that companies building AI need financial help.

Oracle, whose public stock price many investors treat as a rough proxy for OpenAI's health, is also drawing scrutiny over the debt it has taken on to build data centres.

What does a Chinese AI model have to do with any of this?

Quite a lot. A new model from a Chinese start-up spooked investors this week. Chinese companies theoretically have less access to Nvidia's most advanced chips due to US export restrictions, yet their AI systems keep matching or approaching the performance of American ones. If you can build competitive AI with fewer or cheaper chips, the entire argument for spending hundreds of billions on infrastructure starts to look shaky. AI2Day has tracked this unease across 14 stories tagged "Chinese AI" since July, and the nervousness hasn't faded.

What happens next?

Earnings from the other big tech companies this week could calm nerves if the numbers look better than feared. They might not. Several experienced investors spoken to by The Verge's Elizabeth Lopatto said they already expect the industry to overbuild data centres and that many smaller AI companies will fail when spending pulls back. Their bet is that survivors will generate enough returns to cover the losses on the ones that don't.

For ordinary people using AI tools, the near-term effect is likely a continuation of the current price war: companies fighting for users tend to keep services cheap or free. The harder question, which nobody has answered convincingly, is how these companies plan to make the money back. I've been asking a version of that question since this beat started, and the honest answer is that the AI boosters themselves are watching for a market top, because they know one's coming.

Common questions

Does this mean my AI tools will get worse or more expensive?

Not immediately. Companies competing for users have every reason to keep prices low and quality high for now. A correction, if one comes, would more likely shrink the number of companies offering AI tools than raise prices overnight.

Should ordinary investors be worried about tech stocks?

That depends entirely on your individual situation, but the pattern worth watching is simple: if major AI companies keep spending more than they earn and no clear path to profit emerges, valuations built on future promises become harder to defend.