# Hyperscalers can't win with investors as they spend big on AI > Source: > Published: 2026-07-27 09:51:01+00:00 # Hyperscalers can't win with investors as they spend big on AI [Business Insider](https://www.businessinsider.com) Investors fear AI is following the dot-com playbook, with massive infrastructure spending and uncertain returns raising bubble concerns. - Investors have grown increasingly skeptical about hyperscalers' AI spending. - Alphabet's stock dropped more than 7% last week after the company raised its projected AI spending. - But investors like Molly Pieroni say hyperscalers should maintain AI spending to stay competitive. When Alphabet [revised up its 2026 AI spending projections](https://www.businessinsider.com/tesla-alphabet-stock-price-q2-earnings-capex-ai-spending-tsla-2026-7) by $15 billion to $190 billion-$205 billion in its earnings report last Wednesday, investors were quick to show their displeasure. Shares of Alphabet fell as much as 7.4% in trading on Thursday. It was the latest sign that the market is increasingly skeptical that the heaps of cash hyperscalers are throwing at the AI buildout will see a payoff anytime soon. Investors worry that they're watching a [replay of the dot-com bubble](https://www.businessinsider.com/stock-market-bubble-ai-valuations-nvidia-broadcom-dot-com-crash-2026-6), when companies spent big on internet infrastructure, only to see poor initial returns, leading to a crash in stock prices. They recently cheered on Meta after reports that the company is planning on leasing out its spare cloud capacity — cashing in on all the demand — sending its shares soaring as much as 21% in a matter of days. So, should hyperscalers follow the collective wisdom of the market and slow their spending plans? Last Thursday, as Alphabet's stock was getting pummeled, I posed the question to Molly Pieroni, the president at Yacktman Asset Management, which invests in Alphabet. Her answer was a resounding no — an interesting contrast to the prevailing narrative at the moment, coming from someone who has skin in the game. "We always like discipline, but in this case, you're really needing to compete to come out the other end in a spot where you want to be," Pieroni said. "So we're not telling them that they need to pull back on capex and invest less." It highlights the Catch-22 dynamic facing hyperscalers right now. If they pull back on spending, investors might reward them in the short term. But in the long run, they may lose the AI race, while also upsetting investors like Pieroni who think all the spending is justified. In one sense, investors are right to demand financial discipline from companies they hold. But they also may be overreacting to all the spending to a degree. Pieroni pointed out that Google in particular has a healthy balance sheet, and has a fairly diversified business that it can rely on to drive earnings. The company did beat on earnings in Q2, after all. For now, it seems most hyperscalers are continuing to tune out investors' protests against their spending. We'll get more insight into their thinking when Microsoft, Meta, and Amazon report earnings later this week. Last week, I also spoke with Brandon Nelson, [a top-performing fund manager at Calamos Investments](https://www.businessinsider.com/ai-hardware-stock-picks-to-buy-capex-beneficiaries-strl-sitm-2026-7) who is continuing to bet on hyperscaler capex beneficiaries, even after investors have dumped them in recent weeks on fears that the spending could stop. He said the nerves in the market right now are likely to be a temporary blip in an ongoing bull market, and that the spending glut would continue. "1996 through the peak of the stock market in 2000, there were 10 10% corrections in the Nasdaq. Those were not comfortable," he said. "My best guess is we're going through one of those kind of gut checks in the AI infrastructure buildup." [Business Insider](https://www.businessinsider.com/hyperscalers-cant-win-with-investors-ai-spending-capex-alphabet-earnings-2026-7) Get AI news in your inbox Daily digest of what matters in AI.