The honeymoon phase might be over. While Meta, Google, Amazon, and Microsoft are throwing nearly $357 billion annually at AI—a quarter of the entire S&P 500’s R&D budget—major investors are asking the uncomfortable question: When does this actually make money?
The Math Doesn’t Add Up (Yet)
Here’s where it gets spicy: Goldman Sachs predicts these companies will collectively spend $1 trillion on AI infrastructure in coming years. But Barclays just dropped a reality check—cloud providers are expected to spend $60 billion yearly on AI data centers by 2026, but only generate $20 billion in new revenue. That’s a 3x gap.
So basically: massive capex, underwhelming returns.
Jim Cavello, Goldman’s global equity chief, is openly skeptical. He’s questioning whether AI will actually deliver returns comparable to the internet revolution, and suggests the ROI timeline is way longer than Wall Street’s current expectations. Meanwhile, Nvidia remains the only real winner—the shovel seller in the gold rush while everyone else burns cash.
The Fear Factor
Barclays analysts have a blunt take: this isn’t strategic foresight, it’s FOMO. Tech giants are terrified of falling behind, so they’re building data centers faster than demand justifies. Result? Overcapacity and declining hardware prices (maybe). But here’s the catch—Nvidia’s still king, and competitors haven’t seriously challenged their grip. So costs stay high.
The Reckoning Is Coming
Amazon’s 2024 data center spend jumped from $53B to $63B. Meta and Alphabet are breaking records. But Goldman Sachs’ Ryan Hammond dropped the bomb: if revenue projections miss, expect a “valuation de-rating.” Translation: stock sells off hard.
Current AI spending is still below dot-com era levels, Goldman notes—but Barclays counters that existing infrastructure should already be overkill for internet demand. This could get ugly when Wall Street finally stops asking “when will AI be profitable?” and starts demanding actual proof.
The next 6-12 months? Watch for revenue guidance revisions. That’s the real tell.
Voir l'original
Cette page peut inclure du contenu de tiers fourni à des fins d'information uniquement. Gate ne garantit ni l'exactitude ni la validité de ces contenus, n’endosse pas les opinions exprimées, et ne fournit aucun conseil financier ou professionnel à travers ces informations. Voir la section Avertissement pour plus de détails.
La mise de Big Tech sur $357B AI est-elle une bulle ? Les doutes croissants de Wall Street
The honeymoon phase might be over. While Meta, Google, Amazon, and Microsoft are throwing nearly $357 billion annually at AI—a quarter of the entire S&P 500’s R&D budget—major investors are asking the uncomfortable question: When does this actually make money?
The Math Doesn’t Add Up (Yet)
Here’s where it gets spicy: Goldman Sachs predicts these companies will collectively spend $1 trillion on AI infrastructure in coming years. But Barclays just dropped a reality check—cloud providers are expected to spend $60 billion yearly on AI data centers by 2026, but only generate $20 billion in new revenue. That’s a 3x gap.
So basically: massive capex, underwhelming returns.
Jim Cavello, Goldman’s global equity chief, is openly skeptical. He’s questioning whether AI will actually deliver returns comparable to the internet revolution, and suggests the ROI timeline is way longer than Wall Street’s current expectations. Meanwhile, Nvidia remains the only real winner—the shovel seller in the gold rush while everyone else burns cash.
The Fear Factor
Barclays analysts have a blunt take: this isn’t strategic foresight, it’s FOMO. Tech giants are terrified of falling behind, so they’re building data centers faster than demand justifies. Result? Overcapacity and declining hardware prices (maybe). But here’s the catch—Nvidia’s still king, and competitors haven’t seriously challenged their grip. So costs stay high.
The Reckoning Is Coming
Amazon’s 2024 data center spend jumped from $53B to $63B. Meta and Alphabet are breaking records. But Goldman Sachs’ Ryan Hammond dropped the bomb: if revenue projections miss, expect a “valuation de-rating.” Translation: stock sells off hard.
Current AI spending is still below dot-com era levels, Goldman notes—but Barclays counters that existing infrastructure should already be overkill for internet demand. This could get ugly when Wall Street finally stops asking “when will AI be profitable?” and starts demanding actual proof.
The next 6-12 months? Watch for revenue guidance revisions. That’s the real tell.