Amazon, Alphabet, Microsoft, and Meta shelled out $130.6 billion on AI in the first three months of 2026 alone. That’s more than three times the Manhattan Project’s total cost. By year-end, their combined spending could hit nearly $700 billion, dwarfing NASA’s Apollo Program and any single-year corporate outlay in modern history.
Numbers like these demand scrutiny. Wall Street analysts now peg total AI capital expenditures above $1 trillion in 2027, per a CNBC report. Amazon eyes $200 billion this year, up from $72 billion in 2025. Alphabet hiked its forecast to $180-190 billion after buying data-center energy firm Intersect. Meta boosted to $125-145 billion, citing memory chip costs. Microsoft? Another $190 billion, with tens of billions just for parts.
Short bursts of revenue growth mask the strain. All four beat earnings expectations with double-digit sales jumps. Yet Meta shares dropped 7% after Mark Zuckerberg warned of billions more in spending. Alphabet’s free cash flow? Projected to plummet 90% to $8.2 billion. Amazon’s could turn negative by $17-28 billion, prompting SEC filings for debt raises. Microsoft ekes out flat cash flow. Meta’s dwindles.
Rising input costs bite hard. Memory chips surged 50% this year, with more increases ahead. Zuckerberg pinned Meta’s capex hike on that alone: “We are increasing our infrastructure capex forecast for this year. Most of that is due to higher component costs, particularly memory pricing,” he said April 29.
And power. Data centers guzzle electricity, pushing builds to rural “man camps” with free steaks and golf simulators to lure workers, as Bloomberg detailed. S&P Global warns energy shocks could force cuts in the $635 billion (pre-hike) AI tab, per Reuters.
Critics Cry Bubble—But Demand Outstrips Supply
NYU professor Gary Marcus blasted it on X as “sheer insanity,” the “greatest capital misallocation in history.” Billionaire Ray Dalio sees 80% of the dot-com euphoria. Hardware depreciates fast too—AI gear loses value in three years amid Nvidia’s yearly leaps, turning much capex into maintenance, not growth, according to Fortune.
But here’s the counterpunch. Google Cloud admits leaving revenue on the table—can’t build fast enough. Quarterly capex tripled from $162 billion in 2022 to $448 billion in 2025 across five firms, hitting $140 billion in Q4 alone, Visual Capitalist charted. Amazon CEO Andy Jassy remains “confident in the long term capex investments.” Layoffs? 81,747 in Q1 2026, reallocating headcount to GPUs.
Pay influencers $400,000-$600,000 each? Sure, AI ad spend jumped 495% in January 2026. But ChatGPT’s daily burn was $700,000 pre-scale; now it’s inference at hyperscale. HBM memory? Sold out for 2026, Micron’s market ballooning to $100 billion by 2028.
So, debt-fueled sprint or historic buildout? Big Tech spent $410 billion last year, up 60% to $650-725 billion now—matching Goldman Sachs’ $500 billion-plus call from late 2025, which underestimated. Wall Street Journal tallies $670 billion for 2026 across the four, with Morgan Stanley eyeing $2.9 trillion through 2028 on chips and servers alone.
Free cash flow race to the bottom, yes. Amazon borrowing. Google quadrupling debt. Yet revenue from AI clouds surges—Google Cloud up 63%, outpacing rivals, Reuters reports. OpenAI raised $122 billion, burning it on compute in three years.
Bubble fears echo dot-com. But railroads took 6% of 1880s GDP; AI’s at 1.2% now, with room to run. Compute bottleneck favors leaders. Miss this window? Get left behind. Stocks dipped post-earnings. Investors uneasy. Execution decides winners.
One thing clear. This $700 billion burn reshapes everything—data centers, power grids, markets. Profitable payoff or epic write-down? Watch cash flows. And the GPUs.
WebProNews is an iEntry Publication