Media Flags GPU Skepticism

A graphics processing unit is the chip that does the heavy lifting for modern artificial intelligence, and the new question is whether all the chips being ordered will earn enough money to justify the bill. (sequoiacap.com) That question moved into mainstream coverage this month as MIT Technology Review pointed readers to new data showing artificial intelligence companies are bringing in revenue quickly while still spending hundreds of billions of dollars on chips and data centers. Stanford’s 2026 Artificial Intelligence Index, released April 13, said artificial intelligence data centers worldwide can now draw 29.6 gigawatts of power. (technologyreview.com (hai.stanford.edu) The spending is not hypothetical. Nvidia said on August 27, 2025 that its quarterly data-center revenue reached $41.1 billion, up 56% from a year earlier, and that Blackwell data-center revenue grew 17% from the prior quarter. (investor.nvidia.com) The buyers are also naming numbers. Meta reported on January 28, 2026 that capital expenditures were $72.22 billion for full-year 2025, including $22.14 billion in the fourth quarter alone. Alphabet said on February 4, 2026 that most of its capital spending went to technical infrastructure, with about 60% for servers and 40% for data centers and networking gear. (nasdaq.com) (abc.xyz) Wall Street has started putting the total in one place. Bloomberg reported on February 5 that Amazon, Alphabet, Meta, and Microsoft together had forecast about $650 billion of capital expenditures in 2026, most of it tied to the artificial intelligence build-out. (bloomberg.com) The skeptical case is straightforward: if a graphics processing unit is like a factory machine, the owner needs enough paid work to keep it busy. Sequoia Capital argued in June 2024 that Nvidia’s revenue implied a much larger amount of end-customer revenue would be needed across the ecosystem, and said the old “$200 billion question” had become a “$600 billion question.” (sequoiacap.com) Sequoia also argued that the shortage story had changed by mid-2024. Its report said late 2023 was the peak of the graphics processing unit squeeze and that, for many buyers, access had become easier with more reasonable lead times. (sequoiacap.com) The bullish case is just as concrete. Meta chief executive Mark Zuckerberg said on the company’s January 28 earnings call that the business saw “AI-driven performance gains,” while Nvidia chief executive Jensen Huang said in August 2025 that “demand is extraordinary” for Blackwell systems. (s21.q4cdn.com) (investor.nvidia.com) The build-out is now spreading beyond chips into land, power, and long-term contracts. Meta said on February 11 that it was breaking ground in Lebanon, Indiana on a 1-gigawatt data-center campus, one of its largest infrastructure investments to date, and Reuters reported on April 9 that Meta had expanded its CoreWeave cloud deal by $21 billion. (about.fb.com) (msn.com) The media shift is less about whether artificial intelligence is real than about what counts as proof that the spending works. The next earnings reports will show whether cloud sales, advertising gains, and paid artificial intelligence tools are rising fast enough to keep the graphics processors busy. (technologyreview.com) (cnbc.com)