The article seems to hinge on the core assumption that revenue is much less than spending:
> Those expenditures may be approaching $1 trillion for 2025, while AI revenue—which would be used to pay for the use of AI infrastructure to run the software—will not exceed $30 billion this year
While it's clear that the author is summing up the spending from the big players, it's not clear to me that their math is right for revenue. Yes, OpenAI, Anthropic, Thinking Machines, SSI, etc. have pretty limited direct revenue (including zero!).
But this comparison assumes no revenue growth for other top computing users. Some companies are certainly saving money on some tasks and increasing revenue, particularly in fields like customer support. See the confusing figure in section 5 of https://hai.stanford.edu/ai-index/2025-ai-index-report/econo... .
That chart is by number of respondents and not weighted by revenue. Like the MIT study, it would not be surprising that "just pipe this to an LLM" isn't enough for most fields or companies. But a few have likely made material improvements.
10% of respondents saying they've seen a >10% revenue gain could be substantial, if they're bigger firms with high leverage in computing.
Edit to add: the comparison also makes a classic "GDP vs market cap" style mistake. Capital expenditure has multiple years of useful life. Revenue is annual. You'd want to compare depreciation vs revenue.
> Those expenditures may be approaching $1 trillion for 2025, while AI revenue—which would be used to pay for the use of AI infrastructure to run the software—will not exceed $30 billion this year
While it's clear that the author is summing up the spending from the big players, it's not clear to me that their math is right for revenue. Yes, OpenAI, Anthropic, Thinking Machines, SSI, etc. have pretty limited direct revenue (including zero!).
But this comparison assumes no revenue growth for other top computing users. Some companies are certainly saving money on some tasks and increasing revenue, particularly in fields like customer support. See the confusing figure in section 5 of https://hai.stanford.edu/ai-index/2025-ai-index-report/econo... .
That chart is by number of respondents and not weighted by revenue. Like the MIT study, it would not be surprising that "just pipe this to an LLM" isn't enough for most fields or companies. But a few have likely made material improvements.
10% of respondents saying they've seen a >10% revenue gain could be substantial, if they're bigger firms with high leverage in computing.
Edit to add: the comparison also makes a classic "GDP vs market cap" style mistake. Capital expenditure has multiple years of useful life. Revenue is annual. You'd want to compare depreciation vs revenue.