What is your basis for this $600 billion figure, even $600 billion over two years is far more than all of these companies are spending.
The estimates I’ve aware of $.01 - $.10 per query. Google processes about 10 trillion queries a year at an estimate of $.001. There is no LLM processing anything close to that volume and that’s the only way to get to $600 billion.
$60 billion sounds far more reasonable give the number of LLM queries that has been published against the published costs. Thats a big enough number, no need to juice it.
Training is the expensive part. There are also hordes of developers either developing the LLMs themselves or something to work with an LLM, and theyre often quite well-paid.
I don’t trust a random unsourced figure either but queries likely aren’t even half the total cost.
Thank you for realizing what most AI fans fails to realize!
60 billion is a pretty reasonable ballpark for what AI companies are making. It’s closer to 50b annualized (so 35b for real) in 2025 , but same ballpark. What they’re actually SPENDING is an entirely different number. The 600 billion is a quick-and-dirty extrapolation from this sum of capital expenditure by the 5 biggest AI companies. Note that OpenAI direct spending isn’t in there, because they’re not reporting it, but OpenAI is mostly Microsoft.
That brings us to a quick 600b in 2 years in expenses, which are climbing, and some 50b in income, with absolutely zero profit. And that’s a problem, because all that capex is not infrastructure, it’s mostly consumable GPUs that will be worn out in a few years and insane salaries.
Seems reasonable. If the price I have to pay is to step over my fellow citizens who happen to be experiencing an involuntary outdoor lifestyle, that’s fine.
What is your basis for this $600 billion figure, even $600 billion over two years is far more than all of these companies are spending.
The estimates I’ve aware of $.01 - $.10 per query. Google processes about 10 trillion queries a year at an estimate of $.001. There is no LLM processing anything close to that volume and that’s the only way to get to $600 billion.
$60 billion sounds far more reasonable give the number of LLM queries that has been published against the published costs. Thats a big enough number, no need to juice it.
Training is the expensive part. There are also hordes of developers either developing the LLMs themselves or something to work with an LLM, and theyre often quite well-paid.
I don’t trust a random unsourced figure either but queries likely aren’t even half the total cost.
Thank you for realizing what most AI fans fails to realize!
60 billion is a pretty reasonable ballpark for what AI companies are making. It’s closer to 50b annualized (so 35b for real) in 2025 , but same ballpark. What they’re actually SPENDING is an entirely different number. The 600 billion is a quick-and-dirty extrapolation from this sum of capital expenditure by the 5 biggest AI companies. Note that OpenAI direct spending isn’t in there, because they’re not reporting it, but OpenAI is mostly Microsoft.
Some adding up from the WSJ top-4 here for 2024: https://www.wsj.com/tech/ai/tech-giants-double-down-on-their-massive-ai-spending-b3040b33
130 billion from microsoft: https://www.cfodive.com/news/microsoft-capex-grow-slower-rate-cfo-ai/746947/
75 billion from alphabet here: https://www.cnbc.com/2025/02/04/alphabet-expects-to-invest-about-75-billion-in-capex-in-2025.html
72 billion from meta here: https://fortune.com/article/meta-q1-earnings-revenue-profit-beat-ai-capex-raise
11 b from tesla: https://www.reuters.com/business/autos-transportation/tesla-expects-capital-expenditure-exceed-11-bln-2026-2027-2025-01-30
And the same from apple: https://finbox.com/NASDAQGS:AAPL/explorer/capex/
That brings us to a quick 600b in 2 years in expenses, which are climbing, and some 50b in income, with absolutely zero profit. And that’s a problem, because all that capex is not infrastructure, it’s mostly consumable GPUs that will be worn out in a few years and insane salaries.
Seems reasonable. If the price I have to pay is to step over my fellow citizens who happen to be experiencing an involuntary outdoor lifestyle, that’s fine.