Crunching the AI Numbers

By
Line drawing of a king in a crown and robes bent over a counting device.

Should have used Excel. Image via the British Museum.

It can be very hard to get to terms with how big AI is becoming—the sheer scope of this new space. So I worked with ChatGPT and some AI blogs to extrapolate some interesting, large AI-related numbers. Here they are:

$3.26 trillion: Nvidia’s market cap in October 2024, driven by AI chip demand. (Reuters)
$11 billion: Loans from Wall Street to startups for Nvidia’s AI chips. (Financial Times)
$832.8 billion: TSMC’s market cap in October 2024, also due to chip demand. (Reuters)
$60 billion: Investment in AI infrastructure by major tech companies in late 2024, up 60% YoY. (Investopedia)
$25.2 billion: Investment in generative AI in 2023, nearly 9x 2022 levels. (World Economic Forum)
$400 million: Amount raised by AI startup Physical Intelligence, valuing it at $2 billion. (Reuters)
133 million: New jobs expected from AI technologies by 2030. (Tech.co)
37%: Businesses using AI as of early 2024. (DataProt)
$407 billion: Estimated value of the AI market by 2027. (CompTIA)
405 billion: The number of parameters in Meta’s Llama 3 in 2024, up from 117 million in OpenAI’s GPT-1 in 2018. (Ars Technica, Reuters)

ChatGPT dug around on the web and found the numbers, and I rewrote everything. It was a useful exercise but I don’t think I’ll do this much in the future—I don’t think people want a bunch of numbers every week. But I do have a reproducible process for finding stats about AI, based on recent news, whenever I want. And I can show it RSS feeds in any industry and get similar results. I asked it to stop focusing on AI and instead focus on the “Food”-related feeds. It produced a lot of inflation-related stats and then:

0.3%: Increase in the food-away-from-home index in April, reflecting a rise in dining out expenses. (Barron’s)

And that’s how I learned there’s a “Food Away From Home” index (see chart)! 

I’m seeing more and more automated, aggregated newsletters written by AI tools. I think that’s a mistake. The purpose of a newsletter is to connect the company (Aboard) to humans (you) over time, so when you need some help building a big, complex piece of business software, you give us a call, human to human. Right now the newsletter has a very good open rate; if I put AI in charge, you’d all drift away.

However, I am very interested in having lots of stats, facts, and links at my fingertips when I sit down to write the newsletter. That’s a huge amount of work to gather solo, and if the computer helps, it will help me be more thoughtful. Numbers represent a kind of distilled fact—they keep you honest. So I’m going to keep developing this idea.

One of the things that AI is good at is filling in an empty page with absolute nonsense, and then you can respond to that. It’s terrible, but the portions are large. But I think what’s even better than that is AI going out, reading the news, coming back with stats and figures and pointers, and letting you sift a bit before you, the writer/creator-person, sit down and get to work. I want humans to reassert control of the blank page. The computer gives us information we can process and use, and then we decide what gets written down. That’s the way to go.