Companies spend just 0.1% of their budgets on AI. Why so little?
Wallach calculates that doubling or tripling that share gets to a couple trillion dollars a year, but he wonders why businesses are not spending more already.
This Week in Startups · Is AI the new cigarettes? Our VC panel digs into Anthropic’s S-1 | E2356D.A. Wallach, an investor, was going through Anthropic's stock market filing with a panel of fellow investors when he stopped on a number he called the most interesting in the filing. The company said users of its AI were spending roughly one tenth of one percent of their total operating spending on AI, though Wallach said he did not know if that meant its top customers or corporate America broadly.
His question was blunt: "Why aren't they already today spending 1% or a half percent?"
Wallach said Anthropic presented the tiny share as headroom for growth. His own math, as he described it, was that doubling or tripling that tiny share would imply annual spending in the low trillions.
He argued the same number could be read the opposite way. It complicates the argument that cheaper AI will automatically unlock far more spending, he said, because companies are already testing the technology extensively. He asked whether what holds them back is price, or whether the tool is simply not yet useful enough for the work companies do.
He offered the disclaimer he said everyone must give before criticizing AI, that he uses it all day and loves it and that it has changed his life and his business. Even so, he said, his hypothesis was that the low spending was mainly about usefulness, and he kept asking why companies were not already spending twice that amount.
That led to a second question about timing. Wallach said Anthropic is rationally ordering huge amounts of computing power in advance, and every chief executive is rationally running pilots and buying tokens to learn what the technology can do. The open question, he said, is whether those two sides, AI makers buying computing power and companies buying AI, will grow at rates that match on supply and demand.
Jason Calacanis, a fellow investor, offered a partial answer in plain terms. He blamed employees who cannot yet use the tools, not a shortage of supply, using his term "meat sacks" for the people in organizations who have not been able to deploy it. Wallach kept coming back to the number: they are only spending point one percent.
Questions people ask
What does 0.1% of operating spending actually mean?
This definition is general background, not from the transcript. Operating spending is everything a company pays to keep running, including salaries, rent, software, and supplies. If a business spends $1 billion a year to operate, 0.1% is $1 million for AI.
What is an S-1 filing?
An S-1 is the detailed financial document a company files with U.S. regulators when it plans to sell shares to the public. It discloses revenue, losses, customers, risks, and contracts. Investors study it because it is often the first complete look inside a private company.
Does low spending prove AI is not useful enough?
This is my reading, not a view Wallach stated. He asked whether low spending reflects high cost or something else, and did not answer his own question. What he did say is that talk of supply constraints presumes that if the price were lower, companies would be spending more.