A sense of scale: web ads pay for AI/ML

Worldwide spending on all advertising is on the order of 1 trillion USD per year (source: statista). Web ad companies capture large fractions of this. For example, Meta gets about 10%, and Google gets about 20%. Then those companies invest heavily in R&D of various types, including for machine learning (ML) / artificial intelligence (AI). That money pays the salaries of many bright and creative ML/AI researchers, some of whom I’m fortunate to count among my friends and colleagues. Meta spends about 0.03 trillion USD on R&D, and Google spends about 0.04 trillion USD.

For comparison, the entire US government annual budget (outlays) are about 6 trillion USD right now. R&D budgets for the NIH and NSF are about 0.05 trillion USD and 0.01 trillion USD, respectively. The DOD spends more, about 0.1 trillion USD. (Much of that is on basic R&D, not weapons specifically. However, I haven’t found a good source for a more granular analysis just yet.)

Thus, the R&D spends of individual web ad companies are on par with entire US government R&D arms. This is a relatively new phenomenon. I wrote about this in 2012, and the ratios were different then.

Companies that spend money on web ads on Google, Facebook, Instagram, etc. are underwriting most of the R&D budgets in ML/AI. US taxpayer dollars sent to academia are a bit player (as are Canadian taxpayer dollars and funds from taxpayers in other countries).

Granted, it should be acknowledged that the nature of R&D is different in academia and industry. Still, the state of affairs is interesting to note.