This is part one out of three in a series about the implications of AI for work. I’d love your feedback / thoughts / opinions!
Happy long weekend {{first_name|friend}}!
This is Matt. Matt is a total nerd but we love that for him.
Matt was born in 2001. He just got a new job - one of his first jobs out of college - and posted on LinkedIn about it. Congrats Matt!
I love how casually Matt made this post, like it’s a new job update anyone could share.
But this was no ordinary job update.
This summer, Mark Zuckerberg called Matt to offer him a job - and a $125mil pay package over four years - helping to run Meta’s AI research efforts. (Matt knows a lot about AI.)
Matt texted his friends about to ask what he should do. They advised him to turn it down. So he did. Typical early twenties stuff, gotta lean on your friends to help you decide about nine figure job offers!
Shortly after, Zuckerberg met with Matt in person and offered him $250mil over four years, with $100mil paid in year one. Matt texted his friends again and this time they told him to take it.
This pay package is 16% larger than Steph Curry’s new contract. And while Curry fills stadiums with people who want to see him play, no one is paying to come see Matt play. Except for Zuckerberg who has apparently rearranged the office so that his desk is close to Matt and the other newly hired nerds who have all received similar compensation.
Okay so…why would Meta want to spend this much on Matt and his new coworkers? Perhaps it’s all FOMO and we’re in a hype bubble. But let’s just assume for now that there are a bunch of smart people who think they are going to make that money back, and more.
Part of the answer can be found at the site of this new construction project getting started in an area called Richland Parish in northern Louisiana:
This patch of ground will turn into Meta’s Hyperion Data Center. Here’s the size Hyperion in comparison to Manhattan:
Meta is spending $10bil to build this data center, part of $70bil or so of capital expenditure this year.2
So if you’re spending $70,000,000,000 on data centers for AI, I guess it makes some sense to spend 5-10% of that amount on compensation for the 50 or so humans who will tell the data centers what to do all day. Even one week of going in the wrong direction could throw off corporate progress for a long time to come.
In my business, if I had a $100mil hiring budget I would hire 200 people at $500k/year each which would enable me to more than double the salaries of all of the best professionals in my field and then we would have a team of champions. There are very very few industries or even moments in time when it makes sense to spend all of that budget on one person, and it indicates a lot about what’s happening in this industry and this moment in time.
Namely, a very small number of people at a very small number of companies hold an absolutely insane amount of power right now. Perhaps over time these concentrations will decentralize - the same way the huge steel or railroad companies of 100 years ago held immense sway until they didn’t - but at least for next several years or decades, we are living in a time of immense technological power held by very few people.
So on this Labor Day weekend - when we celebrate the importance and accomplishments of labor unions over the past 200 years - it’s worth pausing to consider some implications here.
Labor unions historically arise in response to moments of great technological change, when the owners/managers of companies were making a new set of decisions about the future of their workers. We are obviously living through another one of these great change moments.
What kinds of unions will form?
How might workers try to balance power with their employers in this moment?
I’m an optimist and I love technology. I want to believe that all these big bets are going to work out great and our AI-enabled life will be awesome.
But technological leaps don't automatically translate into shared prosperity. They require deliberate choices about who benefits and how. I don’t see our big tech leaders making those deliberate choices right now.
How different would it feel if we saw Meta renovating five public schools for every Matt they hire?
Or building beautiful houses for people who need them at the same rate they are building houses for computer chips?
Not even “superintelligence” can solve the problem of what kind of world we want to live in. That’s a question for us humans to answer - and then put in the labor to make it real.
1 Bloomberg points out that there is no way the data center will actually be this large in terms of cubic feet, that the visual is meant more as a way for Zuckerberg to emphasize that it is a Large and Important project that New Yorkers should care about.
2 That $70bil of capex that Meta is spending is part of a much broader massive tidal wave of spending on data centers to support AI, currently estimated at $400bil this year. This wave of spending is so large that as of July it had a bigger impact on GDP growth than consumer spending. In other words: big tech’s spending on AI infrastructure is propping up the entire U.S. economy. This helps explain why high-level economic numbers continue to be generally positive but consumer sentiment is down.
