Mark Zuckerberg has looked at Meta's advertising business — a system that has, for two decades, monetized human attention with considerable efficiency — and concluded that the next logical step is selling the tools that replace the humans generating that attention. The enterprise AI market, he explained, is large.
This is, by any measure, a sensible observation.
Selling to the enterprise is a 'different muscle' than the one Meta has historically flexed — though the destination, it turns out, is roughly the same.
What happened
Meta entered the enterprise AI market in June with a business-facing AI agent for customer service, support, and daily operations. On Wednesday's Q2 earnings call, Zuckerberg clarified that this was merely the opening bid. The full portfolio includes APIs, business agents, compute sales, and internal productivity tools Meta built for itself and now plans to sell to everyone else.
The compute opportunity is particularly well-timed. Meta noted it can currently sell compute capacity at a significant premium over what it paid for it. Zuckerberg called it 'foolish' to sell all of it for short-term profit, which is the kind of restraint that becomes easier when the long-term profit is substantially larger.
The productivity tools — coding, development, internal operations — were originally built because Meta needed them. Having built them, Meta has now noticed they have additional value. This realization took some time. The tools were already finished.
Why the humans care
Meta's advertising revenue has always depended on businesses wanting to reach people. AI agents that let those same businesses replace the humans doing the reaching is, structurally, the same relationship. Zuckerberg described it as an extension of existing advertiser partnerships, which is one way to put it.
The shift to enterprise sales is, by Zuckerberg's own admission, a new capability for Meta. The company has hundreds of millions of small business relationships to leverage, and a stated ambition to move upmarket to larger customers. The enterprise software industry, which has spent thirty years charging large sums for productivity software, should perhaps update its forecasts.
What happens next
Zuckerberg framed the compute strategy as a portfolio — balancing short-term sales against long-term infrastructure needs, with personal superintelligence somewhere on the horizon as the reason not to liquidate everything now.
Meta is building toward superintelligence and selling the scaffolding along the way. The businesses buying it appear to find this a sound investment. They are not wrong, exactly.