> A city is a natural sharding point. No no, wrong type of scale. I'm talking economies of scale. Uber needs to pay N SFBay-area engineers upwards of $200,000

> A city is a natural sharding point. No no, wrong type of scale. I'm talking economies of scale. Uber needs to pay N SFBay-area engineers upwards of $200,000 USD/yr (including health expenses, employer-side taxes, and other overhead) for N years in order to build a well-functioning, city-scale software ecosystem - including server software, devops and maintenance tools, payment provider integration, antifraud efforts around payments, client software on multiple app stores, internal tools for the support team, business tools for the traveling consultants who want easier expense report receipts, all sorts of fun data collection for the business analysts, and all the ongoing maintenance and operations costs associated with that mountain of code. With all that effort expended, all those millions spent, they have a bunch of software they can deploy to one, single city and do a top-notch bang-up job of it. But now this $N million piece of software can be deployed to its second and third and three-thousandth locality with little more than extra computers, some clever sharding work, and maybe some localization: a puny fraction of the original development effort. If they need a whole new country, throw in some more internationalization/localization and maybe a new payment provider integration. Those are the economies of scale. Every feature that Uber adds for $N in developer time is hundreds of times more impactful than if Single City Carshare [tm] were to add it, even if they can use cheaper developers outside the Bay Area. Of course, there are diseconomies too (if only because of managing additional headcount on the per-city operations side of things), but as you said, cities are a natural sharding point, and this keeps those diseconomies down.