>
That is a strategy only available to... monopolies.I'm not sure I agree with your characterization of that as a monopoly strategy, and I'm also genuinely curious how you know those are "negative margin" services. To the first, the strategy seems more about vertical integration than monopoly per se. Plenty of businesses aim to make a profit overall, but either have an outright strategy of loss leaders or more commonly have individual services that are absolutely necessary to the rest of the business yet are not profitable in and of themselves. And if the opportunity comes to leverage some of those internal divisions to some degree it can be completely reasonable.
To the first, the console market would be an example of a competitive sector (at least since the mid-90s) where players have quite fairly run one part at an initial loss (hardware in the first few years) with the plan to make it up on add-ons. They can further average break-even or profit over the product lifespan, because the hardware functionality is fixed but will be vastly cheaper to manufacture in Year 7 vs Year 1.
To the second, it's not uncommon at all for a business to require a certain amount of internal hardware or services for operations that is sitting idle a fair amount of the time, generates no direct revenue, or both. This is part of the whole logic behind cloud computing, or the constant tension of critical IT being seen as a "cost center". There isn't much minimal size for that to be the case either, even a small company may easily have a share of underutilized kit. In those cases, the marginal cost can be extremely low, even near-zero. If it can be turned into any sort of useful service or value-add at all with low enough overhead then it's still of value.
So I'm concerned your formulation is overly reductive of any company to its parts. The vertical strategy has real advantages as well as disadvantages, gains that are quite genuinely win-wins which is why they work. For AWS for example, that's a core business for Amazon and also something they require for their own activities. How much of the capacity is sitting idle at any given time so they can be prepared for when actual high priority customers need to spin things up? They certainly require significant margin to respond to variations in demand. Video encoding seems like a great way to fill troughs and average things out, since it can easily be run low priority, and the marginal cost is going to just be the electricity running chips at full vs power saving. Compared to all the cost of having the datacenter and hardware and staff supporting it at all, even just sitting around doing nothing, that's probably a pretty low fraction. Is video coding actually costing Amazon money vs what value they gain, or is it simply an area where they quite genuinely can do it for less, putting otherwise idle necessary capacity to use? And what about Azure or GCP, or even someone just running dedicated encoding hardware? How does Amazon have a monopoly on compute? If they're merely able to do it cheaper thanks to scale, what exactly is the problem? If they raised prices above what competitors or someone could do themselves where is the lock-in for video encoding?