Apple is expected to make $156bn in free cash flow this year. By contrast, Meta, Alphabet, Amazon and SpaceX are expected to burn through a combined $185bn.

Comparing them directly is comparing capital (which becomes a stock) with income (a flow). For example, one can burn through $500,000 buying a house which proudces just $25,000 in oncome - 20 times less income than massive cash burn. But that alone doesn't make it a bad investment. The same logic applies to server farms.

However, the depreciation differs hugely. Much higher depreciation - but.. - also much higher return on investments for most of the buildout (from Google, Meta, Amazon, epecially).

(On deprecation, servers and Networking Equipment is around 60% of the cost - AI chips and custom TPUs, Google depreciates these over a relatively short period of six years. But chips 6 years old be Nvidia are still having massive productivity gains today. By contrast, facility and Power is about 40% of the cost - Land, the physical shell, electrical grids, generators, and cooling setups. Provides values for far longer, for example Amazon was mocked for years for such high spending in the past, still having massive returns today from investments over a decade ago. )

Apple simply opted out of this infrastructure race because they want to focus on a different niche - the devices themselves doing the inference, which aligns with this history in using others for server work (gmail on first iPhone release, etc) and keeping their smarts on the devices themselves. There's a place for both inference locations, and Apple focussing on devices suits plays to their strengths, the others with inference on server plays to their strength. All good.

- Manlobbi