I guess the point I was trying to make was that at some point, especially as growth slows, it is likely that Amazon will switch to be valued based upon earnings and not revenues. All of the parts you are summing are based upon revenue multiples and not earnings (eg the 0.6x WMT comp - Walmart throws off tons of cash and is valued as a multiple of earnings/cash so I'm not sure a revenue multiple using WMT is the right comp. Same with EBAY).
Your revenue comps and values are very relevant as long as AMZN continues to grow rapidly and investors value it based upon growth and not earnings. The point at which growth slows, a lot of the valuation argument switches from revenues to an earnings based approach (and you see a bit of that in the recent sell off which was because the revenue expectation is much lower than people were expecting). That's a tougher way to back into rationale for today's price.
My point is that I think it's likely we fill 95% of the valuation before accounting for growth at today's price, and I'm more interested in economic value...
Given that the stock is tanking, with no end currently in sight, a good entry point to add or to buy may be opening. That's the impetus behind today's post.
Using multiples in general is a shortcut to a full-blown DCF, which may be impossible to do accurately for each part as cost of equity should change from unit to unit.
The multiples I used in that dirty SOTP valuation were at substantive discounts to a multiple that could be attributed to Amazon's businesses. Certainly, Amazon's etailing business is first-class and it's unlikely to lose its footing by any incumbent, Walmart or otherwise, in the foreseeable future.
Does this mean that Amazon's retail business should be valued at the same multiple as WMT? Absolutely not, but we need guideposts. Same for the 3p business, which likely has better margins than eBay's given the scale derived from the synergies of AWS and the 1p operation.
Another point is that the video business has value. Every dollar invested? Who is to say, but that Netflix has subsidized AWS, and now Amazon Video is outspending big media firms says something.
If Amazon Video started selling content separate from Prime, what valuation do we ascribe? While the sector is somewhat depressed, FOXA trades at 4.3x book. If Amazon spends $4B this year in content, is it worth $16B more next year? Is it worth $160B in 10 years if AMZN spends $4B every year? What about the video content that used to be written off in past years? It's new accounting to capitalize since the accountants realized that video will likely be earning rather than losing.
Another point is we can dig deeper and come up with fairly good margin assessments based on value of goods sold on a relative basis to get a better feel for the retail business.
For example, AmazonBasics relative to other large players. Energizer, for example, generates a 46% gross margin and an 11% net margin. Add back 50% of ENR's depreciation (capex was $25m last year and deprec was $50m), and net income grows from $200m to $225m. $225/85m = 265% cash returns on equity?! ~13% owner's income margin?! I don't see a reason why Amazon isn't doing just as well except those margins should be wider since AMZN can leverage its scale to cut out the retail chain to sell paper shredders and batteries.
Also, AWS @ a 5x multiple on sales is very conservative since RHAT is selling for 11x to IBM, and RHAT is not a leader by any stretch. Maybe in this light, AWS could be worth 12-13x sales (or more) when taking runway into account.
Depreciation is an interesting item as well: TTM $14.6B. How much is a cash loss?
Hardware depreciates on a faster schedule than the real life of said hardware. And back-of-env shows that $14.6B of depreciation bumps from last year's years' investment in PPE of $10.8B. In other words, a better DCF would also account for the difference between deprec and replacement.
Considering how savvy the ppl at AWS are, my guess would be double the depreciation cycle (ie 6 years--could be longer) before becoming obsolete and requiring replacement would be somewhat conservative. Switches can last a very, very long time...especially ones that are built to custom spec. Also, it's likely that parts are "dumbed down" so that they are running ram, cpus, fpas using some figure under voltage requirements to reduce wear. Since they have so many units working in tandem, increasing the number of threads can offset performance losses.
That said, parts will depreciate over a 3 year period for tax but 6 in real life--the difference should get added back over the last 3 years, which every year has been growing along with AWS. This logic could put $7B of deprec back into owner's earnings from last year TTM Sept 30, and another $5B from TTM Sept 2017, etc.