The problem with Amazon is not that it doesn’t make a profit, but that you don’t actually know what the profits are.
On the face of it, this sounds like an absurd statement - it doesn’t make any profits. After all, look at this chart - massive revenue growth, zero profit.
If we’re going to do this properly, of course, we should look at free cash flow, or rather (to smooth out the seasonality) trailing 12 months free cash flow (which is also Amazon's preferred metric).
But, of course, this doesn’t show any profits either - in fact what (relatively) little FCF there used to be has now been reduced to close to zero by capex (though this also includes a one-off $1.4bn for purchasing Amazon's headquarters building in Seattle in Q4 2012). Amazon is, very obviously, reinvesting every penny that it can squeeze out of the business back into growth, in pricing, market expansion and capex.
This leads to two views of the company. One, to put it crudely, is that at some point, when it has gained enough market share to get away with it, it will ‘flip a switch’, put up prices or cut capex and start making a return.
The other view is that this isn’t actually possible - that Amazon is a sort of Ponzi scheme. It can only grow by running at zero profit - as soon as it puts up prices or cuts capex the business will collapse, and as soon as the share price stops going up all the staff will leave.
It is certainly true that you can’t just decide to change your business model to be be profitable, as Horace Dedui points out with some precision here. But there's another angle to the Amazon story - it isn’t actually one business.
This chart shows the revenue segments that Amazon reports. These are in different industries, at different stages of development, and in different markets. It seems pretty likely that their underlying economics are different too. Not, that is, the FCF or net incomes that Amazon reports after all that re-investment, but the underlying performance of the divisions.
Moreover, even this isn’t the full story, since Amazon is actually a lot more atomised. Most separate product lines have their own internal owner and P&L by country or region (with a lot of internal transparency, incidentally). Some of them fail and get killed, some have only just started and some are doing very well.
Meanwhile, Amazon is constantly creating new business lines. When they start, like any new business, they're loss making. But they don't 'flip a switch' to get to profitability - they just grow and execute, like any other business.
This means that Amazon's earnings are actually driven a mix of four overlapping factors:
Capex in new distribution
'Artificially' low prices
Operating losses at new ventures
Offsetting profits at established ventures
We have very little idea, from outside, what the mix is. All we know is that Bezos diverts any profit that arrives at the bottom of the P&L back into these to keep the final result at zero. But at least two offer a switch that can be 'pressed' for profit without any damage to the business or any conceptual problem.
To put this another way, Amazon is LOTS of different startup ecommerce businesses on one platform. All the profits from the ones that work are spent on new, loss-making ones.
This prompts two analogies. The first is GE, which also managed its earnings with great care, so that it was hard for anyone outside really to tell what was going on.
The other is Marxism. As Karl Popper pointed out, the problem with the Marxist theory of historical materialism was that you couldn't disprove it. Any scientific theory is susceptible to disproof: something could happen that would show it to be untrue. But there isn't anything that would disprove Marxism - the Marxist can always say 'the conditions for revolution weren't right, but just wait, it will come next century'.
Equally, the problem with saying 'we can't tell from outside how Amazon is really doing, but it will become profitable, just wait and see' is that you could be waiting for ever without ever knowing if you're wrong.