Amazon’s Cash Flow


We published a blog on Amazon around a year ago where we explored the issue of free cash flow. We explained then that Amazon uses three different definitions of free cash flow and why we use a different calculation. Feedback from readers was that this was quite surprising, given that the free cash flow yield is considered one of the most important valuation parameters, and many readers asked if we could delve deeper into Amazon’s cash flow statement and look at some of the factors affecting operating cash flow. Here goes…


Amazon’s 2020 Q4 earnings release began:

“Operating cash flow increased 72% to $66.1 billion for the trailing twelve months, compared with $38.5 billion for the trailing twelve months ended December 31, 2019.”

The next three bullets list the free cash flow improvement in the three different measures Amazon uses $31bn, $20bn and $21bn, depending on which definition you prefer (see our first Amazon blog).


The table shows that Amazon’s operating cash flows increased 72% year on year, slightly less than the 84% increase in earnings. I have shown both the year in year increase and the proportion of 2020 operating cash flows for each line item in the table.

table 1.png

Most companies have to invest in working capital as they grow. Unusually, at Amazon, 80% of the cash was generated from operations and another 20% was generated from working capital – the company enjoys a negative working capital position, and this gives it two significant operational advantages from an analytical perspective:

–          Negative working capital shrinks the total capital employed and enhances returns on capital.

–          Negative working capital reduces the investment required for growth.

Investors are highly attracted to companies with negative working capital as it boosts the two most important financial metrics for the business – returns and cash generation. We shall return to this extremely important subject later and explain why it’s almost always helpful but can in certain instances be quite dangerous.

First, let’s look at the cash generated before working capital. There are three main elements, plus a balancing other item:

Net income – this is the earnings generated by the business and is the common benchmark against which operating cash flow is compared.

Depreciation and amortisation – for Amazon, this is even higher than earnings because of the capital intensity of the business and the low profitability currently, although margins have reached 5.9% (5.2%).

Stock-based compensation – Amazon charged $9.2bn last year for stock based comp. Most CFOs add this number back when calculating adjusted earnings and I explained in a detailed blog on the subject why the calculation was flawed and the accounting treatment should be modified by analysts – a better reflection of the economic impact is what it would cost to buy back the stock issued to employees in order to cancel out shareholder dilution.

This is generally significantly higher than the P&L charge and this should be factored into the analyst’s calculation of the sustainable operating cash flow. Amazon appears to be sensitive to the issue of dilution – the fifth bullet in the earnings release (after the first four on cash flow) reads:

“Common shares outstanding plus shares underlying stock-based awards totaled 518 million on December 31, 2020, compared with 512 million one year ago.”

It’s quite refreshing to read an earnings release with headlines that don’t start with a non-GAAP earnings number and flagging the share count is a signal that management care about dilution. But a net 5m shares have been issued in the year – the share price averaged $2681 so the real economic cost was c$13bn, vs the $9bn charge in the P&L, and this is significant relative to the free cash generated.

Other items – the main other adjustments here relate to tax and other expense – a $2.6bn cash outflow last year, which is not reflected in earnings. I usually monitor this number closely as it can reflect the release of accruals, a ploy used by CFOs to pad out earnings, often to meet eps targets enshrined in an overly generous bonus calculation. This is not that material to Amazon but note the sharp increase year-on-year.

The second element of this is the tax adjustment. When comparing cash generation to earnings, I prefer to take tax out of the calculation. If I am trying to understand the quality of earnings through the lens of cash generation, I find that it’s more effective to do so on a pre tax and pre-interest basis.

In the past interest varied little between P&L and cash flow, but accounting standards have for some time required the calculation of long term liabilities in present value terms. This means that each year, as you get one year closer to the crystallisation of the liability, the liability needs to be discounted by less, resulting in a notional interest charge in the P&L. In my mind, this muddies the water when looking at the results and the cash vs earnings comparison which is one of my key accounting red flag detection tools. I therefore prefer to do the calculation at the ebit level.

Amazon’s tax charge increased from $2.4bn to $2.9bn ie +$0.5bn on a pretax profit increase of $10bn. There was a decline of $0.5bn in the amount of tax deferred in the cash flow vs a $0.8bn increase the previous year. The way to think about this is that Amazon generated $0.8bn more cash last year because taxes were deferred and $0.5bn less this year because more tax was paid. But the tax rate in the P&L fell from 17.0% to 11.8%. I haven’t investigated this, but I doubt that Amazon will be able to maintain this level, as politically it’s not expedient.


Amazon reports working capital under 5 headings:

Inventories – this is straightforward, but the conventional calculation of inventory days vs sales or cost of sales are less helpful indicators as increasingly Amazon’s business is as a third party service provider, rather than a retailer.

Accounts receivable, net and other – the basic retail business should have minimal accounts receivable as the goods are purchased in advance, although there will be credit card processing delays. Some of the AWS business may be transacted on conventional credit terms, while some is on a subscription basis, payable in advance, and this is in unearned revenue.

To look at Amazon in detail, it would be important to understand the payment characteristics of each of the businesses, but if I were just taking an initial look – a 90 minute Test the Hypothesis check per the methodology in my book, The Smart Money Method – I would simply add the accounts receivable and inventory together, and compare that to the total sales – that should give a rough impression of the trend in this element of working capital.

Source: Behind the Balance Sheet from Sentieo data

Source: Behind the Balance Sheet from Sentieo data

This is quite an interesting chart as it’s not what I would have anticipated – Amazon is getting less efficient in terms of working capital required to drive a unit of sales – the service businesses are perhaps more working capital heavy? (I have no particular insights into the specifics of Amazon – the last time I wrote a report on the company was 2014/15. The purpose of this blog is to illustrate the methodology).

Source: Behind the Balance Sheet from Sentieo data

Source: Behind the Balance Sheet from Sentieo data

Accounts payable – the payable days have also gone up, so Amazon’s overall position has not deteriorated. There was a particular pickup last year, which has taken it beyond the 90 day mark on our estimates (using Sentieo data). This is good short term, but we question whether it’s sustainable for Amazon – not paying your suppliers for well over 3 months (we are comparing accounts payable to sales, not cost of sales) is not going to win Amazon many friends. Interestingly, no questions were asked in the analysts’ calls on the last two annual results about cash flow or working capital – it’s all about the next quarter’s earnings.

Source: Behind the Balance Sheet from Sentieo data

Source: Behind the Balance Sheet from Sentieo data

Accrued expenses and other – we would not generally spend too much time on this line but in the case of Amazon it’s a very large number, $44bn or over 11% of sales. These liabilities are related to leases and asset retirement obligations, payroll and related expenses, tax, unredeemed gift cards, customer liabilities, current debt, and acquired digital media content. The level has risen very significantly in the last two years, and Amazon’s disclosure is unhelpful. Last year, there was a $3.3bn increase in payroll related liabilities (headcount was up an astonishing 63%), a $2bn increase in lease liabilities and a $4bn increase in other.

The measurement of this in terms of sales is slightly misleading as the balance includes lease liabilities and short term debt, but to strip these out would involve going back through nearly 20 years of annual reports and we therefore took this short cut. The pickup in the chart thus includes short term debt and lease obligations. We don’t know why Amazon presents these liabilities together as in our view, the presentation could be and should be significantly clearer.

Source: Behind the Balance Sheet from Sentieo data

Source: Behind the Balance Sheet from Sentieo data

Unearned revenue – although this number is often relatively small, it’s important to monitor as it can indicate that subscription levels are rising or falling. This is attributable to AWS and Amazon Prime and there is an additional $2bn of unearned revenue in Other long-term liabilities, presumably relating to AWS. The level has been steadily rising but the growth is slowing on the larger base. Note that even for a company of Amazon’s size, $12bn of extra cash float (including the long term portion) is nice to have.


Amazon’s negative working capital position is an attractive feature of the business and the jump in payables last year is a significant component of its free cash flow. The performance is so impressive that we would question to what degree it would be repeatable, but this would require a more detailed knowledge of the business. Note that the December year-end will be the most favourable point for payables because of the Christmas rush and the payment to suppliers post year-end. The table summarises the working capital position:

Table 2.png

It’s clear that the balance is improving (note that we have excluded from the calculation non trade receivables and accruals). The $21bn improvement is attributable to an increase in trade payables – we wonder how much room there is for further improvement.


One important question we get asked is whether negative working capital has any downsides? And like many factors in investing, it’s great when things are going well, but can turn toxic in a downturn. Supermarkets have negative working capital as do many construction companies. If a supermarket chain has to close a store, there is a cash outflow; when the store was opened the customers pay for the stock before the supplier has to be paid; the reverse happens when the store is closed.

Similarly, when a construction company completes a contract, it needs to find a new contract immediately, otherwise cash received from its customer will have to be paid out to sub-contractors, and its “float” will decline. This is why construction companies often bid aggressively for new contracts when economies slow. It may be hard to envisage growth at Amazon reversing, but a slowdown in growth would therefore be amplified at the level of free cash flow generated. But it would likely continue to boost cash rather than reverse – Amazon is in a much stronger position than a construction company or a travel company. We have seen with Covid just how difficult the unwinding of advance payments can be in that industry.



Amazon’s cash generation is impressive, and the company appears disciplined in its use of stock options and suppliers to finance its working capital. The latter has increased significantly and we would question to what extent Amazon will be able to improve further its working capital position. As with its favourable tax rate, we wonder if the company may find further improvements harder from here, especially if political expediency becomes a more significant factor in strategy.