The rapid expansion of AI related infrastructure by the big-tech hyperscalers, and the effect this capital expenditure has on free cash flow, has been a recent focus for equity markets. But does the capital expenditure reported in financial statements provide a complete picture?
We explain how metrics commonly used by investors understate capital expenditure and overstate free cash flow. Only by ensuring that the effects of leasing and other ‘effective’ flows are fully reflected in free cash flow will this important measure be relevant for equity valuation.
We first looked at free cash flow and capital expenditure of big-tech in 2019. A hedge fund invited us to present to its analysts about various accounting and analytical issues, one of which was the measurement of free cash flow (FCF). They particularly wanted to discuss Amazon, and the three different non-GAAP measures of free cash flow provided by the company in its annual and quarterly reports. The question put to us was which is the most relevant measure of cash flow to use in their analysis.
Our recommendation was actually not to use any of them, and we provided an alternative approach that we considered (and still do) provides a more comprehensive view. Our version includes a more complete measure of capital expenditure and other ‘effective cash flows’ that we think should also be components of free cash flow.
Since 2019 we have continued to monitor the free cash flow metrics presented by Amazon and have used the data in several other presentations to investors. Considering the current heightened interest in reported and planned capital expenditure on AI data centres by big-tech, we thought it time to revisit the topic.
Three cash flow measures presented by Amazon – but only one in 2025
Although Amazon has presented three versions of free cash flow in previous years, in 2025 they have changed to only disclosing one version. We think what remains is the least relevant of the three. More on our views later, first we explain what Amazon disclosed pre-2025, and the interaction of these metrics with lease accounting.
Amazon non-GAAP free cash flow metrics
Free cash flow is an important metric in equity analysis, being the basis for free cash flow yield and for discounted cash flow valuations. However, FCF is not directly found in the cash flow statement or defined by accounting standards, leading to companies and investors using multiple different versions. The closest reported measure in the cash flow statement is operating cash flow, but this is incomplete because it fails to include the capital expenditure needed to support operations.
The differences between the three (pre-2025) free cash flow measures provided by Amazon concerns the treatment of leases and whether, and how, lease cash flows are included.
- Amazon FCF version 1: The first (and in 2025 the only) non-GAAP free cash flow measure presented by Amazon equals operating cash flow, less capital expenditure (net of proceeds from disposals), both as presented in the cash flow statement. Operating cash flow is the standard US GAAP measure, including interest and taxation (unlike IFRS reporters where operating cash flow often excludes interest payments). Capital expenditure is the cash paid in respect of asset purchases and does not include the principal payments under finance leases, which appear as financing outflows in the cash flow statement. Capital expenditure also excludes any increase in leasing right-of-use assets. However, this free cash flow measure does include the interest element of finance lease payments and the full rental payments for operating leases, considering the ‘single lease expense’ method of presentation under US GAAP.1Finance leases are those where the lease term is for most or all the useful life of the asset and which are, in substance, equivalent to purchasing the whole asset. Operating leases are all other leases – they are characterized by a material amount of asset risk remaining with the lessor. The single lease expense approach applied to operating leases under US GAAP involves capitalising the leases in the same manner as for finance leases but, instead of reporting interest accretion for the liability and depreciation for the asset, the lease rental is itself reported as an operating expense. We think operating lease accounting in US GAAP is confusing and investors are better served by the IFRS approach. For more about lease accounting differences, see our article ‘Operating leases: You may still need to adjust’.
- Amazon FCF version 2: The second approach is the same as above but, additionally, with the deduction of the principal payments for finance leases (and other similar financing arrangements) that appear as financing flows in the cash flow statement. This approach has become very common, particularly for IFRS reporters for which there is no distinction between operating and finance leases, and all lease payments are classified as financing flows rather than operating cash flows in the cash flow statement2It is not quite all – very short-term leases are not capitalised at all, and some lease payments that are deemed variable (apart from those that vary with changes in inflation or interest rates) are excluded from the amount capitalised. These payments are included in operating expenses under both IFRS and US GAAP.. This second approach corrects the problem with version 1 where there is no free cash flow impact of capitalised leases (but just the finance leases for US GAAP), even though these flows are closely related to fixed assets and capital expenditure, and regarded by many as, in effect, operating flows. We prefer version 2 to version 1.
- Amazon FCF version 3: The third version that Amazon previously presented is unusual because it includes what is, strictly speaking, a non-cash item. In this, the principal payments for finance leases of equipment are no longer deducted as in version 2 (although principal repayments for property leases are still deducted). Instead, the equipment leasing ‘capital expenditure’ – the new right-of-use assets for leases originating in the period – is deducted. When leases are first recognised as an asset and liability, there is no actual cash flow because two ‘flows’ offset and cancel. The effective inflow due to new debt (lease) finance offsets an effective outflow for the ‘purchase’ of a right-of use asset. However, capital expenditure and financing flows are treated differently for free cash flow, which is why the effective leasing capital expenditure is deducted. We prefer version 3 to either 2 or 1.
New leases involve 2 offsetting transactions – new debt and the purchase of fixed assets
We call the lease capex deducted in version 3 an ‘effective cash flow’. It is what the capex would have been had the single lease transaction actually been two separate transactions – the raising of finance and the purchase of an asset. Applying this approach means that the cash flow effect of assets acquired through leasing is more comparable with the cash flows from outright asset purchase. However, the acquisition of a right-of-use asset may be significantly less than the cost of an asset purchase which arguably still limits comparability.3The two still differ in that only the right-of-use portion of assets is recognised under lease accounting, When the revised lease accounting standards were developed, some investors favoured the use of the ‘whole-asset’ approach to lease capitalisation. The amount capitalised would equal the outright purchase price, and a liability reported for (not just lease payments) but also for the obligation to return the asset at the end of the lease term. This approach did not gain much traction amongst accountants, but it would result in greater comparability of free cash flow measures and would also negate the current advantage of off-balance sheet structures. We think there is some merit in investors using a whole-asset approach for analysis, although this is difficult to implement considering the limited information available.
Amazon non-GAAP Free Cash Flow Disclosure 2024
Version 1 – Free Cash Flow
Version 2 – Free Cash Flow Less Principal Repayments of Finance Leases and Financing Obligations
Version 3 – Free Cash Flow Less Equipment Finance Leases and Principal Repayments of All Other Finance Leases and Financing Obligations
Amazon 2024 10k
Versions 2 and 3 above both allow for the cash flow effects of finance leases, albeit with different interpretations of that cash flow. Version 1 excludes any cash flow effects of finance leases, and therefore versions 2 and 3 must always be lower than (or equal to) version 1. In the last couple of years, Amazon’s 3 different measures of free cash flow have converged somewhat due to a reduction in the use of finance leases. In earlier years the differences were more pronounced, and this could well return if, for example, Amazon uses more finance leases for its AI capital expenditure expansion.
On average, versions 2 and 3 will be the same because the lease capital expenditure must equal the sum of future lease principal payments. Version 3 free cash flow will be lower than version 2 if the use of equipment finance leases is growing, which was the case for Amazon through to 2019 (see the chart below). Interestingly Amazon started to roll back its use of finance leases around the time that new lease accounting rules were implemented and, from 2020, version 3 has been higher than version 2.
10-years of Amazon free cash flow data
Amazon 10k filings 2016 to 2025
From 2025 Amazon has only disclosed the version 1 free cash flow measure. We assume this is because finance leases have become less relevant, and to simplify their accounting. In our chart above we have not tried to recreate what the other two measures would be in 2025 although, as for 2024, we do not think there is a significant difference between the measures.
Amazon non-GAAP free cash flow disclosure 2025
Amazon 2025 financial statements
We think version 1 (excluding all finance lease flows) is conceptually incomplete and should not be used by investors. This is even more important for IFRS reporters where there is no concept of finance and operating leases, and the cost of all leases would be omitted if version 1 were applied. Fortunately for investors, most IFRS reporters that provide a measure of free cash flow in their investor communications use version 2 above.
Our preference is for version 3 (although we would expand it, as we explain below). We think the inclusion of the effective flow for leasing capital expenditure results in more comparability and a more relevant measure that is consistent with leasing being a form of financing, and lease payments being financing flows.
Our alternative approach to free cash flow
Although we think that the (pre-2025) third method of determining free cash flow applied by Amazon is superior to what we see from many other companies, we think some additional adjustments are required. The first is to apply the same lease capital expenditure approach as employed by Amazon, but for all leases; the next concerns an additional form of financing for asset purchases, and the final is an adjustment to a different component of free cash flow entirely, but with similar underlying logic.
(1) Apply the lease capital expenditure adjustment to all leases
The adjustment by Amazon to recognise the new right-of-use assets for equipment finance leases as an ‘effective’ cash flow (used in their old free cash flow version 3) is a good approach. We think this produces a more realistic measure of capital expenditure and is consistent with the underlying concept of free cash flow and how this is used in DCF valuation4DCF valuation can be done using either the version 2 or version 3 cash flow presented by Amazon but not version 1 which will result in an overstatement of value. Using version 2 or 3 will produce the same result if the other aspects of DCF (discount rate and enterprise to equity bridge) are consistent with the definition of free cash flow. See our article ‘DCF valuation models: Have you updated for IFRS 16?’.. However, we think the adjustment should be applied to all leases and not just finance leases of equipment. We also think the same should apply to operating leases.
Extending the approach to all finance leases is straightforward, given the supplementary disclosures to the cash flow statements provided by Amazon. Simply deduct the line item “Property and equipment acquired under finance leases” that is disclosed as supplemental data to the cash flow statement, so for 2025 it is the $2,911m, as shown in the note below. There is no further adjustment needed.
Amazon supplemental cash flow information disclosure
Amazon 2025 financial statements
For operating leases, it is a little more complicated. The line item “Assets acquired under operating leases” ($19,930m in 2025) is the same as that for finance leases, but the two types of leases are presented separately because the rest of the accounting differs. Under US GAAP, even though operating leases are reported as lease liabilities and right-of-use assets, the accounting in profit and loss and the cash flow statement is different.
US GAAP ‘single lease expense’ approach make the calculation more complicated
Operating lease accounting under US GAAP uses the ‘single lease expense’ approach whereby the rental expense is reported in profit and loss instead of the depreciation and interest that is recognised for finance leases. In the cash flow statement, the cash rentals paid are included in operating cash flow but, unlike for finance leases, these are not split between the principal and interest components.
If you wish to calculate a pre-financing enterprise free cash flow, the lack of this split does not matter because both principal and interest need to be excluded. However, the Amazon free cash flow is a version of an equity cash flow, which is after deducting interest. Therefore, for our adjustment, we need to add back the estimated principal component of the operating lease cash flow and deduct the operating lease capex (assets acquired under operating leases), as shown in the table.
The principal component of the operating lease cash flow can be estimated by removing an estimate of the interest accretion equal to the opening operating lease liability multiplied by the lease discount rate which is one of the required disclosures.
We should note that under IFRS things are easier because all leases are accounted for in the same way that finance leases are in US GAAP. This means there is no added complication of trying to disaggregate the operating lease cash flow into interest and principal components.
The result of this and our other adjustments can be seen in the table below.
(2) Adjust for supplier financing of fixed asset purchases
Our second adjustment is somewhat unusual, and not something we would normally include in free cash flow because the amounts involved are usually not material. However, we think it is conceptually appropriate and, in the case of Amazon (and for any high growth / high capex / long supplier credit businesses), rather more significant. The adjustment concerns the amounts owed to suppliers in respect of fixed asset purchases.
Most purchases of assets (or payments for expenses) involve a credit period. In the cash flow statement, the purchase of fixed assets should be the cash paid in the year in respect of fixed assets, but some of this will relate to purchases made in the previous period. By offering credit the suppliers of fixed assets are providing financing to the purchaser. The initial flows arising from this financing do not appear in the cash flow statement because, as in the cases of leases, they offset.
Asset purchases on credit are in effect two offsetting ‘flows’
When an asset is purchased on credit, the transaction is non-cash but comprises two offsetting ‘flows’ – the purchase of an asset (a capital expenditure outflow) and a short-term loan from the supplier (a financing inflow) that is repaid when the credit is settled. We would argue that the most appropriate capital expenditure amount is the initial purchase rather than the later cash settlement of the payable. Of course, generally this is not going to make much difference given that the amounts are the same5Generally short-term payables are not discounted. If they were the capital expenditure would be lower than the payment to settle the trade payable. For more about working capital flows in equity analysis, see our article ‘Receivables, payables and (hidden) financial leverage’. and it just shifts some of the capital expenditure from one period to another. However, if the credit period is long and, particularly if growth in capital expenditure is high, the effect of making this adjustment can be more significant – which is the case for Amazon.
The payable in respect of fixed asset purchases is disclosed by Amazon in the property and equipment (PPE) footnote (see the last paragraph in the following extract).
Amazon property and equipment disclosures
Amazon 2025 10k
What we are interested in is the change in this payable, which for 2025 is $10.2bn ($27.0bn – $16.8bn). This is the amount by which the fixed asset purchases in 2025 exceeded the cash paid in respect of fixed assets in the same year. When the payable increases we reduce our measure of free cash flow by the difference. The result is the same as if Amazon had purchased the assets for immediate cash payment and then separately taken on a short-term loan. Amazon also provides this change in the payables amount in the supplementary data for the cash flow statement, which we show above (See the line “Increase/(decrease) in property and equipment acquired but not yet paid” – $10,155m in 2025).
Amazon only started separately identifying the payable related to PPE in 2025. For our adjusted data we have estimated the amounts in prior years based on our estimate of the average days payable period for 2024/25
(3) Include the stock-based compensation effective cash flow
Our final adjustment has nothing to do with capital expenditure; it is an adjustment to the reported operating cash flow. Like the initial recognition of leases, a stock-based compensation expense is not cash. However, it also comprises two offsetting flows, only one of which is an operating flow. The non-cash grant of shares and options to employees is economically equivalent to paying employees cash (an operating expense and cash outflow) and the employees reinvesting their cash in the purchase of shares and options (a financing cash inflow). It is the (effective) operating outflow that we think should be included in free cash flow metrics.
Stock-based compensation is both a valid expense and an ‘effective’ cash flow
Unlike leasing, stock-based compensation will never result in an actual cash flow (unless the compensation is cash settled) and therefore the separate recognition of the offsetting effective flows becomes even more important in free cash flow calculations. Some may still argue that the lack of a cash outflow for equity settled stock-based compensation means it is less burdensome than other cash components of employee remuneration. We think this topic is now much better understood and few investors continue to believe that stock-based compensation is somehow a ‘free lunch’. To read our explanation of why this is a valid expense, see our article ‘Dot-com bubble accounting still going strong – Tesla’.
The adjustment to recognise the effective operating outflow for stock-based compensation is usually straightforward considering that the amount should be shown on the face of the cash flow statement – see below.
Amazon operating cash flow section of the cash flow statement
Amazon 2025 10k
An alternative approach to dealing with stock-based compensation in free cash flow is to include the cash cost of purchasing shares to offset the dilution that results from share and option grants. Although this is an actual rather than effective flow, we do not think it is the right approach. Not all companies chose to offset the dilution, which means the cash cost is potentially incomplete, and the approach also fails to reflect the optionality cost involved in granting stock options. We think our ‘effective cash flow at the time of grant’ approach is superior.
How our adjustments change Amazon free cash flow
Here is a summary of our adjustments to the free cash flow over the last 3 years. Below this is the 10-year view, including the additional versions of free cash flow that Amazon presented up until they were discontinued in 2025.
Summary of our adjustments to Amazon free cash flow
Amazon 10k disclosures and The Footnotes Analyst estimates. References relate to the numbered adjustments explained above. Assets acquired under lease arrangements include finance leases, operating leases and build-to-suit financing arrangements.
10 years of Amazon free cash flow compared with The Footnotes Analyst alternative
Amazon 10k disclosures and The Footnotes Analyst estimates
Our adjustments clearly make a significant difference. Indeed, over the 10-year period, the aggregate free cash flow as presented by Amazon (their approach 1) is positive $161bn (it is somewhat lower, albeit still positive, for their now discontinued alternatives), whereas the aggregate of our adjusted free cash flow over the same period is negative $149bn, with only 2023 and 2024 showing a positive free cash flow.
We would use forecasts of our adjusted free cash flow for DCF valuation and in free cash flow yield metrics. However, remember that negative free cash flow is not necessarily an adverse investment signal. Indeed, for a capital intensive business like Amazon we would expect to see negative cash flow during a period of high growth – the revenue of Amazon has increased by a compound 18%p.a. over the last 10 years. What matters for valuing this company is when and to what extent you think it will produce positive cash flow in the future.
Following our usual practice, we do not provide any opinion about the financial results of Amazon, the value of the business, or the merits of Amazon as an investment. Nor do we suggest that there is anything wrong with their financial reporting, indeed quite the opposite, we think the company provides a lot of transparency. Our analysis is educational and designed to highlight an alternative calculation of free cash flow that we think is more relevant for equity analysis.
Insights for investors
- Ensure that either the principal repayments of US GAAP finance leases, or the amount of new right-of-use assets ‘acquired’ in the period, is included in free cash flow. Do not simply define FCF as operating cash flow less capital expenditure.
- Not omitting lease flows is even more important when analysing IFRS reporters, for which all leases are reported in the same manner as finance leases under US GAAP.
- We think it is preferable to include lease capital expenditure ‘effective’ cash flows in free cash flow rather than the principal repayments of a lease liability. This ensures greater comparability across companies that adopt different ownership / leasing strategies.
- Like the initial recognition of leases, equity settled stock-based compensation does not involve an explicit cash flow, but does comprise offsetting flows that have different characteristics. We think the effective operating outflow, but not the offsetting financing inflow, should be included in free cash flow metrics.