How Amazon Thinks About Free Cash Flow

Why Bezos Cared About Cash, Not Earnings

When you open a letter to Amazon shareholders from almost any year the emphasis falls on free cash flow not net income. That's no accident of style. free cash flow usually abbreviated as FCF is the cash a company generates by running its operations after paying for the long-lived things it needs to keep running - warehouses servers delivery vans called capital expenditure or capital spending. Jeff Bezos made this case going back to the company's early years as a public company and Andy Jassy has continued to use the same framework since he took over. The argument is simple once you break it down. The value of a company is the cash it can eventually deliver to its owners discounted to today not the accounting profits it reports along the way. Net income is an opinion. Cash is a fact

Bezos went beyond simply choosing cash flow over profits. He advocated free cash flow per share not the aggregate number. A company can increase total free cash flow simply by issuing more shares and putting the earnings to work and that's no use to someone who already owns a share. Splitting by the number of shares eliminates the growth that comes from diluting existing owners rather than the business actually improving.itself a form of share count growth. We'll talk more about that later because it's one of the places where this whole framework gets awkward

The Actual Formula Amazon Uses

Here's the formula in layman's terms. Amazon defines free cash flow as cash provided by operating activities less purchases of property and equipment reported for the trailing twelve months in its quarterly filings. That's a simpler version than some companies use. Amazon also separately discloses a more detailed free cash flow figure that also excludes principal repayments from finance leases as the two measures can diverge significantly over years of heavy equipment financing. Note that secondnumber more detailed. Becomes important later

Net income and operating cash flow start from the same place the income statement but answer different questions. Net income asks how much profit is recognized in this period once all income and expenses including noncash charges like depreciation have been equalized. Operating cash flow asks how much actual cash moved through the company's bank accounts in this period which means starting with net income and adding back expenses that reduced profits but never left the bank andThen adjust the time gaps between when revenue and expenses are recognized and when cash actually changes hands. Subtract capital expenditure the cash spent buying the equipment and buildings the business needs to continue growing and you get free cash flow: the money actually left over after keeping the machine running and investing in its own future capacity

Why Net Income and Cash Flow Diverge: Depreciation and Working Capital

Depreciation It's the main reason net income understates cash flow in a capital-intensive business like Amazon's. When Amazon buys a warehouse robot or server cash immediately goes out the door set aside as a capital expense. However the accounting expense is spread over the useful life of the asset and a small depreciation is charged to earnings each quarter for subsequent years. That depreciation reduces net income without costing a single additional dollar of cash in the period in which it is purchased.record. Add it back to net income and you get closer to the picture of real cash. This is exactly the setup that Bezos was wary of regarding net income. Two companies can spend an identical amount of cash on identical equipment and still make very different profits depending only on how quickly each decides to depreciate them

working capital is the second big driver and is where the story gets specific to how Amazon actually operates. Working capital is the short-term assets the inventory the amounts owed to the company the amounts the company owes its own suppliers which lie between the recognition of a sale and the cash from that sale actually arriving. A sale recorded on this quarter's income statement might not turn into cash for weeks. An expense recorded this quarter may not require cash for months.nothing from that point.Operating cash flow is adjusted directly adding increases in accounts payable (cash that Amazon holds on to for a little longer) and subtracting increases in inventory or accounts receivable (cash tied up in things that haven't yet been sold or paid for)

The Negative Cash Conversion Cycle

If you put those working capital time gaps together you get something that finance people call cash conversion cycle.Measures in days how long cash remains tied up between paying for inventory and collecting cash for its sale. Approximately it is the days a company has inventory plus the days it waits to collect from customers minus the days it takes to pay its own suppliers. Most retailers have a positive cycle. They pay suppliers hold the inventory for a while sell it and then wait a little longer for payment from the customer. Cash goes out long before it comes back in

Amazon has been following the opposite pattern for two decades a negative cash conversion cycle. Customers pay almost instantly a card charge clears in days sometimes for products that Amazon hasn't even paid its own supplier for yet because Amazon has negotiated payment terms with many suppliers that go far beyond that. The gap between those two clocks means that Amazon is effectively financed for a significant portion of its inventory by its own suppliers and not by its own balance sheet. Increasing sales in that type of company spits out cash faster than it burns it because more sales meansmore money from other people into Amazon accounts before the bill is due. That's a real structural source of the cash Amazon has used to fund everything from fulfillment centers to now AI data centers. This is a big part of the reason why a company that reinvests almost everything it earns has rarely needed to raise much outside capital to do so

A Worked Example: From Net Income to Free Cash Flow

Numbers make this concrete faster than more paragraphs will so here's a completely created company not Amazon created to show the mechanics cleanly. Call it Year One and Year Two of the same illustrative business numbers in the billions

Walk the arithmetic. In the first year net income of 2 billion is accumulated by adding 4 billion of depreciation and amortization 1 billion of stock-based compensation and another 1 billion of net cash drawn from working capital mostly suppliers are paid a little slower than customers pay.That adds up to 8 billion in operating cash flow. Subtract 7 billion in capital expenditures and free cash flow reaches 1 billion. A perfectly normal and healthy year

Now the second year. In fact the underlying business grew net income increased 20 percent and operating cash flow also grew 25 percent to 10 billion. Nothing in the core business got worse. But management decided to invest money in new capacity and capital spending increased 71 percent from 7 billion to 12 billion. Subtract that from 10.0 billion operating cash flow and theof free cash flow is negative 2.0 billion. Same company growing earnings growing operating cash flow and a free cash flow number that went from solidly positive to negative simply due to a capital spending decision. That's the trick to correctly interpreting a free cash flow collapse. Look at the operating cash flow before panicking about the bottom line

The 2026 Collapse, in Context

Which brings us to what really happened at Amazon. For the trailing twelve months ended March 31 2026 Amazon reported operating cash flow of $148.5 billion up 30 percent from $113.9 billion for the trailing twelve months ended March 31 2025. Property and equipment purchases during the same window reached $147.3 billion aUp 67 percent from $88.0 billion a year earlier. Subtracting one from the other free cash flow for the trailing twelve months ended March 31 2026 came to about $1.2 billion down from $25.9 billion a decline of about 95 percent

A 95 percent drop sounds like a company falling apart and it got exactly that kind of coverage when Amazon reported first-quarter 2026 results. However run the logic of the worked example with Amazon's actual numbers and the story changes. Operating cash flow the cash the real business wastes before any reinvestment grew 30 percent. This is a healthy acceleration not a deterioration. What changed is capital spending which grew more than twice as fast as cash flow.operating cash powered by data centers custom AI chips and networking capacity for Amazon Web Services plus continued spending on robotics and the Amazon Leo satellite constellation. Andy Jassy has forecast roughly $200 billion of capital spending across the company for all of 2026. Whether that spending stays the same or becomes overbuilt depends on whether the resulting capacity is used at prices that amortize the investment and a quarter of the data can't answer that question

Capex as a Choice, Not Just a Cost

Amazon has run this playbook before. In the mid-2010s AWS's intense buildout of fulfillment centers and data centers depressed free cash flow for years and critics really questioned whether the company would ever convert its scale into real cash profits. AWS then became one of the industry's highest-margin cloud businesses and free cash flow rebounded sharply. Building 2026 looks like the same kind of bet on a new category this time theAI infrastructure rather than a sign that the retail or cloud businesses beneath it are struggling. First-quarter 2026 net sales reached $181.5 billion up 17 percent year-over-year and operating income rose to $23.9 billion from $18.4 billion a year earlier. Both figures point to a healthy and growing operating business that isbelow the increase in capital spending. The real risk here is not that Amazon is not profitable. It is about capital discipline: whether $200 billion of capital spending by 2026 generates an adequate return and whether the company can reduce spending if demand for AI infrastructure does not appear as expected

Case Study: Dell and the Negative Cash Conversion Cycle

The clearest historical example of negative growth in cash-conversion cycle financing is not Amazon at all. It is Dell in the 1990s long before it became a struggling laggard in a smartphone and cloud computing market that then emptied. Michael Dell built the company by selling computers directly to customers built to order largely avoiding retail middlemen. Customers pay at the time of order or in short terms. Meanwhile Dell negotiated payment terms.He had longer payment terms with his component suppliers and kept very little inventory of finished products because he built machines only after receiving an order

Business school case studies on Dell at the time describe the company running its cash conversion cycle in negative territory about a week or more in the company's favor while traditional PC makers that had warehouses of finished inventory ran cycles closer to a month.to consume it. The mechanism is identical to the one that keeps Amazon's balance sheet fed today. Get paid before you have to pay and the scale will become partly self-financing

It's also instructive where the comparison breaks down. Dell's model worked brilliantly until the industry shifted to store-bought products and later to phones and tablets where the build-to-order advantage mattered less and design and ecosystems mattered more. A negative cash conversion cycle is a real financial advantage. It's not a permanent moat on its own. It has to sit on top of a product the market still wants

Where the Free Cash Flow Story Breaks

I just dedicated several sections to making free cash flow look like an honest number that's supposedly not net income. It's a better number in many ways. It's not magic and Amazon's own filings give you three distinct reasons to remain skeptical

The first is the most basic. Free cash flow can be boosted by underinvestment. A management team that wants a big free cash flow number in a given year can simply spend less on capital expenditures: delaying maintenance deferring new capacity skipping projects with really good long-term returns. The metric goes up.strengthening the business but the logic is two-way. A company with a large free cash flow trend and a shrinking investment budget deserves more scrutiny not less because you can't tell from the free cash flow line alone whether spending fell because opportunities ran out or because management wanted a more attractive number this year

The second is specific to Amazon's own definition. Remember the second more detailed free cash flow figure Amazon discloses which also subtracts principal repayments from the financial lease? That distinction exists because a company can obtain similar productive capacity a data center a fleet of delivery vans either by purchasing the asset outright which appears as a capital expense or by financing it through a lease where the principal repayment appears in the financing section of the cash flow statement rather than in the sectionsof operation or investment. Simple operating cash flow minus capex free cash flow may miss that entirely. If Amazon leaned more on financed leases for AI infrastructure rather than buying equipment outright free cash flow might look better although the economic obligation paying for the equipment over time is really no different from debt. That's exactly why the itemized measure exists and exactly why I would never quote the simple figure without checking to see if the lease financing changed a lot.between the periods being compared

The third is stock-based compensation and I think it's the most underweight. The SBC is added back when you move from net income to operating cash flow because it's a non-cash expense from the company's perspective no one transfers real dollars when an employee's shares vest. But it's not free. It dilutes existing shareholders increasing the number of shares that Bezos wanted to split the free cash flow by in the first place. A company can show a cash flowThey are not automatically honest about the true cost of paying people with shares rather than cash and it is worth checking this separately each time rather than assuming that a clean free cash flow trend means shareholders are being treated equally well

How I Actually Use Free Cash Flow

My read for what it's worth is that free cash flow is the right first number to check on a company like Amazon but a really bad last number. The way I would actually use it if I were building a view on a stock like this instead of just writing about it is to treat a free cash flow variance as a question never an answer in and of itself

The first thing I check is which side of the equation moved. Did operating cash flow fall or capital spending increase? Those are completely different stories carrying the same headline number and mixing them up is the easiest way to misinterpret a quarter like this. Only after separating the two do I analyze whether the increase in capital spending is described as demand-driven and adjustable language that Amazon and similar companies use when they think they can throttle spending if returns disappoint versus aof capital that is contractually fixed for years regardless of what demand does

Then I try to compare the story to the parts of the business that have nothing to do with capex. Revenue growth operating income segment-level growth like what AWS is doing. If all of this remains healthy while free cash flow plummets this is a management team deciding to reinvest which I find really reassuring rather than alarming. If revenue growth were also slowing at the same time as free cash flow was falling I would interpret the same capex figure asa much more terrifying sign because then you cannot know if the spending is even allocated to what the market wants

I'll say clearly that I find this really difficult to model with true precision. No one outside the company knows the actual expected return on $200 billion of AI infrastructure spending and I'm suspicious of anyone who confidently claims to know. What I can do is track inputs operating cash flow growth capex growth and segment revenue growth quarter after quarter and update as the picture develops. This is a process not aa forecast and I think it is the honest way to use this metric instead of expecting a single number from the last twelve months to tell us something definitive

What to Watch Next

Three things matter more than the headline free cash flow number from here. First the pace of capital spending relative to guidance: whether quarterly spending stays near the reported $44.2 billion in Q1 2026 or accelerates further toward the roughly $200 billion full-year target. Second AWS's revenue and margin growth as AWS grew 28 percentin the first quarter of 2026 and is the segment whose returns will ultimately justify or not justify AI infrastructure spending. Third management comments on capex flexibility as Amazon has historically described a significant portion of its infrastructure spending as demand-driven and adjustable meaning the company can slow the pace if returns disappoint. Free cash flow should begin to recover once construction matures and depreciation catches up to utilization butThe timeline for that recovery is the real open question not the current decline

The Bottom Line

Amazon's free cash flow fell to around $1.2 billion in the trailing twelve months starting in the first quarter of 2026 not because the underlying business weakened but because the company decided to invest record capital expenditures in AI infrastructure. Bezos developed a habit of looking at free cash flow per share instead of earnings because cash is harder to disguise than accounting earnings and depreciation working capital and now financial leases and compensation basedin stocks are exactly the places where the gap between the two numbers hides. Reading a free cash flow figure correctly means separating operating performance which improved from a deliberate investment decision which is the real story and then checking the parts of Amazon's own definition the leases and the share payment that can still make even the honest number less honest than it seems