To understand the “timing lag” framework through Amazon CEO’s remarks during the earnings call: Today, the most noteworthy point from Amazon CEO Andy Jassy’s earnings call was not his statement that AWS will eventually reach $1 trillion in revenue—this is important, but not the most critical. The key takeaway is that he finally provided a clear timeline for the return on capital expenditures, something entirely absent from Google’s earnings call. Jassy’s remarks perfectly illustrate the “timing lag” framework we previously discussed. Andy Jassy views capital expenditures primarily in two components: data centers, and the servers and networking equipment placed within them. These have distinct capital cycles: 1) The construction cycle for data centers typically takes one to two years. Once a data center is operational and servers are installed, it immediately begins generating substantial revenue. With a lifespan of approximately 30 years, these facilities can continue generating revenue for over three decades without requiring a repeat of the initial capital investment. 2) The lifecycle for servers and networking equipment is much shorter. These devices are typically purchased several months before a data center becomes operational, meaning Amazon must have strong visibility into customer demand before committing to this spending. If demand is not there, Amazon does not make the capital expenditure. For servers and networking equipment, the average investment achieves payback in less than three years. Server lifespans are currently at least five to six years, and most of Amazon’s current AI capacity contracts are for at least five years. This means that for two to three years after payback, Amazon will generate significant free cash flow from servers and networking equipment. AWS has a strong track record of shortening the payback period for server equipment—we’ve made meaningful progress without compromising customer experience and have found ways to extend the useful life of these devices. 3) For data centers with a lifespan exceeding 30 years, AWS should be able to realize at least five to six generations of server economics, as previously explained. Subsequent generations after the first benefit from avoiding repeated data center investments, resulting in even better overall economics. When demand forces so many data centers to be built simultaneously before they can be monetized, capital expenditures rise sharply in the short term, creating headwinds for free cash flow until those data centers come online and their servers are utilized over several years. As time passes, revenue growth will eventually outpace incremental capital expenditure growth (which will occur at some point), leading to highly attractive levels of revenue, free cash flow, and return on invested capital. AWS experienced this dynamic during the first era of cloud computing—but over a longer timeframe, because demand accumulated more gradually than it does in the AI era. Now, we’re seeing profit margins and returns in AI align with—and even slightly outpace—those observed during the same developmental stage of core business segments. From the second paragraph above through here is essentially Andy Jassy’s direct wording. According to Jassy’s statement, servers and networking equipment achieve payback in less than three years on average, with server lifespans of at least five to six years and most AI capacity contracts lasting at least five years. If this narrative translates into actual cash flows, the typical lifecycle of a server batch would look like this: Year 0: Purchase of servers and networking equipment Years 1–3: Cumulative recovery of initial investment Years 4–5/6: Transition into clearly positive free cash flow Subsequent upgrades: Replace servers while continuing to use existing data centers, power, and land. This model theoretically generates ROIC above the cost of capital. This is precisely the core idea of the “timing lag” framework we discussed earlier: “Cash is paid first (capital expenditures are incurred upfront), capacity comes online later, orders convert to revenue afterward, and profits and free cash flow are realized last.” It’s about spending first, then observing how quickly and efficiently returns materialize.