Amazon $300+ yes, probably. But not just yet.
In my previous article I’ve explained why I think Amazon is expensive right now. I argued that at around $233, Amazon was being priced for a future that still required a meaningful amount of faith. As of April 12, 2026, I think that basic concern still holds.
After a sharp rally that pushed the stock into roughly the $240 area on April 10, the central debate matters even more: not whether Amazon is a great company, but whether the current price already gives it credit for too much of what still has to be proven.
Let me be clear about what I am arguing. I am not saying Amazon is broken. I am not saying management is hiding anything. And I am not denying that AWS is showing real AI traction.
What I am saying is simpler: at these levels, the stock appears to require a very demanding outcome; one in which AI monetization ramps quickly, massive infrastructure spending earns strong returns soon enough, and the rest of Amazon’s business remains stable while all of this is happening.
That is a lot to assume at once.
Why the Economic Life of AI Infrastructure May Matter More Than the Accounting Life
One underappreciated factor in this debate is how Amazon’s depreciation assumptions have evolved over time — and why the economics of AI infrastructure may make those assumptions more fragile than they appear.
The issue is the growing gap between the accounting life of infrastructure and its economic life in a much faster-moving hardware environment.
Amazon previously extended the estimated useful life of servers and networking equipment from four years to five years in 2022, and then from five years to six years effective January 1, 2024. According to Amazon’s 2024 annual report, the 2024 change reduced depreciation and amortization expense by $3.2 billion for the year and increased net income by $2.5 billion.
But in 2025, the direction partially reversed. Amazon said it shortened the useful life of a subset of servers and networking equipment from six years back to five years, citing the faster pace of technological development, particularly in artificial intelligence and machine learning. It also recorded approximately $920 million of accelerated depreciation and related charges in Q4 2024 tied to early retirements of certain servers and networking equipment. Together, those two factors were expected to reduce 2025 operating income by about $1.3 billion.
That matters for a simple reason: it suggests that the six‑year assumption may be less durable in AI‑related infrastructure than investors currently assume.
In traditional cloud, longer useful lives were easier to defend because performance cycles were slower and refresh needs were more predictable. In frontier AI, that may no longer be true. Nvidia’s public roadmap from Blackwell to Vera Rubin and beyond points to a much faster cadence of improvement in compute, memory, bandwidth, and system efficiency. Even if older infrastructure does not become obsolete overnight, it may become economically less competitive much sooner than legacy cloud hardware did.
That is why the more important question is whether a $200 billion capex cycle has enough time to generate compelling economic returns before the next wave of infrastructure demands makes another major reinvestment cycle economically necessary.
If the answer turns out to be less favorable than the market currently assumes, then the path to free‑cash‑flow recovery becomes much narrower.
A Simple Sensitivity: Why Shorter Economic Lives Would Matter
I want to be precise here. Amazon has not said that its full AI stack will revert to three‑year lives, and it has not disclosed a single “effective life” for all AI‑related infrastructure. So what follows is not a reported fact. It is a scenario analysis based on standard depreciation mechanics.
In a stylized accounting sensitivity, if one were to spread $200 billion of annual capex over six years, that would imply roughly $33 billion of annual depreciation. Spread over three years, the same notional capex would imply about $67 billion. That is a swing of roughly $34 billion in annual depreciation expense. This is not a forecast of Amazon’s reported results; it is a simple illustration of how much asset-life assumptions can matter when capex reaches this scale.
This is a stylized illustration. The real answer will depend on what share of future capex is tied to front‑tier AI hardware subject to shorter renewal cycles. It does not mean Amazon will report exactly those numbers. It does mean that when capex gets this large, asset‑life assumptions materially affect how quickly future cash‑flow normalization can show up in reported economics. So the real analytical issue is not whether a three‑year reversion is certain. It is whether investors are assigning enough probability to a world in which AI‑related infrastructure proves economically shorter‑lived than the market currently assumes.
Free Cash Flow Is Still the Core Tension
The cleanest place to start is with free cash flow itself.
Amazon’s trailing‑twelve‑month operating cash flow rose to $139.5 billion in 2025, up from $115.9 billion in 2024. On the surface, that looks reassuring. But free cash flow fell sharply to $11.2 billion from $38.2 billion, as net purchases of property and equipment jumped from $77.7 billion in 2024 to $128.3 billion in 2025. Management has since indicated that capital expenditures in 2026 will be around $200 billion, driven largely by investment in AI infrastructure.
That is the core tension in the stock. Amazon is still producing substantial operating cash, but far more of that cash is now being absorbed by infrastructure spending. So while the earnings story may look manageable on the surface, the cash‑generation story is much less comfortable.
That is also why I do not think the valuation debate can stop at the P/E ratio. At a stock price in the $233–$238 range and using 2025 free cash flow of $11.2 billion, Amazon’s market capitalization implies a price‑to‑free‑cash‑flow multiple of roughly 225x to 230x, with an implied free‑cash‑flow yield of around 0.43% to 0.45%. Even if one argues that the earnings multiple is less extreme than in earlier periods of peak enthusiasm, the cash‑flow multiple still looks very demanding.
There is no serious debate that Amazon can still grow earnings if AI monetization works. But when the stock is this dependent on future cash‑flow normalization, the assumptions embedded in that normalization matter far more than they would in an ordinary capex cycle.
Longer Lives Made Sense in Cloud. AI May Change That
Amazon is not alone in having extended hardware lives over time. Across the cloud industry, longer useful‑life assumptions became more common as infrastructure matured, designs became more standardized, and the economics of scale improved. That trend made sense in a world where hardware refresh cycles were becoming more stable.
But AI changes the character of the investment cycle. Microsoft, Google, and Amazon are all investing heavily into infrastructure that now sits much closer to a technology frontier defined by rapid jumps in accelerators, memory, networking, and power efficiency. The relevant question is no longer just, “How long can this server physically run?” It is increasingly, “How long does this hardware remain economically competitive for high‑value AI workloads?”
That is an important distinction. A machine can remain usable long after it stops being the best economic choice for the workloads that matter most.
The Bull Case Is Real — but Still a Forward Case
To be fair, the bullish argument is not weak.
In his latest shareholder letter, Andy Jassy disclosed that AWS’s AI business is now running at more than $15 billion annualized in the first quarter of 2026. Reuters also reported that Amazon’s chips business — including Graviton, Trainium, and Nitro — is running at more than $20 billion annualized. Jassy further defended Amazon’s plan to spend roughly $200 billion in 2026, arguing that there are meaningful customer commitments behind that investment and that much of the spending should be monetized in 2027 and 2028.
Those are real disclosures and meaningful third‑party data points. They matter. They clearly strengthen the case that Amazon is not simply building empty capacity and hoping demand appears later.
But they do not fully settle the valuation question. They strengthen the future case. They do not eliminate the present tension. The stock is still asking investors to believe that these demand signals will convert into visible economic returns quickly enough to justify today’s price — and that the infrastructure supporting those returns will not require another major reinvestment cycle before the current one has paid off in economic terms. That is not impossible. It is just not yet proven in the cash‑flow profile.
Why This AI Capex Cycle Is Not the Same as the Old AWS Build‑Out
One of the most common counterarguments is that Amazon has been through major infrastructure investment waves before, especially during the original AWS expansion. That is true. But the current AI cycle is not identical to the earlier cloud build‑out.
Traditional cloud infrastructure was capital‑intensive, but it did not operate under the same kind of rapid hardware performance race that now defines top‑end AI systems. Today’s AI infrastructure is tied to a much faster cadence of improvements in compute, memory, interconnect, and system‑level efficiency. That does not mean every dollar of AI capex is at risk of immediate obsolescence. It does mean the monetization window may be less forgiving than investors became accustomed to during earlier cloud cycles.
That is where the concern becomes more subtle and, in my view, more important. The real issue is that the accounting life of infrastructure may not map neatly onto its economic life in a market where competitiveness can shift much faster than it used to.
In other words, the risk is not simply the $200 billion being spent now. The risk is that maintaining competitive relevance could require another very large round of spending before the current one has fully paid for itself in economic terms. In old cloud, that question was less urgent. In AI, it may be central.
The Market May Be Treating Future Returns as More Certain Than They Really Are
This is the heart of my concern.
If one believes AWS AI, custom silicon, and data‑center scale will produce a much larger earnings base in the years ahead, then Amazon can certainly justify a strong valuation. But the stock is not just pricing in success in some abstract sense. It seems to be pricing in a relatively smooth sequence: demand stays strong, monetization ramps fast, margins improve, the capex burden eventually eases, and the rest of Amazon’s business absorbs the transition without meaningful damage.
That sequence may happen. But investors should admit how much has to go right for it to happen cleanly.
At this point, much of the support for the bull case still comes from run rates, customer commitments, strategic logic, and management guidance about future monetization. Those are important inputs, but they are not the same thing as already seeing a strong rebound in free cash flow after the investment. The asset‑life question discussed above makes that future recovery even more sensitive to timing and execution. Until that cash conversion becomes more visible and sustained, I think it is fair to say the market is extending Amazon a great deal of credit in advance.
Where the Bear Case Could Be Wrong
The bear case is not airtight, and it is worth being explicit about that.
The first challenge is timing. Amazon has often invested heavily ahead of the revenue curve. If the $15 billion AWS AI run‑rate accelerates sharply through the second half of 2026, the market may stop focusing on near‑term depreciation pressure and instead reward the company for capturing a larger share of an expanding AI market. In that scenario, today’s $200 billion capex plan may look less like overextension and more like a well‑timed land grab than today’s bear thesis suggests.
The second challenge is custom silicon. Graviton, Trainium, and Nitro are not just cost‑savings stories; they may also become a margin‑recovery engine. If Amazon can increasingly route inference workloads through its own silicon, as management suggests is strategically important, AWS economics could improve meaningfully.
That would not eliminate the risk, but it would change the balance of it.
The question would shift from whether Amazon is spending too much to whether its spending is giving it a more durable cost advantage than skeptics assume.
The third challenge is the resilience of the consumer moat. Amazon’s ACSI score of 82, down only one point year‑over‑year, suggests that broad‑based customer deterioration is not yet severe. The company’s logistics density, Prime ecosystem, and fulfillment scale still give it a cushion that many competitors cannot match. That does not solve the AI valuation problem, but it does give Amazon more room to absorb mistakes than a weaker company would have.
These are real counterarguments, and they are precisely why the bull case deserves respect. But acknowledging upside scenarios is not the same thing as assuming they are already fully justified in the price.
Retail Is No Longer the Old Growth Engine — and That Changes the Cushion
Another part of the valuation problem is that the mature side of Amazon’s business no longer carries the same growth premium it once did.
Retail still matters enormously. It anchors Prime, fulfillment density, the consumer relationship, and the broader ecosystem. But the premium case for the stock today is much more concentrated in AWS, AI services, custom chips, and the operating leverage Amazon hopes to create from them over time. That matters because it means the stock’s upside increasingly depends on future returns from a capital‑intensive technology build‑out, not on a broad‑based reacceleration of the company’s mature businesses.
That does not make the retail business irrelevant. It makes the margin for error smaller. If AI monetization takes longer than expected, investors are relying more heavily on the resilience of a business mix that is no longer growing the way Amazon’s overall story used to grow.
Customer Service: Not a Proven Breakdown, but a Risk the Market May Be Underestimating
This part of the debate requires precision.
The hard data do not support a dramatic claim that Amazon’s customer experience has already collapsed. The 2026 ACSI Retail and Consumer Shipping Study showed Amazon tied for the top score among online retailers at 82, down one point year over year but still at the top of the category. That is not evidence of a broken moat.
At the same time, I do not think the qualitative warning signs should be dismissed. Amazon has eliminated about 30,000 corporate jobs since October 2025, including another 16,000 in January 2026, and Reuters reported that these cuts were part of a broader effort to reduce bureaucracy and push efficiency as AI changes the company’s labor needs. Recent consumer‑facing coverage also suggests that reaching a real human in Amazon support has become increasingly difficult. On April 6, Tom’s Guide published a piece explaining the effectively hidden path required to get a callback from a live Amazon customer‑service agent after navigating automated menus and self‑service flows.
That is not the same thing as proving a broad service breakdown. But it is enough to frame a real execution risk. Amazon built trust not just on price and selection, but also on convenience and resolution. If automation and cost discipline gradually add friction to customer support while the company is simultaneously asking investors to remain patient on AI returns, then part of Amazon’s historic qualitative advantage may come under more pressure than topline satisfaction surveys currently reveal.
Final View: A Great Company Can Still Be an Expensive Stock
My conclusion remains straightforward.
Amazon is still a world‑class company. AWS is clearly gaining traction in AI. The custom‑chip strategy may prove more powerful than skeptics expect. And management has given investors more substance than just vague promises.
But as of April 12, 2026, with the stock having recently traded in the $233–$240 range, I still think the market is underwriting too much perfection.
Investors are effectively being asked to assume rapid AI monetization, strong returns on a roughly $200 billion capex plan, durable competitiveness in a fast‑moving hardware cycle, and a stable core business while all of that unfolds.
Maybe Amazon delivers all of that.
But at these levels, I do not think the stock offers enough margin of safety for how much still has to go right.
We’ll see what next quarter earnings bring. Maybe they will confirm my thesis. Maybe they will make me change my mind. If I am wrong and the train leaves without me, so be it. Congratulations to those who were willing to pay up and were proven right.
🟢 Disclosure: The author does not hold a position.
⚠️ I produce these analyses for my own enjoyment and because I’m always looking for new opportunities. I am not a financial professional, and I don’t have access to professional-grade tools or proprietary data. Everything here is built from publicly available information and my own reasoning — which means I can be wrong. This analysis may include forward-looking statements based on current expectations and projections; these are subject to risks and uncertainties that could cause actual results to differ materially from what is discussed here. I may not always see the full picture, and my views will change as new information emerges or as I come to understand data points I initially overlooked or underweighted. However, I am under no obligation to update or keep this information current as the situation evolves. I only operate with cash positions — no leverage, no margin, no shorting. I never bet against the market or individual companies. My opinion may be based on fundamentals, market behaviour, or a mix of both. The company is not its price, and the price is not the company. I express my own opinions. I am not receiving compensation to share this. I have no business relationship with any company whose stock is mentioned in this article. Nothing here is financial advice. Do your own due diligence.