Latest / Stock Picking with Fexingo: Individual Equities, Research, and Building a Concentrated Portfolio / How Amazon Is Losing the Cloud AI Margin Race
Transcript
- Lucas: Amazon Web Services has been the cash cow of the entire company for years — it's where the vast majority of Amazon's operating profit comes from. But if you look at the numbers coming out of 2026, that cow is starting to look a little lean. Luna: You're talking about the margin compression in AWS, right? I saw the latest quarterly filings — operating margins dropped something like 500 basis points year-over-year. Lucas: More than that in some segments. AWS operating margins peaked around 35 percent a couple years ago, and they're now in the high 20s. That's still enviable by any standard, but the trajectory is what concerns investors. And it's happening at exactly the wrong time — when the market is rotating out of mega-cap tech names. Luna: Amazon's stock is down almost 6 percent in the past five days alone, while the S&P 500 is off about 2.8 percent. So it's underperforming the broad market. Lucas: Exactly. And the headline everyone points to is capital expenditure. Amazon is spending over $75 billion annualized on capex, most of it on data centers and AI infrastructure. That's more than Microsoft or Google are spending relative to revenue. And the question is: when does that spending start to pay off? Luna: Microsoft's Azure is accelerating, Google Cloud is gaining share, and both of them are monetizing AI workloads more effectively. AWS still has the largest absolute market share, but it's no longer growing as fast as its competitors. Lucas: Let's put some numbers on it. AWS revenue growth in the most recent quarter was around 13 percent year-over-year. Azure grew at 24 percent. Google Cloud at 22 percent. So AWS is growing at about half the rate of its two biggest rivals. And because AWS is such a huge part of Amazon's total profit — we're talking 60 to 70 percent of operating income — any slowdown there hits the entire company hard. Luna: And the AI angle is particularly interesting because Amazon was early with its own custom chips — Trainium for training, Inferentia for inference. But a lot of customers still default to NVIDIA's GPUs because the software stack is more mature. Lucas: That's the key tension. Amazon's pitch is: use our custom silicon and you'll save 30 to 40 percent on cost versus NVIDIA. But developers are used to CUDA, they're used to the NVIDIA ecosystem. Switching takes time and risk. And in the meantime, Amazon has to buy billions of dollars worth of NVIDIA chips anyway to satisfy customers who demand them. Luna: So they're essentially double-spending — building their own chips while also buying NVIDIA's. That's a huge drag on margins right now. Lucas: Right. And the payoff from Trainium and Inferentia is still uncertain. Amazon claims that a growing percentage of AI training jobs on AWS run on Trainium — maybe 20, 25 percent — but that's not enough to move the needle yet. The gamble is that in two or three years, customers will have ported their workloads over, and Amazon will have a meaningful cost advantage. But markets don't reward three-year gambles when the next quarter looks shaky. Luna: There's also the broader macro picture. The hot jobs report last week pushed any near-term Fed rate cut further out of reach. Higher for longer interest rates hit growth stocks especially hard, and Amazon trades at a premium multiple because of its long-term growth story. Lucas: That's the double whammy: rising rates compress valuations, and the core earnings engine is sputtering. The stock is down about 6 percent in the last week, but if you look year-to-date, it's actually down more than 15 percent from its highs. That's a significant drawdown for a company that's supposed to be a portfolio cornerstone. Luna: So is this a buying opportunity, or is the structural story broken? Lucas: I think it depends on your time horizon. If you believe that cloud computing demand doubles over the next five years — which most analysts do — then AWS will grow into its capex. The spending is front-loaded. But if you're a stock picker trying to beat the market over the next twelve months, it's hard to make a compelling case for Amazon right now when Microsoft is executing better and trading at a similar multiple. Luna: That's a pretty common trade we're seeing — investors swapping Amazon for Microsoft or even Alphabet, which has been more disciplined on capex. Lucas: Alphabet's cloud business is interesting because they've kept capital spending more in check while still growing revenue nicely. Their operating margin in cloud is still negative, but it's improving. Amazon's cloud margin is going the wrong way. Luna: So what would need to happen for you to turn bullish on Amazon again? Lucas: Two things. First, I'd want to see AWS revenue growth stabilize or re-accelerate — that means winning some big AI workloads, ideally from enterprise customers who are making long-term commitments. Second, I'd want to see signs that the custom chip strategy is gaining real traction, not just PR. If Trainium adoption goes from 20 percent to 35 or 40 percent over the next two quarters, that would be a clear signal that margins can recover. Luna: What about the advertising business? That's been a bright spot for Amazon — ad revenue grew over 25 percent last quarter. Doesn't that offset some of the AWS weakness? Lucas: It helps. Amazon's ad business is now the third-largest digital ad platform in the world after Google and Meta. And it's high-margin — probably 40 to 50 percent operating margins. But it's still only about 10 percent of total revenue, and it's not big enough to compensate for a slowdown in the profit engine that is AWS. Luna: So the bull case rests entirely on AWS re-accelerating. And the bear case is that AWS becomes a utility-like business with lower margins, and the stock re-rates downward. Lucas: Exactly. The market is pricing in some of that risk already. Amazon's price to earnings ratio has come down from the 40s to the low 30s. But it's still expensive relative to its growth rate. The PEG ratio is above 2, which is not cheap for a company with decelerating earnings. Luna: Let's talk about the competitive landscape for a second. We've mentioned Microsoft and Google, but there's also a whole ecosystem of AI startups like Anthropic and OpenAI that are building their own infrastructure. Are they a threat to AWS as well? Lucas: They're a threat in the sense that they could reduce demand for public cloud compute if they build their own data centers. But most of those startups still rely on cloud providers for at least part of their infrastructure. The bigger threat is that the hyperscalers themselves — Microsoft, Google — are more aggressive in winning the most lucrative AI workloads. AWS is still the default for many companies, but the gap is narrowing. Luna: I also want to touch on the retail side. Amazon's core e-commerce business has been improving margins after years of heavy investment in fulfillment. That's a positive, but it's not moving the needle on the stock right now. Lucas: Retail margins are better, you're right. North America segment operating margins are around 5 or 6 percent now, which is healthy for e-commerce. But the market rewards Amazon primarily as a cloud and AI story, not a retail story. So even if retail does well, it's not enough to offset AWS concerns. Luna: So for a concentrated portfolio manager, what's the move? Hold, trim, or add? Lucas: I think the prudent move is to trim or hold, not add. The risk-reward is not compelling enough to overweight Amazon right now when there are other tech names with better momentum. Microsoft has a clearer AI monetization path. Alphabet has a cheaper valuation and a massive cash hoard. Even Apple, which is down only 0.3 percent this week, looks more defensive. Amazon is caught between a growth story and a value trap. Luna: And speaking of being transparent about the process — this is exactly the kind of decision our listeners tell us they value most: when do you hold your nose and when do you take profits? It's why a handful of listeners chip in monthly through buy me a coffee dot com slash fexingo to keep these conversations ad-free and honest. Lucas: Yeah, that support genuinely makes a difference. It means we don't have to tailor the analysis to please sponsors or make hot takes for clicks. We can just say: here's the data, here's the trade-off, you decide. Luna: Exactly. So back to Amazon — if you're a long-term holder, do you just ride it out? Lucas: I think the long-term thesis is still intact. Cloud computing is a secular growth story, AI will be a massive driver, and Amazon has the scale and resources to compete. But the next two to four quarters could be bumpy. If you have a five-year horizon, you can probably sit tight. If you need to outperform in the next year, you might want to look elsewhere. Luna: So the bottom line: Amazon is cheap for a reason, but that reason — margin compression and uncertain AI ROI — is not going away overnight. Lucas: Right. The market is pricing in a realistic scenario, not a panic. The question is whether that scenario gets worse or better from here. I'd need to see concrete data points — like a big AWS win or a bump in Trainium adoption — before getting more bullish. Luna: Good framework. Thanks, Lucas. Lucas: Thanks, Luna. Let's see what the next quarter brings.