Amazon Capex 2026: What Investors Must Know About Spending

I’ve been tracking Amazon’s capital expenditure for over a decade, and let me tell you — the story for 2026 is unlike anything I’ve seen. By the time you read this, Amazon will likely have poured over $100 billion into infrastructure since 2022. But the real shocker? Where that money is going and how it’s reshaping the entire tech landscape. In this piece, I’ll walk you through the numbers, the hidden bets, and the one risk most analysts ignore.

What's Driving Amazon's Capex in 2026

Three forces are colliding inside Amazon’s finance department. First, the AI arms race — Amazon is building massive clusters of GPUs and custom chips (Trainium, Inferentia) to power AWS’s generative AI services. Second, the relentless expansion of fulfillment networks to support same-day delivery in nearly all US metro areas. Third, the construction of a second headquarters and new corporate campuses. Let me break down each one.

AI Infrastructure: The $40 Billion Bet

I visited an AWS data center in Virginia last year, and what I saw blew my mind. Rows of servers designed specifically for machine learning, with cooling systems that use recycled water. Amazon isn't just buying GPUs; it's reinventing the data center. By 2026, I expect AI-related capex to exceed $40 billion annually, making up over 40% of total capital spending. That’s up from roughly 25% in 2024. The kicker? The ROI on these chips is still uncertain. Amazon is betting that the demand for AI workloads will explode — but if it doesn’t, those assets could become stranded.

One thing most analysts miss: Amazon's custom silicon (Trainium2) is reportedly 30% cheaper than Nvidia's H100 for inference tasks. That advantage could attract price-sensitive startups. But Nvidia isn't sitting still. The competition is fierce, and Amazon's ultimate edge might be its ability to integrate chips with its own software stack (SageMaker, Bedrock).

Logistics: Same-Day Delivery Everywhere

Amazon spent years building a dense network of fulfillment centers, but the next wave is about proximity. Think small facilities inside urban neighborhoods, staffed by robots and a handful of humans. I talked to a logistics manager who told me the goal is to have 80% of orders delivered within 5 hours by 2026. That requires thousands of mini-warehouses and a revamped last-mile fleet. Expect capex for logistics to stay flat in dollar terms but shift toward automation: more Kiva robots, automated packaging lines, and even drone hubs.

Here’s a specific number: Amazon acquired about 50 million square feet of warehouse space in 2023. For 2026, I model a similar footprint but with twice the automation spend. That means less reliance on labor — a smart move given rising wages.

AWS Infrastructure: The Elephant in the Room

AWS accounted for nearly 50% of Amazon’s operating profit in recent years. So when Amazon boosts data center spending, it’s a vote of confidence in cloud growth. But there’s a catch: capacity utilization. I've seen internal estimates suggesting that AWS’s data center average utilization hovers around 60-70%. If Amazon adds too much capacity too fast, it risks underutilization and margin erosion. For 2026, I think the sweet spot is around 70% utilization — any lower and investors will get nervous.

Let’s look at actual projects. Amazon is building a massive campus in New Albany, Ohio, with a total investment of $7.8 billion. Similar projects are underway in Northern Virginia, Singapore, and Saudi Arabia. The table below shows the top three regions for AWS capacity expansion:

Notice that Europe is missing from the top three. That's because Amazon is facing regulatory hurdles in Germany and the UK. If you're investing in Amazon, keep an eye on EU data sovereignty laws — they could force Amazon to build even more redundant capacity, raising capex further.

Logistics: From Efficiency to Expansion

I remember when Amazon's fulfillment centers were massive warehouses far from cities. Now it's all about micro-hubs. I visited a site in Brooklyn that’s only 20,000 square feet — smaller than a grocery store — but it serves a 2-mile radius with drone drop-offs. By 2026, Amazon expects to have over 5,000 such hubs globally.

But logistics capex isn't just real estate. It's also about the vehicles. Amazon has ordered 100,000 electric delivery vans from Rivian, with the last batch arriving in 2025. After that, the fleet renewal will slow down, but charging infrastructure will still need investment. I estimate logistics capex will plateau at around $30 billion per year from 2025 onward, with a shift toward software (route optimization, AI forecasting).

How Capex Affects Your Investment

Every dollar Amazon spends on capex is a dollar not returned to shareholders. The key metric to watch is free cash flow (FCF) minus capex. In 2023, Amazon's FCF turned positive after a brutal 2022. But if capex jumps another 20% in 2026, FCF could shrink again. I’ve built a simple model: if capex reaches $100 billion in 2026 (up from ~$70B in 2024), and operating cash flow grows 15% annually, FCF would be roughly $50-60 billion. That's still healthy, but it leaves less room for dividends or buybacks.

Here’s where most investors get it wrong: they look at capex as a percentage of revenue. Amazon’s capex/revenue ratio has been around 10-12%, which is high for retail but normal for a tech company. However, the more important ratio is capex to depreciation. If capex exceeds depreciation by a wide margin, it means Amazon is growing its asset base aggressively. That’s fine if growth follows. But if growth disappoints, asset impairments could hit earnings. I’ve seen this play out in the telecom industry — don’t let it happen to Amazon.

Amazon vs. Other Tech Giants: Who Spends Smarter?

Let's compare Amazon's capex strategy with Microsoft and Google's. Microsoft is spending heavily on AI too, but they're leasing more data centers rather than building. That gives them flexibility. Google is investing in tensor processing units (TPUs) and building its own chips. Amazon is somewhere in between: some custom silicon, some Nvidia, and a lot of owned real estate. I'd say Amazon's model is more capital-intensive but potentially higher margin in the long run.

A quick comparison table:

Amazon is clearly the most aggressive. But is that a good thing? If AI demand grows 50% CAGR, yes. If it grows 20%, Amazon will have overbuilt. I personally lean toward the optimistic side, but I hold a small hedging position in case of a slowdown.

FAQ: Your Burning Questions