Wall Street changed its mind about AI spending in mid-2025: hyperscaler capex now looks smart when it feeds a real cloud-computing business and reckless when it doesn't. This guide shows you how to evaluate Microsoft, Amazon, and Alphabet on their AI-linked cloud revenue, and how to track that exposure against broader benchmarks in one place.
What does "AI capex only pays off with cloud" actually mean?
It means investors now reward AI spending that flows directly into a growing cloud-computing business, and punish spending that doesn't. AI capex is the money hyperscalers pour into data centers, GPUs, and networking to train and serve artificial intelligence models. A MarketWatch report on Aug 1, 2025 framed the new consensus: capex is only "good" when it converts into paying cloud revenue from customers renting compute.
The logic is simple. A company like Microsoft can spend $20 billion a quarter on infrastructure and see it justified if Azure keeps growing 25% or more year over year. A company spending the same amount with no cloud rental business to show for it looks like it's burning cash with no return path.
Why the market reframed hyperscaler spending
The reframe happened because early AI enthusiasm gave every big spender a pass, and that stopped working. Investors got specific: they now separate revenue-generating AI infrastructure from speculative AI infrastructure. This mirrors a broader repricing we covered in our look at where AI valuations stop making sense in 2026.
- Cloud-linked capex: spending that shows up as Azure, AWS, or Google Cloud revenue within a few quarters.
- Product-linked capex: spending on internal AI features (search, ads, Copilot) with a slower, fuzzier payoff.
- Speculative capex: spending on capacity ahead of demand, betting customers arrive later.
How do Microsoft, Amazon, and Alphabet compare on cloud-linked AI capex?
All three run massive cloud businesses, but they differ in scale, growth rate, and how directly their AI spending ties to rentable cloud revenue. Here is the side-by-side that matters for the "capex pays off with cloud" thesis.
| Company | Cloud arm | Cloud growth (approx) | AI-capex payoff clarity |
|---|---|---|---|
| Microsoft (MSFT) | Azure | ~28% YoY | High: OpenAI partnership feeds Azure directly |
| Amazon (AMZN) | AWS | ~18% YoY | High: largest cloud base, Bedrock and Trainium chips |
| Alphabet (GOOGL) | Google Cloud | ~30% YoY | Medium-high: fastest growth, smaller base, TPU edge |
Microsoft (MSFT): the cleanest cloud-capex story
Microsoft has the tightest link between AI spending and cloud revenue because of its OpenAI relationship and Azure. Every dollar of GPU capacity it deploys can be rented back through Azure OpenAI Service, which turns capex into recurring cloud revenue fast. This is why MSFT is the name most analysts point to when they say AI spending "works."
Amazon (AMZN): scale plus custom silicon
Amazon's advantage is the sheer size of AWS, still the largest cloud platform by revenue. Its AI capex feeds both third-party model hosting (via Bedrock) and its own Trainium and Inferentia chips, which lower the cost of serving AI workloads over time. Slower headline growth reflects a bigger base, not weaker economics.
Alphabet (GOOGL): fastest growth, chip advantage
Alphabet's Google Cloud is the fastest-growing of the three off a smaller base, and its custom TPU chips let it train large models without paying full Nvidia margins. The catch: a larger share of Alphabet's AI spend still supports internal products like Search and ads, so the pure cloud payoff is slightly less clean than Microsoft's.
Why should you track AI capex exposure separately in your portfolio?
You should track it separately because a portfolio can look diversified while being heavily concentrated in a single macro bet: the payoff of AI infrastructure spending. If you own MSFT, AMZN, GOOGL, plus an S&P 500 index fund, you may be tripling down on hyperscalers without realizing it.
- The Magnificent Seven make up roughly a third of the S&P 500 by weight.
- Owning the index plus individual megacaps stacks the same names on top of each other.
- An AI-capex disappointment hits your index fund and your single stocks at once.
PortfolioTrackr handles this by letting you tag holdings into custom groups, so you can see your true AI-linked cloud exposure as a single percentage of net worth instead of guessing. This is the same overlap problem that makes a dedicated tracker beat a manual spreadsheet once you hold more than a handful of positions.
How do you measure a hyperscaler's AI capex against a benchmark?
You measure it by comparing each company's capex-to-revenue ratio and cloud growth rate against a benchmark like the S&P 500 or the Nasdaq-100. The goal is to see whether spending is accelerating faster than the revenue it's supposed to generate.
- Capex intensity: total capex divided by total revenue. Rising intensity needs rising cloud growth to justify it.
- Cloud growth vs capex growth: if capex grows 50% but cloud revenue grows 20%, the payoff is stretching further out.
- Relative return: how each stock performs against the Nasdaq-100 over 6 and 12 months.
In PortfolioTrackr you can pin a benchmark to any holding or group and watch relative performance update in real time, which turns this from a quarterly spreadsheet chore into a live dashboard. If you also hold chip names, our breakdown of DeepSeek and AI chip fragmentation explains why the hardware layer can move independently of the cloud layer.
What are the risks if the cloud-capex thesis breaks?
The main risk is that AI demand plateaus while capex is already committed, leaving hyperscalers with expensive, underused data centers. Because these facilities take years to build, spending decisions made in 2025 lock in costs through 2027 and beyond regardless of demand.
- Demand air pocket: enterprise AI adoption slows and cloud growth decelerates below capex growth.
- Margin compression: depreciation on new data centers drags operating margins before revenue catches up.
- Efficiency shock: cheaper models (the DeepSeek scenario) cut the compute needed per query, softening demand for rented GPUs.
- Rate sensitivity: higher-for-longer rates raise the cost of capital-heavy buildouts, a dynamic we cover in our rate-hike portfolio guide.
None of these break the long-term case on their own, but each can compress multiples on names priced for flawless execution. Watching the gap between capex growth and cloud growth is your earliest warning signal.
How do you set up an AI-cloud watchlist in a tracker?
You set it up by grouping your hyperscaler holdings, attaching a benchmark, and tracking relative performance plus each earnings date in one view. The point is to stop checking three broker apps and start seeing the whole AI-cloud bet at once.
- Create a group called something like "AI Cloud" and add MSFT, AMZN, and GOOGL.
- Pin the Nasdaq-100 or S&P 500 as the benchmark for the group.
- Set price and earnings alerts so you catch cloud-revenue prints the day they land.
- Add any related ETFs or chip names so overlap is visible instead of hidden.
If your positions are spread across brokers like Interactive Brokers, Schwab, and Alpaca, our guide on connecting brokerage accounts to a portfolio tracker walks through pulling them into a single dashboard. That consolidation is what makes exposure math accurate instead of approximate.
The bottom line
The market's new rule is clear: AI capex only pays off when it converts into rentable cloud revenue, which is why Microsoft's Azure, Amazon's AWS, and Alphabet's Google Cloud are the reference names for the thesis. Track them as a group, benchmark them against the Nasdaq-100, and watch the gap between capex growth and cloud growth. Do that and you'll spot a stretched payoff long before it shows up in the share price.
Track your portfolio in real time: free for 3 days
Live P&L across stocks, crypto, and global markets. WhatsApp and Telegram price alerts. AI trade import. Unified dividend tracking. No brokerage connection required.
Start Free Trial
Frequently asked questions
Which stock benefits most from the AI cloud capex thesis?
Microsoft (MSFT) has the cleanest story because its OpenAI partnership feeds Azure directly, turning AI spending into rentable cloud revenue quickly. Amazon's AWS wins on scale and custom Trainium chips, while Alphabet's Google Cloud grows fastest off a smaller base. All three tie capex to cloud, but Microsoft's link is tightest.
What is AI capex and why does it matter to investors?
AI capex is the money hyperscalers spend on data centers, GPUs, and networking to train and run AI models. It matters because it's huge, often $15 to $25 billion per quarter per company, and the market now rewards it only when it converts into growing cloud revenue rather than speculative capacity.
How much of the S&P 500 is exposed to Big Tech AI spending?
The Magnificent Seven make up roughly a third of the S&P 500 by weight, so most index investors already hold heavy AI-capex exposure. Owning MSFT, AMZN, or GOOGL individually on top of an index fund stacks the same bet, concentrating risk more than the portfolio appears to show.
How can I track my total AI cloud exposure across brokers?
Use PortfolioTrackr to tag MSFT, AMZN, GOOGL, and related ETFs into a custom group, then view that group as a percentage of your net worth. Connecting accounts from Interactive Brokers, Schwab, or Alpaca consolidates everything so overlapping AI exposure is visible instead of hidden across separate broker apps.
What signals show the AI cloud capex thesis is breaking down?
Watch for capex growth outpacing cloud revenue growth, falling operating margins from data-center depreciation, and cheaper AI models cutting compute demand per query. If a hyperscaler raises spending 50% while cloud revenue grows only 20%, the payoff is stretching further out and multiples become vulnerable to compression.