The One Hyperscaler That Stands Out Among Rising AI Capex Concerns
Assessing each hyperscaler’s price targets based on their AI capex expansion plans.
«The 2-minute version»
With Meta, Amazon, Alphabet, and Microsoft projected to grow their capital expenditure by 46% YoY to over $317B, it comes at a time when their revenues and earnings are also projected to slow down due to tougher comps.
Reason to worry? Not yet, especially given that Microsoft, Alphabet, and Amazon are all facing capacity constraints, particularly power, as they build out the necessary infrastructure and kits to support the growing demand for AI workloads.
Plus: With DeepSeek’s R1 innovation, hyperscaler CEOs are generally optimistic about the business implications, as a cheaper and more efficient model might spearhead enterprise and customer adoption of AI across use cases, leading to a higher demand for compute. On top of that, Mark Zuckerberg has ambitious plans, where he sees Meta AI becoming the leading AI assistant in 2025, while his vision to make glasses the form factor of AI may also come to fruition. All of that requires strategic long-term investments in infrastructure in order to deliver the required quality of service and scale.
But there is a catch: You see, with higher capex, it usually accompanies higher depreciation expense, which can eat into overall profitability. While Meta is pursuing efficiencies by extending the life of its useful servers, Amazon and Alphabet CEOs, Andy Jassy and Sundar Pichai admitted that they may face short-term P&L headwinds from depreciation-related expenses. In the meantime, Microsoft is underperforming among the whole hyperscaler complex as it sees a disconnect in its earnings and free cash flow per share.
Will there be one hyperscaler outperformer in 2025? While Meta stole the show in 2024, will it be able to maintain its lead? Investors are increasingly paying attention to revenue contribution from AI products and services across hyperscalers. For instance, Microsoft has surpassed $13B in AI annual revenue run rate, growing 175% YoY driven by its Azure AI services and enterprise Copilot adoption.
Let’s Set The Stage
Big Tech is about to make it rain AI.
That’s right.
Meta META 0.00%↑, Alphabet GOOG 0.00%↑ , Microsoft MSFT 0.00%↑ , and Amazon AMZN 0.00%↑ collectively plan to spend over $318B on AI and data center capex . this year, up 46% from last year.
Here’s how that breaks out:
Quite a splurge, I must say. 💰💰💰
During their respective earnings calls, all four CEOs emphasized the spending as existential, a “once-in-a-lifetime opportunity” to capture the market, especially as we see an explosion of companies building with AI (in the coming years) accompanied by declining compute costs, thus leading to widespread AI adoption among enterprises and consumers across use cases.
However, unlike in 2024, revenue and earnings growth rates are expected to slow down across all four companies. It’s nothing to get immediately alarmed about. But, given that these companies reported an average revenue and earnings growth rate of over 15% and 50%, respectively, in 2024, they face some tough comps in 2025.
Add to that the growing macroeconomic uncertainty from the unknown impact of tariffs on earnings growth, and you may be wondering which hyperscaler is the best positioned to deliver alpha in 2025.
That’s what we are about to find out.
⏰ We are running a limited time offer for all new annual subscribers to lock in $80 for a lifetime. Paid subscribers will get full access to 9-12 Research posts like this one every quarter with select stock picks in the AI ecosystem. Prices will go up higher at the end of March.
Capex is growing across all 4 hyperscalers- that’s not a bad thing
Over the last few weeks, there has been a lot of chatter around AI capex, especially after DeepSeek’s R1 model launch rattled investors on the broader implication of such an innovation on the overall AI landscape.
While some argued that a decline in AI computing costs might accompany a demand drawdown for expensive AI chips and cloud infrastructure, others believe (hint: Nadella’s Jevons Paradox tweet) that cheaper models like R1 might actually unleash a new wave of organizations building with AI, translating into higher overall demand for compute.
There are three things that stood out to us from the latest earnings calls of all four companies that impact the short-term and long-term growth narrative of the hyperscalers.
1. Hyperscalers Are ‘Capacity Constrained,’ a.k.a. Still Racing to Meet AI Demand
When it comes to building AI data center capacity, there are 2 components to it: infrastructure (that consists of land, power, etc.) and kits (CPUs, GPUs, networking, etc.).
During the earnings calls, Microsoft, Amazon, and Alphabet all admitted to facing “capacity constraints”, particularly power, and expect similar conditions to exist in the next quarter, before easing in the second half of 2025.
Furthermore, Microsoft’s CFO, Amy Hood, also outlined that they have pivoted their capex structure to allocate over half of their cloud and AI-related spend towards procuring land and power resources. These are also known as long-lived assets, which will support monetization over the next fifteen years. In the meantime, Nadella also said in an interview that Microsoft is no longer chip constrained, especially after it bought nearly double the amount of Nvidia’s NVDA 0.00%↑ Hopper GPUs in 2024.
The same is also true for Amazon, where it spent $83B in capex in FY24, up 57% YoY, with the vast majority allocated to building AI capacity for AWS (Amazon Web Services), while the remaining portion is allocated to its Stores (North America and International segments) business with the aim to improve the delivery speed and cost to serve.
Finally, Alphabet is also directing its AI capex towards servers and data centers in order to support the growth of their business across Google Services, Google Cloud, and Google DeepMind. In 2024, the company expanded the GCP (Google Cloud Platform) network through 11 new regions and 7 subsea cable projects, with data centers now delivering 4x more compute per unit of electricity than five years ago. Along with that, cloud usage has also surged over 8x in the past 18 months. However, when probed on whether revenue growth could have been higher with more capacity, this is what Alphabet’s CFO Anat Ashkenazi said during the earnings call:
“We do see and have been seeing very strong demand for AI products in the fourth quarter in 2024. We exited the year with more demand than we had available capacity. So we are in a tight supply demand situation, working very hard to bring more capacity online. As I mentioned, we've increased investment in CapEx in 2024, continue to increase in 2025, and will bring more capacity throughout the year.”
2. DeepSeek’s R1 Model Catalyzes AI Demand Further - Hence More Capex
On the topic of DeepSeek’s R1 reasoning model, all four CEOs sounded optimistic about its long-term business implications for their companies and therefore believe that it is in their best interests to stick to their long-term outlook on capex, assuming that compute demand will be significantly higher in the coming years.
But as Andy Jassy, CEO of Amazon, pointed out, the demand curve for compute may not be a straight line. In fact, it may be lumpy in the short to mid-term, especially given the timing of enterprise adoption cycles, capacity considerations, and technology advancements.
But that doesn't change the long-term outlook. Instead, failure to secure the necessary land and power resources could lead to them being forced to secure sub-optimal land with worse price/performance profiles or lose business to competitors who have capacity.
Plus, let’s not forget the fact that these companies are also sitting on massive war chests of cash and highly profitable core businesses that can fund their AI-related capex without them needing to secure the equity and debt financing that is typical of such investment booms.
This is how Andy Jassy described its capex strategy to align with its long-term strategic vision:
“The way the AWS business works and the way the cash cycle works is that the faster we grow, the more CapEx we end up spending because we have to procure data center and hardware and chips and networking gear ahead of when we're able to monetize it.
We don't procure it unless we see significant signals of demand. And so, when AWS is expanding its CapEx, particularly in what we think is one of these once-in-a-lifetime type of business opportunities like AI represents, I think it's actually quite a good sign medium-to-long term for the AWS business.
And while it may be hard for some to fathom a world where virtually every app has generative AI infusing it with inference being a core building block just like compute, storage and database, and most companies having their own agents that accomplish various tasks and interact with one another. This is the world we're thinking about all the time and we continue to believe that this world will mostly be built on top of the cloud with the largest portion of it on AWS.”
At the same time, Sundar Pichai, CEO of Alphabet, also eased investor concerns on the commoditization risk of its Gemini models with the rise of open-source ones like R, stating both their 2.0 flash models compare to DeepSeek’s V3 and R1, while their competitive strength lies in the full-stack development and end-to-end optimization, which should solidify their positioning to support AI workloads across billions of users and on the cloud side.
Finally, Meta also has ambitious plans for GenAI, where Mark Zuckerberg, CEO of Meta, has set far-reaching goals for Meta AI to become the leading AI assistant in 2025, while Llama and open source become the most widely used AI models.
During the earnings call, the management laid out its FY25 capex plans, where it expects to spend roughly $60-65B, as it brings 1GW of capacity in 2025, while building a 2GW AI data center to support both their GenAI and Core AI initiatives to deepen user engagement, drive superior advertising revenue and prepare for liftoff for its Meta AI glasses (as part of its Reality Labs segment).
3. Capex Growth Should Normalize Next Year
However, the rate of growth in capex should moderate from 2026 onwards, as can be seen below.
Particularly, when it came to Microsoft, Amy Hood outlined that the growth rate in their capex should slow in FY26 (they currently finished reporting their Q2 FY25 earnings), especially as they see capacity constraints ease to better match demand.
While Amazon, Alphabet, and Meta haven’t discussed their capex plans beyond FY25, we can expect a similar trajectory in capex, which should boost the overall free cash flow of these businesses in the long term.
Everything Has A Cost, Including Capex
However, on a shorter time horizon, there is no denying that these companies are seeing a significant expansion in capex, particularly capex as a percentage of their projected 2025 revenue.
Here’s how it breaks down:
While we have already discussed the strategic decisions behind the expansion in capex, it has a short-term cost to it. Larger depreciation expense.
Simply put, depreciation is an accounting practice that allows businesses to spread the cost of physical assets over a period of years.
Therefore, when capex grows at an accelerated pace, it leads to higher depreciation expense, which could pressurize overall operating and free cash flow margins. Unless, of course, the company manages to grow its revenues at a faster rate or pursue efficiencies in other forms.
Let’s start with Meta.👇
Over the last several quarters, Meta has seen its operating margins expand as a result of growing ad monetization efficiency and streamlining operating expenses, as can be seen below.
However, in FY25, its revenue growth rate is expected to slow down to the mid-teens range, down from 21.9% a year ago. However, after learning a costly lesson in 2022, Meta is already pursuing efficiencies by extending the useful life of their servers and associated networking equipment, where they estimate that they will be able to use both their non-AI and AI servers for 5.5 years, translating into capex savings and a decrease in depreciation expense.
At the same time, what is also helping Meta anchor investor sentiment is that its earnings and free cash flow per share are moving in lockstep, thus boosting overall confidence in the company’s execution and long-term strategy.
On the other hand, Amazon and Alphabet have clearly outlined that they expect to see margin headwinds from depreciation-related expenses in FY25.
Particularly when it comes to Amazon, the management is expecting to see a total headwind of $1.3B in operating income for its AWS segment from a combination of retiring and reducing the useful life of their servers and networking equipment. I will also point out that AWS is the most profitable business segment in Amazon, responsible for contributing over 50% to Total Operating Income.
Finally, Microsoft is currently facing the brunt of investor pessimism, despite its 31-32% YoY forward growth projection for its Azure cloud, which I believe is quite strong. Part of the weakness is likely tied to the fact that, unlike Meta, it is seeing a disconnect in its earnings and free cash flow per share, as can be seen below. In fact, in its latest earnings call, its free cash flow plummeted 29% as a result of growing capex.
However, I believe that it should be a short-lived phenomenon, especially as the growth rate of capex normalizes at the end of FY25 (mid-2025), with capacity constraints easing to better match the demand landscape.