Big Tech Is Burning Cash on AI — Here's Who Actually Gets It Back

[00:00:00]Big Tech is locked in a massive spending war. Microsoft, Amazon, Alphabet, and Meta, they're all pouring hundreds of billions of dollars into data centers, power grids, and custom chips. But, Wall Street is kind of starting to lose patience, and everyone is asking the same question right now, which is essentially, when do shareholders actually get their money back? Now, the everyday headlines might make it look like a simple software race, but it's actually a physical infrastructure battle. The wrong move could trap billions of dollars in dead cash for the next decade. And so, today Jose and I are going to be talking about the ultimate question on a lot of investors' minds right now. Which hyperscalers actually winning the AI spending war?

[00:00:39]And now, I'm going to bring Jose in to look at the big macro picture.

[00:00:43]>> Now, before we continue with today's episode, if you want market-beating stock picks from our analysts, make sure to check out the pinned comment and the description. Using that link gets you a promotional offer as our thanks for being a viewer. Thank you, and let's get back to today's episode.

[00:00:58]>> Yeah, thank you, Rachel. And I mean, I I I agree, Rachel, that the crazy amount of money entering this space is just so hard to wrap your head around. When you look at the big four, including Meta, so Amazon, Microsoft, Google, and Meta, total annual spending is tracking toward an eye-watering 700 to 800 billion dollars. This goes way beyond just ordering AI chips. These are companies buying crazy amount of real estate and funding massive electrical grid projects just to keep the data centers running.

[00:01:30]>> That's a really important point, right?

[00:01:31]Cuz on the upside, these firms, they're essentially building the computing utility grid for what could be decades, even the next century ahead. But, the market is, and perhaps understandably so, judging their balance sheets very differently. You could take Meta, for example. They're deploying tens of billions of dollars into specialized data centers to optimize, you know, Instagram and Facebook ads to really drive that advertising engine from which they make their money. And it's sort of a defensive tax to protect their ad margins, and it's working. But the flip side of it is it's also putting really heavy pressure on their immediate cash flows.

[00:02:08]>> Exactly, Rachel. I mean, you see the same pressure hitting other companies like Microsoft. Their early investments in OpenAI gave them a great software AI model lead out of the gate. But look at the balance sheet risk right now. Their capital spending just backed by over 80% year over year. Now, to be fair, Microsoft's core operating cash flow remains crazy strong. They are still printing money at the baseline, but the building pace has caused their free cash flow to drop, which has significantly repriced their free cash flow premium, proving that they are stuck in this kind of cycle or wheel, where they have to keep spending just to host AI models like Anthropic and OpenAI.

[00:02:51]>> Well, and I think this is the paradigm that a lot of these companies are facing right now. I mean, for example, if Microsoft were to stop building, OpenAI's development slows down. There's this constant pressure, but it introduces a risk that I think a lot of analysts, including us, are closely watching, and that is circular funding.

[00:03:08]I mean, this has become a huge trend in the tech sector. And what I mean by that, what is circular funding? So, you have a tech giant that, say, puts $10 billion into a hot new AI startup. Well, on paper, that might look like a standard venture capital investment. But the fine print in the contract often compels that startup to turn around and spend that exact same $10 billion renting data center space right back from that specific investor or, you know, hyperscaler, if you will. And so, that is something that we are continuing to see within the space, and it's raising a lot of questions about the sustainability of the buildout.

[00:03:39]>> Yeah, obviously, Rachel, the optics looks a bit scary, right? Think about the accounting flow. The parent company hands over cash from its overall investment ledger, and a few quarters later, that exact same cash comes right back into their cloud division as a recorded customer revenue. The upside is that it jump starts the developer ecosystem, which is crucial, and makes cloud growth look incredible. But, the real structural risk happens if these startups fail to build a self-sustaining business model before that injected cash runs out. Once they run out of that money, that specific cloud revenue, in theory, would completely vanish.

[00:04:17]Luckily, though, luckily, at least right now, Rachel, with at least these types of deals, we're seeing with the big players like OpenAI and Anthropic, which do tend to have a massive demand at the moment.

[00:04:27]>> Yeah, I think the concern is that if the startup can't find real paying end users, we have a problem, right? If the users aren't paying the startup, the startup can't pay the hyperscaler, and that high cloud growth rate can stall out. And I think that risk factor, among others, is why some investors are starting to prioritize companies that are relying on a wide range of corporate clients and are also working to lower their own chip costs. You know, we're not here to say that Microsoft are better or bad companies by any means, but I do think these are real risks that investors should be aware of.

[00:04:56]>> Agreed, and that's why this whole capital landscape is shifting. I mean, Amazon has initiated a $200 billion capital deployment, while Alphabet, as we were talking, a full range between 180 and 190 billion. Both companies are moving away from just your venture hype cycles and focusing on building their own AI chips to lower the cost of total ownership for the data center build-out that we're seeing.

[00:05:20]>> Yeah, I mean, a lot of this is also about who can serve an AI workload at the lowest cost per token. And I'm going to take a look at how Alphabet is managing this transition. You know, Alphabet's latest earnings, I think, were just another example of them crushing the bear thesis, right? I mean, we saw over the last few years a lot of critics suggest that AI would destroy Google search margins, that Google Cloud was, you know, stuck in a distant third place behind AWS and Azure. But, I think that the data is telling a very different story, right? So, Google Cloud's quarterly revenue reached $20 billion in the most recent quarter. That was up 63% year-over-year. That is an acceleration that's almost unheard of.

[00:05:59]It easily quite handily beat Wall Street's expectations.

[00:06:02]But, it's really the strategy behind this number that I think investors need to pay attention to. You know, this growth is driven by established corporate accounts moving their legacy data into active machine learning and AI environments. You know, this is also backed up by Google Cloud's contracted future backlog. That nearly doubled quarter-on-quarter. That hit $460 billion. You know, current revenue tells you what happened last quarter, but backlog tells you what is contractually guaranteed to hit the business over the next 3 to 5 years. So, that's something that's also really important for investors to understand. And there's another really key structural advantage here, and that comes from Alphabet's decade-long investment in its own custom chips. It's TPUs.

[00:06:47]You know, Google has been refining its own internal silicon for years now, and by offering buyers direct access to their TPU hardware, they are really actively working to decouple the profit margins from external component suppliers. Obviously, they're still reliant on suppliers, but this is a trend that we are seeing where they're increasingly offering that access.

[00:07:08]You know, we had seen obviously that Alphabet pushed their full-year capital spending target up towards as high as $190 billion.

[00:07:16]You know, the market didn't like that.

[00:07:18]But, when you're seeing that paired with a $460 billion committed enterprise pipeline, that spending can really transition from speculative building to locking down capacity for guaranteed future cash flows. And remember, they also have over 350 million premium subscribers across YouTube and Google One that they can monetize directly without paying high customer acquisition costs.

[00:07:40]>> Agree that $460 billion backlog provides strong long-term visibility, and I am bullish on this segment, Rachel. But, we always have to look at that operational risk. While Google directing 190 billion towards data center, their fixed server depreciation costs are creeping up. If corporate clients take longer than expected to build actual profitable applications out of that backlog, Google will be carrying kind of this massive fixed bills on underutilized capacity.

[00:08:12]How safe would you say is their bottom line if corporate budget pulls back next year? Now, before I give you this, Rachel, I do want to say it's fair to say that this risk is not just for Google, but right now for all the players here in the cloud space business.

[00:08:27]>> Yeah, I mean, it's a valid risk. I But, I think one of the things that makes me increasingly confident in Alphabet is is the strength of their core ad engine, right? I mean, Google search, obviously, YouTube advertising as well, continue to operate at a really high level of efficiency. It's generating a tremendous amount of cash flow. So, you know, they aren't leveraging their balance sheet with heavy debt to fund this cloud and AI expansion. They're very much covering it out of current operating cash. Um and I think that that is something that investors need to pay uh attention to. I mean, they were also able to initiate an increase to their quarterly dividend while absorbing the costs of this hardware cycle. So, that really gives them, I think, the continued uh trajectory they need even to write out any near-term enterprise deceleration.

[00:09:14]>> Yeah, that cash generation from the ad business is a strong stabilizer. But, what about enterprise positioning, Rachel? Microsoft, for example, has been decades deeply embedded inside global corporate businesses. If enterprise compliance officers view Google as a direct competitor in consumer apps, will they hesitate to store proprietary data at layers on Google Cloud?

[00:09:37]>> You know, at the end of the day, enterprises go where their data already lives. You know, companies are not going to spend millions of dollars migrating their information over to a competitor uh just for the sake of you know, provider neutrality, whatever you want to call it. Uh a significant portion of the world's structured enterprise data already resides natively inside Google Analytics, BigQuery, which is also a Google property. And Google has historically won because they have all that data hosted in-house and running it now on their own internal TPUs is the cheapest, most efficient option on the block for uh developers. Uh but you know, looking at on the infrastructure side, Amazon is also showing a huge surge of growth, and that's something that I think a lot of investors are paying attention to.

[00:10:19]>> Yeah, I mean, if you would have talked about Amazon about a year ago, you would have marketers have suggested that it was caught pretty much flat-footed uh in this AI race. But looking at the actual data, AWS is executing a large-scale infrastructure expansion aimed at positioning their cloud as the primary gateway for corporate AI workloads.

[00:10:39]Looking at their most recent earnings Q1 results, AWS recorded 28% year-over-year revenue growth. While that percentage is lower than Google on the relative basis, remember the sheer size of AWS. This is the fastest growth rate AWS has posted in 15 quarters, moving the division to a historic 150 billion annual revenue run rate. For the quarter, AWS revenue came in at 37.6 billion, reflecting a 2 billion sequential increase over the prior quarter, the largest Q4 to Q1 monetary increase in in the history of the division. But the real story here comes into the custom chip engine, Trainium.

[00:11:23]Amazon's custom silicon business has reached a 20 billion annual revenue run rate, growing at triple-digit percentage numbers. On the standalone basis, this run rate would make Amazon one of the top three data center chip companies in the entire world. And Wall Street completely missed that in the quarter one print is that Amazon is also part of the CPU layer. They just locked in a historic multi-billion dollar deal with Meta to deploy tens of million of custom AWS Gravitons, which are their CPU products. Why is this massive? Because the market thinks AI is only about training large models on GPUs. But as Meta advances into a gen AI systems, that takes real time and real actions, it execute codes, and it searches, it uses tools over and over again. The compute model shifts heavily back to high performance CPUs, and Meta is diversifying into Amazon's custom silicon because no single architecture can handle a gen AI workloads cleanly.

[00:12:26]When you add that to the fact that Amazon has over 225 billion in long-term revenue commitments strictly for their AI chips, and then overall 346 billion contracted backlog for their AWS, their pipeline is pristine. They are spending 200 billion dollars on infrastructure because actual paying enterprise demand is outstripping supply. They are building a physical tool booth that every enterprise has to pay, whether they're running GPUs or CPUs or other types of AI chip solutions.

[00:12:58]>> Yeah, those Tranium commitments are pretty substantial. But I think the one thing that's worth noting is to support this 200 billion dollar buildout, Amazon has utilized the debt markets for a 25 billion dollar bond issuance. And that increased their long-term debt position to around 120 billion dollars. And they also saw a little bit of a downward adjustment in their operating margins, just about a 2% decline. But I think I think the question is if Amazon is expanding its leverage profile, maybe its cloud operating margins are compressing slightly. Do you think there's a risk there for long-term shareholders or do you do you think this is maybe uh you know, short-term headwinds?

[00:13:35]>> Hm, Rachel, I think uh the the minor margin adjustment reflects an accounting timing mismatch rather than a structural demand issue.

[00:13:46]Amazon is front-loading cash outlays to secure real estate uh to lock semiconductor production allocation, to establish power infrastructure link.

[00:13:55]These upfront developments cause affect the operating ledger immediately. But, that backlog is recognized over a multi-year period. As those data centers are brought online, the debt deployment is in my opinion a deliberate capital choice to establish long-term infrastructure assets while preserving their double-A credit rating.

[00:14:17]>> I think the other thing to consider is uh also the chip landscape. I mean, training capacity is heavily booked today. We're seeing Nvidia iterating its product roadmap rapidly though. So, I think there's also the question if third-party hardware supply increases, we see pricing declines over the next year, is there a risk that Amazon's internal silicon strategy loses that cost advantage we're talking about?

[00:14:38]>> I am Amazon's custom silicon isn't, in my opinion, designed to compete with Nvidia on peak performance. Their focus is mainly optimizing optimizing price performance mainly for inference, which is that daily running and fine-tuning of models by enterprises. For the majority of corporate applications, renting top-tier third-party chips for a standard data processing is economically inefficient.

[00:15:05]Now, the data indicates that the infrastructure spending war is shifting into an operational execution phase.

[00:15:11]This level of deployment is tied directly to expanding enterprise revenue run rates across the sector. Whether you prioritize Alphabet's 63% cloud growth and internal chip advantage or Amazon's $150 billion AWS run rate and neutral infrastructure model, the physical cloud layer is capturing a significant share of the current spending. We want to hear your perspective as well. Which infrastructure strategy do you think is going to win over the 5-year horizon?

[00:15:40]Are you buying Google? Are you buying Amazon? Are you buying Microsoft or Meta? And if you found this video useful, click the like button, subscribe to the channel, and share the video. If you want market-beating stock picks from our analysts, check out fool.com/epic-50.

[00:15:56]Thank you for watching, and we'll see you in the next market update.