Spending on AI grabs all the attention from this mixed bag
Microsoft FQ1 26 – Spitting distance to AWS.
- A mighty performance from Azure went hand in hand with yet another huge increase in capital expenditure that made even the most bullish investors slightly skittish.
- FQ1 26 revenues / EPS were $77.7bn / $3.72, just ahead of estimates of $76.6bn / $3.68.
- The star of the show was Azure, which grew revenues by 39% YoY, which would have been higher if Microsoft had not been capacity-constrained.
- This is why capex was up $10bn QoQ to $34.9bn in FQ1 26 and is expected to rise again in the coming quarter.
- Half of the capex is going on GPUs and CPUs, which have a 3-5 year financial life, while the other half is on much longer-lived assets like land leases for data centres and so on.
- The current growth looks set to continue, but given that Azure remains capacity-constrained, revenue guidance was at the lower end of expectations.
- FQ2 26 revenues are expected to be $79.5bn – $80.6bn ($80.1bn), which is only just in line with the consensus estimate of $80.1bn.
- This is in part due to the ongoing capacity constraint at Azure, which, combined with the ever-increasing capex bill, was not well received.
- This is why the shares fell by 4% in after-hours trading, highlighting the tight valuation.
- FY 26 PER is currently 33.4x with FY27 at 28.4x, which is high when considering some of the other AI-related investments that can be had.
- Here, I would highlight Alphabet (see below), TSMC, Samsung and Qualcomm, all of whom are AI-exposed but trade at much lower numbers.
- I hold Qualcomm and Samsung for precisely these reasons.
Meta Platforms Q3 25 – Other People’s Money
- Meta Platforms reported good results, but another monstrous capex bill that has no direct line of sight to revenues (unlike Azure, AWS and GCP) spooked investors, sending the shares down 7% in after-hours trading.
- Q3 25 revenues / Adj-EPS were $51.2bn / $7.25 ahead of estimates of $49.5bn / $6.72, but that was where the good news ended.
- The core business continues to fare well as users are steadily growing and Instagram is hitting 3bn MaUs during Q3 2025.
- This is what drove revenues, but at the other end of the financial statements, there are a lot of outgoings.
- Reality Labs lost another $4.4bn during the last 3 months, and $18.8bn was spent on data centre infrastructure.
- This is a very different proposition as all of this is for internal use and not for resale, as Meta does not operate a commercial cloud business.
- Hence, this is Mr Zuckerberg spending shareholders’ money (over which they have no say) on his pursuit of artificial general intelligence (a race he is unlikely to win if it is even winnable at all) and the Metaverse, where he has a much more justifiable investment case.
- 2025 capex is now expected to be $70bn – $72bn, up $3bn from prior outlook of $66bn – $72bn, and there is more to come in 2026 as “it has become clear that our compute needs have continued to expand meaningfully”.
- Meta expects that “capex dollar growth with be notably higher in 2026 than 2025”, where, assuming guidance is met, 2025 YoY capex will grow by a total of $35bn.
- This means that Meta is guiding to 2026 capex of “notably larger” than $105bn, which is a staggering number and screams peak bubble to any rational investor.
- This, combined with expenses growing faster in 2026 than they did in 2025, raises the possibility of very low or no EPS growth next year.
- It is here where we see Meta’s core risk once again, which is if Mr Zuckerberg decides to spend all of Meta’s money on a vanity project, there is nothing that anyone can do to stop him.
- All investors can do is vote with their feet, which is precisely why the shares fell in after-hours trading, and I would not be surprised to see more to come depending on what happens to earnings forecasts.
- Meta can be fantastically profitable if Mr Zuckerberg wants, and so there could easily be another huge bounce on the other side of this spending spree.
- Hence, I am staying well away from this for now.
Alphabet Q3 25 – More Justification for spending.
- Alphabet also raised capex expectations materially, but with a great set of results and its status as comfortably the main challenger to OpenAI, the market was happy to cheer the company on, lifting the shares by 8% in after-hours trading.
- Q3 2025 revenues-exTAC / EPS were $87.5bn / $2.87, nicely ahead of estimates of $85.1bn / $2.26 as the Google business remains completely unaffected by the use of generative AI at competing platforms.
- 15% revenue growth, combined with no growth in sales and marketing expenses, allowed profits to grow much quicker.
- Google Cloud also put in a good performance, growing by 33% YoY to $15.2bn in revenues, but this should have been better given Azure’s much larger size and even higher growth rate.
- The net result is that Google after a very shaky start, Google is holding OpenAI’s incursion into its territory at bay and with the huge ecosystem it already has, Google has a good chance of migrating its search business to generative AI and holding onto its market position.
- The growth outlook is steady, which, combined with assets like DeepMind, Waymo, YouTube, as well as a reasonable valuation, makes this the only one of the Mag 7 that I would want to own right now.
Take Home Message – Azure could take leadership.
- All eyes will now turn to Amazon, which reports earnings this evening and where I am not expecting great things from AWS.
- The evidence is pointing to a large part of the AI boom passing Amazon by, meaning that Azure could, in 2027, take market leadership from Amazon.
- Amazon remains a laggard in AI, and this is now having a tangible impact on its financial performance.
- I continue to have no interest in owning this stock.
