Qualcomm, Meta & Microsoft – 2 out of 3

Qualcomm FQ2 25 – Unreasonable punishment

• Qualcomm reported good results, but a slightly weaker-than-expected revenue forecast caused the market, already nervous about tariffs, to send the shares down 6%, creating an opportunity for anyone with a time horizon longer than 5 minutes.
• FQ2 revenues / adj-EPS were $10.84bn / $2.85 just ahead of expectations of $10.84bn / $2.82.
• On the fundamentals, everything is ticking along nicely with Qualcomm pretty much doing exactly what it promised it would when it started the drive to diversify the company away from smartphones.
• The premium segment in Android is doing well, while PCs have taken some share and are on target to crush the 12% market share target set.
• Automotive remains small but keeps on growing and is well on track to be a significant contributor to financial performance in due course.
• Qualcomm Technology Licensing has also performed well, quietly renewing its contracts with only Huawei left to renew and churning out large amounts of cash.
• All of this good news was glossed over as soon as the forecast for FQ3 25 was made clear.
• Here, Qualcomm expects FQ3 25 revenues to be $9.9bn – $10.7bn ($10.3bn) compared to consensus of $10.33, making this a miss of $30m or 0.3% relative to the consensus estimate.
• This could easily have been caused by rounding, as Qualcomm is giving guidance to 1 decimal place while consensus is given to two.
• Compounding the irrationality was FQ3 adj-EPS which is expected to be $2.60 – $2.80 ($2.70) some 1.1% (or 4x more than the revenue miss) ahead of consensus at $2.67.
• This was a another very solid set of earnings, and as Qualcomm remains almost unaffected by tariffs, taking the shares down 6% is nonsensical.
• Hence, I am increasingly thinking that this is an opportunity to add to my position in Qualcomm, with which I remain extremely comfortable.

Microsoft FQ3 25 – Cloud but no rain

• Microsoft reported good FQ3 results as growth in Azure went back above 30% YoY, and the rest of the business remained steady and, crucially, pretty much immune from any tariff impact.
• FQ3 25 revenues / EPS were $70.1bn / $3.46 ahead of estimates at $68.4bn / $3.32.
• Most of this was due to Azure where growth is back above 30% and came in at 33% YoY.
• There had been some fears that Microsoft’s disentanglement from some data centre contracts like CoreWeave were a sign that AI demand was topping out, but this has not proven to be the case as I expected.
• Instead, the contract shuffling is all about Microsoft rethinking its relationship with OpenAI and no longer being the sole provider of compute and reducing its dependence on its technology.
• To me, this is a smart move as OpenAI is a badly run company where internal conflicts and contradictions could easily cause another existential implosion like the one at the end of 2023.
• I have long viewed this dependency as the one serious mistake that Satya Nadella has made in an otherwise almost faultless stint as CEO of Microsoft.
• The fact that Microsoft appears to have recognised this and is moving to fix this vulnerability is good news and is what I think was solely responsible for the contract juggling that we observed earlier this quarter.
• Elsewhere, Microsoft’s business continues to develop steadily, much like Qualcomm’s, and so with one of its key risks being mitigated, I would be willing to pay a higher multiple for the shares.
• However, at 28.1x FY 2026 PER, the shares are still pretty expensive, especially when I can have Google at less than 20x and Qualcomm at a number below even that.
• If I had to have a Mag 7 stock, it would be Google or Nvidia.

Meta Platform FQ1 25 – Even more datacentres please.

• Meta reported good results and immediately decided to spend the proceeds on yet more data centres as the race to build as much AI capacity as possible intensifies.
• Q1 25 revenues / adj-net income were $42.3bn (up 16% YoY) / $16.7bn, well ahead of estimates of $41.4bn / $13.5bn.
• Net Income was particularly impressive, growing 35% YoY and was partly driven by Meta using AI in its businesses to make them far more efficient than they have been historically.
• An unusually low tax rate of 9% compared to 13% in Q1 24 was also a significant contributor.
• Also key to revenue growth is the slow (6% YoY), but steady expansion of the daily active user base that now stands at 3.43bn.
• Guidance for Q2 25 was in line with expectations, with revenues of $42.5bn – $45.5bn ($44.0bn) and lower expenses than expected for the full year.
• However, Meta immediately decided to spend more on data centres and increased its capex forecast for 2025 to $64bn – $72bn from $60bn – $65bn.
• This partly reflects an acceleration in build, but also the expectation that data centres are becoming more expensive to build.
• Mr Zuckerberg is absolutely determined to be one of the leading platforms in AI, and his strategy to do this via open source is currently going quite well.
• However, this determination to never be dependent on anyone else’s platform is currently costing shareholders more than $4bn per quarter as Reality Labs continues to burn an eye-watering $46.7m every day.
• With the core business in great shape, this is not a real problem, but if the economy turns down and things wobble, investors will again start getting uncomfortable carrying this business.
• There is little doubt in my mind that data centres are being overbuilt which means that when the correction comes, Google, Microsoft, Meta and Amazon are going to have a problem.
• However, they believe that the risks of underbuilding and getting left behind are greater than the risk of overbuilding which only makes sense if one believes that we are close to the point at which the machines become more intelligent than humans.
• All of the empirical evidence suggests that we are not at this point, and so there will inevitably be a correction.
• This will clobber the pure plays like OpenAI, Anthropic, Mistral, and so on far harder than anyone else, but I still don’t particularly wish to be around when this happens.
• Meta is on 23.5x 2025 PER, putting it at the more reasonable end of the spectrum, but I would still prefer Google, where I think the risks of it losing its business to OpenAI are overstated.