Arm FQ3 26 & Amazon Q4 25 – Different Takes.

Arm FQ3 2026 – Smartphone proof

• Arm reported good results and guided nicely for the coming quarter in an indication that the data centre business and product mix have offset the memory shortage that is holding up the smartphone market.
• FQ3 revenues / Adj-EPS were $1.24bn / $0.43, just ahead of consensus of $1.23bn / $0.41 as Arm’s market share in both the regular data centre and AI data centres continues to grow, and its content in smartphones also did well.
• The market share of Neoverse-based chips is expected to pass 50% this year and the computational content of smartphones is continuing to rise.
• The data centre is now large enough at Arm to offset any weakness coming from the smartphone market, which is what I think allowed Arm to guide nicely where both Qualcomm and MediaTek were forced to include the impact in their expectations for the next three months.
• Here, FQ$ revenues / Adj-EPS are expected to be $1.42bn – $1.52bn ($1.47bn) / $0.54 – $0.62 ($0.58), which is in line with consensus and represents steady growth YoY.
• As a supplier of intellectual property and processor designs, subsystems and potentially, one day, chips, Arm fits the profile of a supplier to the AI capex splurge and should continue to benefit from it.
• This is what is driving Arm’s expectation that its revenue from the data centre will eclipse revenue from smartphones in a couple of years.
• Furthermore, I don’t think that Nvidia’s deal with Intel does it very much damage, as increasingly, the data centre is all about power consumption, which is one area where Arm has always and continues to fare better than x86 designs.
• Hence, I think that the estimates for Q4 26 and fiscal 2027 look pretty secure, which combined with volatility in the share price are helping the valuation.
• Arm has been on 100x or more PER, which is now ameliorating to 62.9x for the year ending March 2026 and 50.5x for March 2027.
• These are still big numbers, but steady progress is being made, meaning that unless there is a large AI-related correction, the shares should be able to hold current levels even in the volatile market conditions in which we find ourselves.

Amazon Q4 25 – A crazy battle of one-upsmanship

• Another massive capex bill weighs on the outlook for Amazon, which appears to be playing a crazy game of one-upmanship that increasingly looks like it is going to end in tears.
• Q4 2025 revenues / EPS were $213.4bn / $1.95, broadly in line with consensus of $211.4bn / $1.97, but profit guidance was soft, and the capex bill for 2026 will be truly stratospheric.
• Here, Q1 26 revenues / operating profits will be $173.5bn – $178.5bn ($176.0bn) / $16.5bn – $21.5bn ($19.0bn), which were in line with revenue of $175.7bn but behind operating profit estimates of $22.2bn.
• The good news is that AWS is finally seeing some of the growth that has ignited its rivals Azure (+39% YoY) and Google Cloud (+48% YoY), but its size means that it has been more constrained at 24% YoY.
• In order to achieve this, it has decided to spend a gargantuan $200bn on data centre capex in 2026, which is already showing signs of hurting profitability thanks to the inevitable jump in depreciation that will follow these investments.
• The problem for Amazon (unlike Google) is that it does not have the cash flow to support a spending binge of this magnitude.
• Cash Flow From Operations in FY 2025 was $139bn, and with only $122bn on the balance sheet, Amazon will have to start raising money from the market if this continues into 2027.
• Google, by contrast, has much stronger cash flow, and so as long as the core business continues to deliver, it can continue to spend at these crazy levels without materially damaging the balance sheet.
• The problem I have with all of these investments is that their size is really only because of everyone else.
• Everyone is afraid that one of their competitors will get to artificial super-intelligence first, which has triggered a crazy game of one-upmanship as everyone tries to outspend everyone else.
• The market is fully aware of this, which is why Amazon’s share price is down 10% in after-hours while those that can better afford to play this game have seen a more muted response.
• The other problem is that the business model of compute is fundamentally challenged, and the data centres either need to get cheaper to build or the revenue earned per GW needs to increase to get them onto an economically viable footing.
• In this environment, the only place to be is in the picks and shovels of the spending boom and here Nvidia, AMD, Micron, Broadcom, Samsung and SK Hynix are in the front line.
• However, these have all increased their valuations substantially, and so it is in the others which the market has largely ignored, where the most upside remains.
• These would be Qualcomm and MediaTek, where I am looking closely at MediaTek and have increased my position in Qualcomm, and I am still holding onto Samsung Electronics.
• I would not touch Amazon, Microsoft, OpenAI or Meta as their profitability will come under pressure once the depreciation charges kick in.