JPM | APAC THEMATICS - Tech... (i) Positioning: Degrossing done?? / (ii) Fundamentals: META & MSFT not enough to shift the narrative / (iii) Valuations: Finally, a case for Memory on P/B.

Research evidence excerpt

JPM | APAC THEMATICS - Tech... (i) Positioning: Degrossing done?? / (ii) Fundamentals: META & MSFT not enough to shift the narrative / (iii) Valuations: Finally, a case for Memory on P/B.

Specialist Sales

APAC Specialist Sales J P M O R G A N

30 July 2026

(2) FUNDAMENTALS - Not enough from META or MSFT to change the near-term narrative

Hyperscaler results overnight failed to provide conviction around AI capex growth. Meta -8% (earnings and guidance miss) and

Microsoft +8% (Azure beat) had contrasting prints… but neither said anything new on capex, either in terms of raising 2026

capex or giving conviction that 2027 will be high enough that it can drive another round of upgrades. Meta tightened their 2026

capex guidance range but didn’t lift the upper end ($130-145bn vs $125-145bn prior)… and CFO Susan Li explicitly said they

will “not provide a specific capex outlook for 27 at this time”. Zuck said the demand for compute was strong and capacity tight -

but the miss on margins and 3Q revenues, was clearly not great. As for Microsoft, they didn’t provide clarity on CY27 capex, just

guiding it would “continue to grow”.

Duration of capex growth remains the focus. Gokul has been saying for a while the debate in tech is switching from “peak” to

“duration” of capex. Tech skeptics say that when the shift from growth to duration happens… the cycle is starting to fade. Maybe.

What is true though, is that it’s more difficult for the market to get its dopamine hit around knowing where 27 or 28 capex will

land - with the hyperscalers hesitant to say. Recent deals by Nvidia to try to underpin the next leg of the capex cycle also hasn’t

made any difference - instead the rise in hyperscaler credit spreads was seen as more important. Ultimately, what we need to see is