UBS says AI demand remains solid; hyperscaler cloud growth forecast at 58%

In UBS's view, semiconductor and hardware names are well supported heading into an earnings season where actual results, rather than storylines, are expected to drive the AI trade. The near-term test comes quickly: ASML reports on Wednesday, and the major cloud providers follow later in the month. If cloud expansion falls short of the acceleration UBS is forecasting, richly valued megacap stocks would probably suffer most. Rates are a second risk: a hot US inflation print on Wednesday that pushes Treasury yields higher would weigh on long-duration tech valuations regardless of fundamentals. UBS's choice of quality and defensive tech reflects the risk of sharp swings even in a supportive environment.

-

The AI surge, in UBS's view, is still running on genuine demand, but with the rally nearly five years old, the bank wants investors to be choosier about where they take part.

Summary:

• Per UBS, recent industry data still point to solid demand for AI compute and infrastructure.
• Sensor Tower data show Meta's Muse AI assistant has surpassed 6.6 million downloads and 1.8 million daily active users in its first month.
• TSMC posted a 51% gain in quarterly revenue, Samsung indicated quarterly operating profit had climbed nearly ninefold, Hon Hai topped sales forecasts and Marvell lifted its fiscal 2028 revenue projection.
• For the September quarter, UBS expects hyperscaler cloud revenue growth to average 58%, after 48% in the three months to June.
• The bank cautions that worries about growth, financing and safety could still bring bouts of volatility.
• It recommends a diversified stance weighted toward quality chip and hardware names, megacap platforms and defensive tech.

According to UBS, AI demand remains robust and fundamentals are likely to steer the AI trade through the coming earnings season; the bank pointed to strong chipmaker figures and the first signs that consumers are willing to use AI agents.

UBS highlighted Meta's Muse AI assistant, which, based on Sensor Tower data, has racked up more than 6.6 million downloads and 1.8 million daily active users a month after its launch. AI agents are programs that act on a user's behalf rather than simply answering questions. The bank said it was too early to measure how other recently launched agents are being adopted, but described the initial numbers as an encouraging indication that wider uptake could boost AI demand further.

Chipmakers are pointing in the same direction. TSMC's quarterly revenue rose 51%, while Samsung flagged quarterly operating profit up almost nine times ahead of its full report. Hon Hai Precision Industry, which assembles AI servers, came in above sales forecasts, while Marvell, a chip designer, lifted its projection for fiscal 2028 revenue. UBS expects third-quarter earnings to provide more evidence that demand for AI compute and infrastructure is strong.

Monetisation among the hyperscalers — the largest cloud providers — will be a central focus. In the September quarter, UBS expects average cloud revenue growth at the main providers to hit 58% year on year; the comparable figure for the June quarter was 48%. It sees consumer AI agents as a possible next growth engine, opening up revenue from subscriptions, advertising and transactions, and argues that broader adoption would push up cloud utilisation, making the connection between heavy infrastructure spending and revenue more visible.

UBS acknowledged the risks. With the AI trade approaching its fifth year, it said concerns about growth, financing and safety could go on causing bursts of volatility. It advised investors to keep participating in what it sees as a structural opportunity while being more selective, urging a diversified selection that leans toward quality chip and hardware stocks, megacap platforms and defensive tech firms.

The view faces a quick test, with ASML due on Wednesday and the major cloud providers scheduled later in the month.

---

Key terms explained

AI agents: Programs that act on a user's behalf — making bookings, purchases or arrangements — instead of just answering questions the way a chatbot does. Because they act as well as respond, they create new earning possibilities for businesses, from subscription fees to commissions on transactions.

Hyperscalers: The small group of technology giants that operate the world's largest cloud computing networks and lease computing power and storage space to companies. They are the largest buyers of AI chips and data centre equipment, making their spending and performance a key gauge for the entire AI trade.

AI compute: The core processing power needed to train and run AI models, provided mainly by specialised chips in large data centres. Growing demand for compute is what generates orders for chipmakers and server builders.

Monetisation: The way companies convert AI investment into revenue. Investors have questioned whether the enormous sums spent on data centres will pay off, making evidence of AI-related sales growth a closely watched item in each earnings season.

Cloud utilisation: The proportion of a cloud provider's installed capacity that is actually in use and being paid for. Higher utilisation means money spent on data centres is producing revenue rather than sitting idle.

AI server assembler: A company that builds the full computer systems that house AI chips, assembling processors, memory, cooling and networking into racks for data centres. Hon Hai, also called Foxconn, is the biggest.

Megacap platforms: The largest publicly listed technology companies, each worth well above a trillion dollars, whose products are used by billions of people and whose scale lets them profit from AI.

Defensive tech: Technology firms with more stable, less cyclical earnings, such as established software or services businesses, that usually hold up better when markets become volatile.

---

What does hyperscaler cloud growth mean?

The companies: Hyperscalers are the small handful of technology giants that run the world's largest cloud computing networks, mainly Amazon Web Services, Microsoft Azure and Google Cloud. They provide computing power, storage and AI tools to businesses over the internet, so customers do not have to build their own data centres.

What the growth figure measures: Hyperscaler cloud growth is the yearly increase in revenue from these cloud operations. Analysts frequently take an average across the main providers to get a sense of how fast demand for cloud services — and increasingly AI services — is growing.

Why it matters for AI: Hyperscalers spend tens of billions of dollars a year on data centres and AI chips. Their cloud revenue growth is the clearest indication of whether that spending is paying off. Faster growth points to customers genuinely using and paying for AI capacity, while slower growth would raise doubts about overbuilding.

Why markets watch it: Since hyperscalers buy more AI chips and servers than anyone else, their cloud results spread through the entire AI supply chain, from chipmakers such as TSMC to server builders such as Hon Hai.

Where it stands: UBS is looking for hyperscaler cloud revenue growth to come in at an average of 58% on a year-over-year basis during the September quarter, compared with 48% three months earlier. Whether those figures hold up will become clearer when the main cloud providers update the market later this month.

"The most advanced AI on the planet and the desk head tells me to go fetch his dry cleaning."