UBS says tech pullback has opened up attractive entry points
UBS is using the recent tech drawdown to reinforce its bullish AI thesis rather than retreat from it, framing the sell-off as a valuation reset rather than a demand problem. The bank’s emphasis on token consumption and hyperscaler cloud growth points to where investors should look for confirmation that AI spending is being monetised. Semiconductor valuations compressing to around 22 times forward earnings, below their post-ChatGPT average, gives UBS scope to argue the risk reward has improved even as long-dated Treasury yields and AI financing concerns keep sentiment fragile. The call to diversify into defensive tech, such as payment networks and data centre REITs, signals UBS sees near-term volatility persisting even as its medium-term conviction stays intact.
—
UBS is treating the tech sell-off as an opportunity, not a warning sign, for the AI trade.
Summary:
- Nasdaq fell 1.9% and the Philadelphia Semiconductor Index dropped 8.2% over the past 10 days as rising long-dated Treasury yields and AI capex or financing concerns weighed on sentiment
- UBS points to 24% week-over-week growth in token consumption on OpenRouter data, with token volume up more than 30-fold over the past year, as evidence demand is outpacing efficiency-driven price declines
- UBS forecasts total AI capex reaching $1.2 trillion in 2027, up 33% from an estimated $900 billion this year
- Hyperscalers reported average cloud revenue growth of 48% year over year in the second quarter, beating consensus; UBS forecasts this accelerating further to 58% in the current quarter
- UBS cites Amazon CEO Andy Jassy’s comment that AI investment reflects a timing mismatch rather than a structural drag, with data centre capital able to generate revenue for more than 30 years
- Semiconductor valuations have eased to about 22 times forward earnings, below the post-ChatGPT average of 24 times and well under the recent peak of 33 times, alongside consensus earnings growth of 92% this year for key semiconductor names and 38% for Nasdaq 100 companies
UBS says the recent sell-off in technology shares looks more like an opportunity than a warning sign for the artificial intelligence trade, arguing that stronger token demand and accelerating cloud revenue growth continue to underpin its bullish outlook on the sector.
In a note to clients, the bank pointed to a difficult stretch for tech stocks, with the Nasdaq down 1.9% and the Philadelphia Semiconductor Index off 8.2% over the past 10 days as rising long-dated Treasury yields and lingering questions over the sustainability of AI capital expenditure and the circularity of some AI financing arrangements pressured sentiment. UBS said Nvidia’s earnings could add further volatility, with the market looking for both a solid earnings beat and forward guidance confirming that demand for AI chips remains intact.
Without commenting on any individual company, UBS said it retains conviction in the broader AI growth story and continues to view it as a central pillar of its positive outlook for markets. The bank pointed to token consumption, the basic unit of AI computing use, as one of the more reliable gauges of underlying demand. Citing OpenRouter data, UBS noted token volumes grew 24% week over week and have expanded more than thirtyfold over the past year. While token pricing has fallen this year, UBS said recent signs of stabilisation suggest demand growth is more than offsetting efficiency gains, a dynamic it expects to keep supporting AI capital spending. The bank now projects total AI capex will reach $1.2 trillion in 2027, a 33% increase on its estimate of $900 billion for this year.
UBS also highlighted evidence that monetisation is catching up with spending. Major hyperscalers reported average cloud revenue growth of 48% year over year in the second quarter, ahead of consensus and an acceleration from prior quarters, with UBS forecasting growth to reach 58% in the current quarter. The bank acknowledged the risk that investment could outpace returns, but referenced comments from Amazon chief executive Andy Jassy, who has suggested the AI investment cycle reflects a timing mismatch between spending and revenue rather than a structural drag, adding that data centre capital can support revenue generation for more than three decades. UBS said this kind of confidence in long-term free cash flow and returns on invested capital should help sustain the current investment cycle.
On valuations, UBS argued the tech volatility has created increasingly attractive entry points for select names. It noted a broad basket of major semiconductor companies now trades at around 22 times forward earnings, below the average of 24 times since the launch of ChatGPT in 2022 and well under the recent peak of 33 times, while earnings growth expectations continue to rise. Consensus estimates now point to 92% earnings growth this year for key semiconductor companies, followed by 40% growth in 2027, with Nasdaq 100 companies expected to grow earnings 38% this year and 26% next year.
Given this mix of solid near-term fundamentals and persistent risks, UBS recommends investors hold a selective, diversified allocation to technology stocks. The bank continues to favour the enabling layer of the AI value chain, including semiconductors and cloud computing, but suggests investors also consider building exposure to more defensive tech segments such as payment networks, data centre real estate investment trusts, and select smartphone makers.
This article was written by Eamonn Sheridan at investinglive.com.