UBS analysts forecast AI spending will reach $1.1trillion in 2027, up from $900bn in 2026. In their own words, “the big numbers are probably in the infrastructure build-outs. Every month an increase in expectations of how much the hyper-scalers will spend.”
In a recent podcast from UBS in collaboration with the WSJ, co-heads of investment at UBS Asset Management Sonja Laud and Barry Gill on ‘The Red Thread: The Line that Shapes the Market’ talked about technology diffusion and early adopter advantage.
The presenter said, “Investors are looking at this and recognising that inflection point, that movement towards a more considered approach to AI use… in the context of asset management, when you look at… when it comes to constructing portfolios allocating their investments, what are the risk drivers in this new era of AI?”
The respondents said, “the challenge now is there’s an interconnectedness at the fundamental level and that then gets reflected at the security level… it’s the hidden correlations that exist between portfolios that bring you down.
The lack of stability of correlations is the other factor that needs to be considered, and I think both of those are going to present very significant risks to portfolios, but also because they’re going to drive dispersion will create a lot of opportunities for investors.”
On earnings (re)allocation
One example of the deep linking between big AI players is Micron Technology, on the cusp of releasing its quarterly results; the share price saw a marginal increase in trading on Friday (+0.16% MU). Its much vaunted partnership with Nvidia (+0.22% NVDA), though, will not take effect until next year.
Revenue forecasts for 2027 reallocated the projected income from Nvidia contracts into the later part of the year. Micron’s shares have grown 277% in value so far this year.
Nvidia is nearing completion of its Rubin artificial-intelligence chip platform, but Susquehanna analyst Mehdi Hossein said to MarketWatch that supply lags the order time by more than 30 weeks, and that average prices of these chips have increased over 30% sequentially.
“As such, we expect the Rubin platform to be a more meaningful demand driver for HBM4 in 2027,” with a “minimal” impact on revenue for this year.
Micron is staged to roll out its next-level high-bandwidth-memory technology (HBM4), but the analyst has had to transfer its income estimates from forward orders to the next reporting period, diminishing profit forecasts. The full advantage of the supplier contract with Nvidia will not, as originally thought, take place in H1 2027 but will take effect in the second half.
Hosseini said, “All in all we expect memory fundamentals to remain healthy” and is forecasting EPS to attain $200, above the consensus of $165.
First Past the Post
The importance of having an early-adopter advantage is epitomised by the launch of Meta’s new AI assistant Muse. A Thursday report from Sensor Tower revealed that Muse has had more than 3.4mn downloads in the US and Canada since its Sept 8 launch.
It also emerged that Gork, Space X’s (+0.44% SPCX) multiple agentic platform, has had 2.2mn downloads since its August launch. Morningstar analyst Malik Ahmed Khan says that, next to Google’s Alphabet (GOOGL +0.46%, GOOGL+0.61%), “Meta is the only company that has experience scaling multiple platforms to multi-billion-user levels.”
Behavioural insights have enabled the company to designate 400 usage cases for Muse from auditing unnecessary subscriptions to bargaining over parking tickets. Standalone product the Muse chain
“gives Meta another access point to its query AI ecosystem, complementing products such as its AI glasses and its broader suite of apps and service,” Counterpoint analyst Anshika Jain wrote in a Thursday note.
To address info security of its users, every Muse agent is segregated in its own Muse Secure VM, and Meta has assured the public Muse data will not be shared with its advertising business and subscribers. An upgraded security model, Muse Confidential VM, will be released later this year, encrypting all user conversations and data behind a private key even Meta cannot access.
But to process payments and confidential communication will need access to PII – third-party logins and payment information and authorisation. Portfolio manager at Harding Loevner, Urday Cheruvu, told MarketWatch, “I’m not sure the trust is there for anyone, the reason being this is so early and so new.”
Leave a comment