Top of the Morning: CEO Macro Briefing Book - Insights on AI
The desk argues that the accelerated adoption of artificial intelligence (AI) across U.S. businesses is reshaping operational efficiencies and influencing market dynamics. Per the full note , AI adoption is projected to grow, with estimates indicating that approximately 25% of businesses have incorporated AI technologies, varying significantly by sector. Notably, technology firms lead this trend with a 45% adoption rate, while sectors like utilities lag at around 11%, suggesting an uneven distribution that could have macroeconomic implications and varying effects on currency valuations moving forward.
What the desk is arguing
The desk posits that the ongoing integration of AI in business operations is a significant driver of future efficiency and profitability. Businesses are increasingly leveraging AI to streamline processes, which, as highlighted, is crucial in a competitive landscape. As AI penetration deepens in various sectors, expect profound impacts on productivity metrics and possibly inflationary trajectories.
Supporting this, UBS reports significant discrepancies across industries; while technology embraces AI robustly, utilities and other sectors are catching up quickly—indicating a broader shift in market expectations regarding efficiency gains. Given the potential for this transformation, traders should monitor related currency pairs closely, particularly those influenced by U.S. economic performance and sectoral productivity.
Where it sits in our coverage
Our consensus target for the relevant currency pair is 1.075, with a range from 1.04 to 1.12. Aligned firms include: - jpmorgan: 1.10 (Mar26) - bofa: 1.04 (Mar26)
This view aligns with jpmorgan's positioning at the upper end of the forecast spread, suggesting that there is mutual recognition of the potential positive impact of AI integration on economic output and, therefore, on currency valuations.
How other firms see it
Several firms, including jpmorgan and goldman, are aligned in seeing the benefits of AI on economic performance, anticipating that this will result in a more favorable economic outlook. Conversely, bofa takes a more cautious approach, reflecting concerns about potential regulatory hurdles or inflationary pressures arising from rapid AI adoption.
The trajectory of USD/JPY may closely mirror these evolving sentiments, particularly with respect to monetary policy shifts shaped by AI's economic impact. Watch for updates related to Central Bank communications that may influence market expectations in this context.
How firms align with this view
Aligned with the desk view
Contrary positioning
Key takeaways
- 01Approximately 25% of U.S. businesses have adopted AI, with significant sectoral variances.
- 02Technology firms lead in AI implementation at around 45%, while utilities lag at roughly 11%.
- 03AI adoption is expected to reshape operational efficiencies and could influence inflationary trends.
- 04Market participants should consider the implications of AI on productivity and currency valuations.
Market implications
Traders should monitor the USD/JPY exchange rate for signs of volatility driven by AI-influenced economic shifts. Additionally, watch for potential market movements around upcoming economic indicators that could reflect the impact of increased AI adoption on broader economic performance.
Risks to this view
The primary risk to this outlook is the emergence of regulatory constraints that could hinder or slow down AI adoption, dampening productivity gains. Additionally, any significant downturns in economic performance or increased inflation rates could lead to a reassessment of currency valuations linked to AI-driven expectations.
Hi everyone, Dan Cassidy here. Welcome back to Top of the Morning on the UBS Market Moves podcast channel. Today we are going to continue with our conversations around the CEO Macro Briefing Book series for today, focusing in on insights on artificial intelligence, AI.
Now this is a special edition of the regular presentation that harnesses the research of the Chief Investment Office for our business owner clients. With that, joining me here today from UBS CIO within UBS FSI, glad to have with me at the table, Senior Asset Allocation Strategist, Paul Hsiao. Paul, great to be with you here at the table.
Thank you for once again dropping by. Thank you very much for having me. Absolutely.
Now Paul, talk to us a bit about the state of AI adoption for businesses in the U.S. What have you been seeing? Yeah.
So Dan, just a step back. This presentation, CEO Macro Briefing Book Insights on AI, is just one of the latest presentations we have on AI. Earlier this year, the asset allocation team came out with AI economy, a roadmap showcasing how we're seeing how AI affects both macro and markets.
This is a more targeted presentation. It's basically 10 questions that we're hearing from our clients about the state of AI and adoption. So going back to the original question about what's the state of AI and business adoption in the U.S., what we're seeing is that it's growing.
A conservative estimate from the Census Bureau shows that around a quarter of businesses have adopted AI nationally with great variation. Technology firms are leading with around 45% and then less AI exposed industries like utilities is lagging behind the national average at around 11%, but they're one of the fastest growing segments. So throughout the U.S., we're seeing AI adoption climb.
Other indices that use different methodology like the RAMP AI Index that surveys around 70,000 firms within the U.S. about their paid AI subscriptions show a much higher adoption rate, about more than half of firms in the U.I. adopting AI in some form of business function. So again, with those sector differentiations, technology leading and then less AI exposed sectors lagging, but they're one of the fastest growing segments. As we're seeing this expansion of use cases across businesses, industries, Paul, what are businesses saying?
What kind of feedback have you been hearing in terms of AI usage at the moment? I think in a word, it's efficiency but not earnings. So people who do use AI within employees say that they're saving around one to four hours depending on what tasks they use and what sort of AI models they use.
Is it just chatbots or is it just more agentic functions that can automate a task? So that's between one to four hours given a specific task and I think that's a pretty large chunk of the workday that can be automated through AI. That all said, we take a look at the Q2 earnings season.
A lot of executives are saying that what we're seeing in the numbers which is AI adoption is increasing, but it's just not having that tangible impact on earnings just yet, but they're quite positive on that. One reason is that even with all these AI adoptions, it's quite supplemental rather than if we believe that this is a wholesale change just like how the internet was, the work has to be rethought of with AI as the forefront and that's something that a lot of businesses haven't done because it just takes a lot of time to get those processes in place. So it's kind of similar to what we saw at the turn of the century when we started adopting electricity.
That technology was there, but it took one or two decades to actually have night shifts or the work schedule around electricity actually in the office. So we're seeing that happening right now with the efficiency, but done not just in earnings. So you're seeing an increase in operating efficiency, not large scale disruption at this point.
That's right. So when we think about tokens, Paul, there has been a recent decline in token price. Do you believe that's reflective of lower demand?
So for our listeners, tokens are thought of as a basic building block or currency of AI usage. It's a string of texts that when you input a prompt, that prompt gets broken up into small strings of text. And then when you receive something, that also is usually a text function that is also an output.
And obviously more complex questions with more complex answers requires more tokens. So with any sort of laws of demand and supply, if we think that demand for overall AI computation is rising, we would see more competition of these tokens. But the opposite has happened in the last couple of weeks, even though that in the larger trend over the last, let's say, two years, the dollar amount per token usage has been steadily rising, no matter what front you have taken it.
There's a lot of reasons. One is just LLMs, large language models, have become much more efficient at processing these tokens. Two is that the processors made by large firms like NVIDIA, that architecture has been made with AI at the forefront, so they're processing these queries a lot more efficiently.
So that's a software and hardware reason why tokens are cheaper. And the third is that there's just more competition from internationally. I think about a year and a half ago, markets really reacted strongly to the Chinese release of DeepSeek that was a cheaper, more efficient model that at the time was thought to rival some of OpenAI's function.
And now I think the delineation is much more clear, where U.S. models are at the forefront of how complex, how agentic, how much computational power that they can do. So the complex reasoning things are really U.S.-led. But for a lot of firms that really are just using AI for more efficiency production, they don't need that expensive or as complex of a model.
So they're using less intensive models that use less tokens in order to prioritize this usage. And that has some pretty big business implications because for the median AI cost per employee, that's around $12 for employee for businesses. But for the top 1% of users, that number has risen to $7,500 a month from $2,500 in just January this year.
So for those power users, token usage is really a concern and should be for our business owner client segments thinking about how to budget properly for next year. Well, very helpful, Paul, when you put some numbers around it. So thank you for that clarity.
As we begin to close out the conversation, midterm elections here in the U.S. are just a couple of months away at this point. We've been talking about this for months now. AI is increasingly becoming a political point of interest.
I spoke to several of your colleagues recently about a data center buildout in particular, how that's impacting communities. How should businesses, Paul, think about AI as we head into the U.S. midterms? I think another tricky thing for businesses right now when they think about integrating AI to their workflows is what sort of regulation that they can expect because the pressure for regulation I think is building both nationally and locally.
You spoke about data centers and for a lot of folks, even those folks who enjoy the productivity benefits of AI, it's just that the grand public views AI as a negative in the U.S. which is quite unique amongst DMs because they see it as something that is increasing their utility costs, something that could potentially take away their jobs and also a big sort of social change that gets talked about in the same way we talk about highways or the internet or railways for example. So there's a lot of uncertainty around that. What the public is saying is that they are concerned about AI usage more than excited that they would rather have data centers not built in their backyard because it increases their utility costs.
At the same time, we do have both Republican and Democratic lawmakers pushing things like the Frontier Act in Congress that aims to put some guardrails over what AI adoption can or cannot do and how LMs can derive their sort of inputs. For example, copyright protection is something that's still a question mark for a lot of legislators. That all said, we see that states are leading the way, California, Illinois, Colorado, New York have put in some sort of regulation where either data centers can go or what sort of accounting methods that AI records should have for the long run or what sort of things they can or cannot compute.
I think pressures for an overall national strategy I think would be the ultimate outcome when or when it doesn't – when it comes in place, I'm not exactly sure. But I think ahead of the midterm elections for those candidates that are more pro-AI or not, they might run into some headwinds because overall public sentiment in the US is quite negative on AI right now. Between policy implications, use cases, demand drivers, you've covered a lot for us today, Paul, here on Top of the Morning.
So thank you for dropping by. To you, our listeners, again, I do want to point you to the publication Paul Hsiao has been making reference to on today's episode, that is the CEO Macrobriefing Book, Insights on AI. This is a special edition of the ongoing CEO Macrobriefing Book series.
The publication is now available for you up on UBS.com slash CIO. For clients of UBS, please reach out to your UBS financial advisor if you would like to receive a copy of the publication directly. Paul, until next time, thank you again.
Appreciate it. Thank you for having me. Thank you for tuning in.
Be sure to visit UBS.com slash studios to view the entire UBS studios suite of podcast channels along with our video offerings such as UBS Trending. You can also follow us on Instagram for content highlights at UBS Trending. UBS Studios is part of the UBS Chief Investment Office within UBS Global Wealth Management.
Visit UBS.com slash CIO to view the latest research. This material has no regard to the specific investment objectives, financial situation or particular needs of any specific recipient and is published for informational purposes only. For information, please visit our website at UBS.com slash workingwithus.
For a full legal disclaimer applicable to the independent investment views produced by UBS, please visit our website at UBS.com slash CIO dash disclaimer.
Sources & References
How we cover this story