Talking Heads – Artificial Intelligence - Everything, Everywhere, All at once

Podcast with Derek Glynn on Disruptive Tech

Daniel Morris: Hello and welcome to the BNP Paribas Asset Management Talking Heads podcast. Every week, Talking Heads will bring you in-depth insights and analysis on the topics that really matter to investors. In this episode, we’ll be discussing artificial intelligence. I’m Daniel Morris, Chief Market Strategist, and I’m joined today by Derek Glenn, Portfolio Manager. Welcome, Derek, and thanks for joining me.

Derek Glynn: Thanks for having me. It’s great to be here.

DM: If we think about the evolutions in technology over the last several decades, the sector becoming quite prominent in investors’ minds with the arrival of the Internet. And then you had the big boost to demand with COVID, and we get AI. There’s been another boost for the sector with the tariffs from the Trump administration. Other parts of the market which are more goods oriented suffered relatively more from the tariffs than tech has. So, a lot of tailwinds. A key development earlier this year was the release of DeepSeek by a Chinese AI lab. Could you tell us more about that and some of the longer-term implications?

DG: I’ll just start with a definition. Generative AI is a technology that can generate content from a text-based or image-based prompt. With more and more data and the right type of training, it’s been able to reach impressive levels of advanced reasoning. It’s unlocked a lot of interesting use cases as it relates to DeepSeek. They developed a reasoning model that can think and review its work before it answers a question. What was unique was how efficient they apparently were in training the model and how cheap it was for users to query or ask the model. The cost of using Deep Seek’s model was priced at just a fraction of a leading comparable model in the US.

The most important takeaway for me was that DeepSeek put a spotlight on this trend of AI models becoming cheaper and cheaper to train and use. Prices today for querying models are roughly one 100th on average of what they were a few years ago. As we know from economics, as prices fall, consumption tends to rise. Our belief is that AI will be everywhere. It will be embedded in many systems, applications and technologies. It will touch every part of our economy. And more efficient models unlock this opportunity.

DM: Aside from the lower costs that you talked about, are there other factors that lead you to believe that AI adoption rates will increase significantly in the coming years?

DG: AI adoption is poised to accelerate because we’re at the intersection of three important trends. The first is this continued decline in the cost of processing queries. The second is continued improvement in model intelligence. Most leading models now exceed a human baseline level of intelligence across several benchmark tests, including math, visual reasoning, image classification, PhD level, science questions and more. At the same time, error rates are declining.

Fundamentally, the combination of those two reinforces this third dynamic, which is a broadening in the number and type of use cases for AI. It’s important to keep in mind this is a general-purpose technology: it’s going to impact all aspects of the economy, and it can also improve many areas of our personal lives. AI use cases like education and technical assistance are well understood. But a unique use case is using AI for therapy or companionship. Many people don’t have access to doctors or specialists. There may be no appointments available or it’s too cost-prohibitive. AI is likely not a perfect replacement, but in some instances, it could be better than nothing. it is improving the quality of life for many. That’s an important example because it highlights this variety of use cases as well as the reach and general applicability of AI.

To summarise those points, it’s a combination of cheaper, smarter and general purpose that implies adoption rates should increase over time. To put some data around this, McKinsey and company did a survey last year where they found 78% of enterprises have adopted AI in one or more functions or departments. That’s up from 55% a couple of years ago. However, only 16% of enterprises have adopted it in five or more departments. This data suggests it’s still early. Enterprises are beginning to experiment with AI, but it still hasn’t spread throughout the entire organisation. Looking ahead in the long run, I’m most excited about AI and different physical form factors like robotics or augmented reality glasses. There will be far more use cases and even greater adoption rates when these AI systems have physical capabilities and access to real-time information about the world around us.

DM: What you just talked about really highlights we really haven’t seen anything yet. Now, at the same time, a key debate among investors is around the magnitude of spending, in particular by the major cloud service providers. How do you see that trend developing in the years ahead?

DG: The advance made by DeepSeek caused some to question the pace and trajectory of capital expenditures because DeepSeek was very efficient in their model training: they essentially did more with less. Our view is that capital expenditures for the major cloud service providers and a leading social media company will continue to grow. It may come in waves of steep investment followed by periods of digestion, but overall, the trend should be higher. For 2025, we’re estimating more than 40% year-over-year growth to almost 290 billion US dollars. Most of this spend is on building out datacentres and equipping those facilities with servers and graphics processing units.

It’s more challenging to predict the pace of investments, but our thinking is we’ll continue to see year-over-year growth, albeit probably at a slower rate than the 40% that we expect for 2025. Big picture, there’s still an AI arms race unfolding between these megacap technology companies. AI is a very large and still emerging opportunity and it’s likely there’s greater risk to them missing it than in overinvesting. So, we think they’re going to pursue this type of spend to secure their future.

There’s also debate about the extent to which the cloud service providers can earn an acceptable return on this investment. Our view has been fairly consistent in that we think they are well positioned. I don’t believe they’re overspending, and they have many ways for them to monetise these investments. I like to say all roads in AI lead to the big three cloud service providers. They provide the storage, the compute, and they have the most widely available access to the graphics processing units. For companies to leverage AI, they’re also developing applications that leverage the technology.

And finally, they’re beginning to use AI internally to drive efficiencies. Software developers are getting much more productive, sometimes by up to 30 to 40%, because AI is helping them autocomplete code as they write it. It’s also important to keep in mind these companies tend to have strong core businesses. They generate a lot of free cash flow, so they can fund these capital expenditures. Those core businesses are leveraging AI as well. There’s many ways for them to monetise the technology as relates to the hardware side of the equation. Semiconductor and semiconductor equipment companies remain well positioned. They’re fundamentally enabling this technology by providing the chips, the equipment, the networking, the hardware that are necessary in the training and use of AI.

That trend’s likely well understood and we are cognizant of potential short-term risks related to tariffs and export controls. We’re in a market environment that requires careful navigation of these risk factors. Overall, AI is a secular growth opportunity, and it should propel stronger financial trends in the years ahead for many players.

DM: If I could just pick up on two points. You see AI as a growth opportunity. If you look at earnings expectations for the industry over the last year, it’s been on a pretty steady trend upwards in contrast to a lot of other sectors. Another point to make, when you mention the capex, is just how important that is for the US economy. If you look at growth in the first quarter, there was notable weakness in consumer demand, but that was offset to a significant degree by business investment. A big chunk of that business investment was taking place in technology sectors and in software. Last question then, Derek, are there other possible beneficiaries of generative AI that might be underappreciated by the market?

DG: High-quality businesses with proprietary data and few competitors are well positioned to benefit from generative AI and that’s underappreciated by investors. These types of businesses can be found in any sector, but they’re particularly prevalent in the information solutions industry. Companies can unlock the full potential of that data with AI by developing new products, like some predictive analytics tool, for example, and that could drive topline revenue.

The second way these companies can benefit is from increasing productivity internally or reducing costs. AI is great at automating workflows, for instance, in areas like customer service, sales and application development. Businesses with few competitors can choose to flow through those AI savings to the bottom line. It is an opportunity for them to expand margins. And finally, these possible beneficiaries, they signal the potential for [a] broadening-out of winners in the market. It won’t just be the large-cap semi and technology companies. We expect there to be many winners from generative AI, including those that reside outside of the IT sector.

DM: If I can summarise some of the key points, if we go back to that DeepSeek announcement, the key takeaway was the ability to develop AI technologies at a much lower cost, leading ultimately to an increased use of AI. We’re also seeing a continued improvement in the model intelligence in some areas exceeding a human baseline. If we think about what the future might hold, you anticipated continued significant capital expenditure as companies pursue the development of these models. Well, Derek, thank you very much for joining me.

DG: Thanks so much for having me.

DM: That’s it for this week’s episode of Talking Heads.
If you would like more information about our capabilities in technology investing, please reach out to your BNP Paribas Asset Management contact or check out Viewpoint, our website for investment insights at viewpoint.bnpparibas-am.com. Viewpoint brings commentary and analysis in a variety of formats, from investment outlooks to asset allocation videos and podcasts, to help investors make better informed decisions. You’ve been listening to the BNP Paribas Asset Management Talking Heads podcast with me, Daniel Morris, and Derek Lynn, Portfolio Manager. Please do join me next week. Until then, take care.