Investing in technology and AI: Opportunities in a rapidly evolving landscape

What are the most exciting AI / technological developments you are seeing right now?

The artificial intelligence landscape is evolving from simple tools to complex autonomous systems. Agentic AI models can reason, plan, and act across workflows. Agentic capabilities are moving from concept to deployment. 

Companies are beginning to leverage the technology to automate workflows and conduct research. One of the most popular use cases of agentic AI is ‘vibe coding,’ where developers prompt and orchestrate systems rather than write code line-by-line, fundamentally redefining software creation.

At the same time, AI is moving beyond the screen. Advances in robotics and embodied AI are enabling machines to perceive, reason, and act in the physical world. We anticipate an expansion in use cases as AI models integrate into physical form factors that empower them with ‘eyes’ and ‘ears’.

Intelligence is also getting pushed to the edge on smartphones and wearables. On-device AI enables real-time, personalised experiences without reliance on the cloud. 

Finally, we are excited about the potential for quantum computing, but it remains in its very early stages with limited near-term commercial applications.

Taken together, these trends point to a world of ambient, embedded intelligence across digital and physical environments.

How do you currently view technology stock valuations?

Technology stock valuations are mixed. In general, valuations within the hardware and semiconductor industry groups are high amid heightened investor expectations, while software and IT services stocks look depressed.

For hardware and semiconductor stocks, we see elevated price-to-earnings ratios compared to recent history, but still well below the bubble peaks of 1999–2001. While the multiple expansion can be justified in part by structurally higher margins, it also reflects expectations for future revenue growth that could be interrupted by a digestion period for capex. 

In some areas like hard disk drives and memory chips, current earnings are likely to prove unsustainable as a large part of the growth has been driven by price increases that will reverse as supply and demand become more balanced.

Software valuations are near 10-year lows based on the enterprise value to next-12-month revenue ratio, which at 3.5 times is down significantly from the COVID-19 peak of over 14 times.  This extreme move reflects concerns about the impact of agentic AI models and coding tools on traditional software companies.

For some stocks, the concerns are justified, as some application software platforms may be disintermediated and the overall competitive intensity may increase for the industry. However, we believe that some incumbents will prove to be winners, especially those who provide deeply embedded, mission-critical applications that serve as a system of record, where domain knowledge and business expertise create defensible moats.

As capex continues to grow, what does this mean for the hardware and semiconductor sectors?

The increasing capital spending outlook of the cloud service providers and AI leaders is the primary catalyst for revenue growth and earnings revisions for a broadening group of semiconductor and technology hardware stocks. 

When ChatGPT 3.5 debuted in November 2022, a majority of the initial spend was focused on accelerator chips and the AI servers that house them. Over the past year or two, the leadership among semiconductor and technology hardware names has diversified as market participants look for the next area(s) of infrastructure investment to support AI training and inferencing applications.   

For example, optical networking components and systems are being used to enable scale up (connecting more GPUs – graphics processing units - at lower latency within a server rack), scale out (connecting server racks), and scale across (enabling training of AI models across data centres separated by longer distances) of AI supercomputers. 

And with the rise of agentic AI, inference workflows are changing in ways that require more CPUs (central processing units) for tasks like workflow orchestration and data management.  The ratio of CPUs to GPUs is improving from 1:8 in training, to 1:1 in inferencing applications.

Key risks include the potential for one or more digestion periods during the AI infrastructure build-out phase, which could result in lower unit demand for semiconductors and hardware, as well as lower pricing for commodity items like hard disk drives and DRAM (memory chips) where pricing has spiked due to lack of supply. 

What are the possible catalysts that will help drive even greater AI adoption?

First, multi-modal models meaningfully expand the addressable use cases by incorporating formats like images, audio, and video. We expect AI adoption to grow as users interact with models across these new content formats which enable richer, real-world workflows rather than narrow, text-based tasks.

Second, enterprises are beginning to adopt AI though it remains early. There are certain areas where enterprises’ return-on-investment appears strong, including customer service and coding. AI enterprise adoption is poised to accelerate as companies gain comfort around data governance, establish the proper security measures, and build ‘AI-ready’ data foundations. 

Third, inference costs are declining rapidly, lowering the marginal cost of intelligence and enabling more frequent usage. We expect inference costs to continue to decline due to ongoing advancements in hardware and software. 

Finally, model quality is advancing quickly. Each generation is demonstrating stronger reasoning and higher performance on key benchmark tests, expanding the set of tasks that AI can reliably handle.

A virtuous cycle of better performance, lower cost, and clearer ROI should support AI adoption over time. 

How do you see AI and technology's long-term investment, and economic, potential?

From an investment perspective, we see compelling stock-specific opportunities across the technology stack, including in semiconductors, hardware, and software, as well as in adjacent industries supporting the build-out of data centres.

The major cloud service providers are spending hundreds of billions of dollars to provide the compute and related services necessary for model training and inferencing. It remains early but recent trends suggest they can earn adequate returns above their cost of capital. Backlogs are surging, cloud revenue growth is accelerating, and margins are holding at attractive levels. 

In the near-term, the massive investments in infrastructure to enable AI is supporting areas of the industrial economy tied to data centres and energy. Though we are mindful of the potential displacement of certain jobs in the long run, we generally believe AI will be a boon to the economy as it improves productivity by automating workflows and tasks.