In this analysis from Tim Humphreys, co-manager of the IFSL Marlborough Global Essential Infrastructure Fund, he highlights opportunities for diversification as well as growth, by investing in the relatively ‘volatility-lite’ arena of infrastructure
“There’s been a lot of talk about an AI bubble. From our vantage point, we see something very different.”
These were Nvidia CEO Jensen Huang’s opening remarks during the company’s final earnings call of 2025. To paraphrase Winston Churchill: never in the field of revenue reporting have so many derived so much relief from so few words.
Fears over stretched valuations in the arena of artificial intelligence had mounted in the days prior to Huang’s big reveal. Even the International Monetary Fund joined the growing chorus of concern. [1] It was generally assumed a disappointing set of figures could spark a spectacular sell-off.
By announcing the world’s largest business had once again beaten expectations, Huang restored a measure of calm. Even so, investors continue to balance excitement about the potential of AI with warnings about a bubble.
None of this is to suggest there isn’t a case for investing in Nvidia and AI’s other poster children. After all, a company whose market capitalisation exceeds that of whole indices ought to have a certain appeal.
If there’s one lesson we might take from that eagerly awaited call, though, it’s that backing mega-cap AI stocks requires confidence that these technology titans can continue to satisfy what are undoubtedly high hopes.
Given that AI could transform almost every aspect of day-to-day life, it’s hard to imagine many investors having no interest whatsoever in a potentially epoch-defining transformation. But can they tap into it without engaging in the collective nail-biting likely to precede Huang’s next pronouncement?
In my view, they can. The trick lies in recognising that the long-term success of AI doesn’t depend only on a handful of tech titans. This brings us to the opportunities in the sphere of infrastructure.
Architects versus enablers
The era of “hyperscaling” is already under way. It entails creating the far-reaching infrastructure necessary to deal with ever-bigger datasets – and, by extension, ever-bigger workloads – as AI continues to improve.
The capital expenditure sums now routinely bandied about would have once been considered mind-boggling. They include the $1.4 trillion that ChatGPT developer OpenAI plans to spend on data centres over the next eight years. [2]
Yet it’s not the job of OpenAI and its peers to bulldoze sites, erect buildings and install equipment. The likes of Amazon, Google, Meta and Microsoft aren’t going to take it upon themselves to hook up power supplies, fit drains, turn on the taps and so on.
Such tasks – all of them essential to the hyperscaling effort but frequently overlooked by investors – instead fall to a sizeable cast of stakeholders. These include construction firms, network operators, energy suppliers and utility companies.
It can therefore be helpful to think of the wider AI landscape in terms of a distinction between architects and enablers. The tech giants are in the former camp, while the many businesses they need to turn their vision into a reality are in the latter.
This translates into a classic “picks and shovels” scenario. Just as tool-selling merchants facilitated the gold rushes of the 19th century, infrastructure providers have to make the AI revolution happen.
On the whole, of course, it was those merchants who prospered most. Dishing out picks and shovels was a pretty steady way to earn a living, whereas hacking away at unyielding rock faces in the hope of finding a nugget or two was inherently hit-and-miss.
Similarly, the dynamics that tend to influence earnings from infrastructure are unlike from those that that tend to shape earnings from AI. They might be described as rather less mercurial – which is to say they’re relatively volatility-lite.
A matter of exposure
We can demonstrate this by comparing the recent volatility of the holdings in our own fund with the recent volatility of the tech-heavy IAIQ Index. You may agree that the results are striking.
Over a three-year period, according to our analysis, the highest correlation for an individual stock was for Constellation, a US electricity provider, at 1.04. The company has already signed numerous deals with hyperscalers.
Overall, however, there was a correlation of just 0.3. In other words, there was virtually no overlap between our portfolio and the index – even though we hold plenty of businesses that are central to AI’s ongoing rise. How so?
As an infrastructure-focused strategy, our fund is invested in companies whose earnings are normally determined by long-term contracts, regulated returns and physical service demand. These drivers are quite different from those that dictate the earnings of, say, Nvidia.
In fact, we can prove that very point. The volatility correlation between our holdings and Nvidia during the same three-year period was only 0.03.
All of this indicates it’s eminently possible to invest in AI without being overly exposed to the frequent ups and downs that afflict the best-known names. There remains scope for AI-linked upside, yet there’s also scope for resilience during risk-off periods or rotations.
Granted, there may still be some investors who are thoroughly unmoved when, say, a utility benefiting from hyperscaling gently raises its earnings-per-share guidance. They might instead prefer to stick with the seemingly boundless growth associated with the usual suspects.
For those who would like to make it through to Nvidia’s next earnings call without gnawing their nails away, though, there could be much to be said for looking beyond the obvious.
Tim Humphreys is co-manager of the IFSL Marlborough Global Essential Infrastructure Fund.
[1] See, for example, Reuters: “AI investment boom may lead to bust but not likely systemic crisis, IMF chief economist says”, October 14 2025 – https://www.reuters.com/legal/transactional/ai-investment-boom-may-lead-bust-not-likely-systemic-crisis-imf-chief-economist-2025-10-14/.
[2] See, for example, NPR: “Here’s why concerns about an AI bubble are bigger than ever”, November 23 2025 – https://www.npr.org/2025/11/23/nx-s1-5615410/ai-bubble-nvidia-openai-revenue-bust-data-centers.
