AI’s Spending Boom: Opportunity or Overreach?

Back in late 2023, we were quite cautious about the AI trade. Press headlines and broker reports radiated excitement about generative AI and propelled Nvidia and other AI infrastructure names into the spotlight.

However, we questioned whether the capex spend behind it was sustainable as the build out was financed by venture capital and experimentation. We were sceptical about the pace of adoption and the arrival of new and valuable use cases, as well as the underlying business model and monetisation.

When the facts change…

As this cycle progressed, our view changed. In early 2025, we saw the potentially disruptive threat of generative AI changing from a well-funded science experiment into a strategic arms race for the large cap tech incumbents. They are being forced into making ever-bigger investments into data centres and AI infrastructure to keep up with OpenAI.

Amidst market volatility in 1H25, we invested in high quality AI infrastructure names such as Broadcom, Nvidia, Arista Networks, Amphenol, TE Connectivity and Vertiv. AI infrastructure names now make up around 30% of the Platinum International Technology Fund.

Today, we think the market is likely underestimating how this investment cycle could evolve into a multi-year capex binge – but a binge funded by some of the most profitable businesses in the world.

As this quarter’s Platinum International Technology Fund commentary highlights, growing ChatGPT user engagement, expansion into advertising and OpenAI’s ability to raise progressively larger sums of capital will likely force Alphabet and Meta to keep investing.

As we discussed in our July article about Apple, we also think Amazon and Microsoft could be at risk. We see ample capacity to fund this capex binge as the five big tech companies combined have up to half a trillion US dollars in free cash flow that can be spent.

Searching for the next top model

We still have questions about where all of this is heading long term. Generative AI companies like OpenAI are still searching for a sustainable business model. Competition among models is fierce and performance advantages are typically nullified within 3 to 6 months, rendering the return on high upfront costs to train models to effectively zero.

Hyperscalers and neo clouds that are buying GPUs and renting them out to loss-making AI companies make very low returns on capital as the majority of the economics are captured by Nvidia through higher prices. The entire AI ecosystem is thus reliant on external funding (driven by hype and optimism) and by the belief that the free cash flows of big tech companies can continue to grow.

We remain both vigilant and open minded.