The AI spending race accelerates

With so much attention focused on the SpaceX Initial Public Offering (IPO), some other events in capital markets got less attention than they might have done. SpaceX has been the largest IPO on record, but it wasn’t actually the largest equity raise in the public market in the second quarter. That award went to Alphabet, which raised around $85 billion in June (although $40 billion of that can be sold over time). 

Among unlisted companies, we’ve seen Anthropic and OpenAI raise significant new money. At the same time, large tech businesses have tapped the bond markets – notably Nvidia’s recent $25 billion debt raise and even larger bond issuance from Amazon and Alphabet earlier this year. Large tech businesses are spending aggressively on their Artificial Intelligence (AI) build-out, and investors, for now, are happy to fund them.

We can see the impact of this playing out in all sorts of metrics. The chart below, for instance, shows capital investment by the US tech sector as a percentage of their total sales and their cash generation. These figures will likely rise further from here. Estimates for capital spending by so-called hyperscalers in 2026 have topped $1 trillion. 

We can see the impact of all this in macro data. The chart below shows spending on data centres as a percentage of GDP. We’ve seen a steady rise in over the past fifty years and a sharp increase more recently, taking the proportion of spending above the recent peak from the Internet boom.

All this represents a significant shift in how investors think about many tech businesses. Over the past decade, many investors have thought about technology as being “asset-light” – meaning that they didn’t spend that much on building factories etc. As the chart above shows, that’s currently not the case. But the chart also shows that this wasn’t always true, at least in aggregate. Tech businesses in the late 1990s and early 2000s spent a good proportion of their sales or cashflow on capital investment. We’ve been trained to think of tech as being “asset-light” but maybe that’s just because we’ve been too focused on the past decade.

Being asset light has worked for tech investors in recent years. The chart below compares the Return on Equity – a broad measure of how much value businesses create – between US tech and non-tech businesses in the US equity market. We can see the spread between them has widened significantly since the Global Financial Crisis.

But with capital investment increasing sharply, we could be entering a period where the returns on these businesses fall. That’s not necessarily certain or even a problem. From a long-term perspective, you could argue that businesses should be investing behind new technologies to avoid being made obsolete and to create new avenues of growth. Certainly, we can see that earnings in technology have expanded dramatically (see the chart below), particularly over the past couple of years. 

So where does this get us? New technologies require an upfront investment in research and development and infrastructure to prove that they work and then scale them up. We think we’re in the early stages of scaling up and we’re seeing tech businesses raise a lot of money to fund that process. It’s impacting the tech supply chain and driving a lot of economic growth, particularly in the US.

There’s often a lot of discussion around the equity valuation for these businesses, but at this point we’d argue that the more important question is around scaling up and the growth of end demand. Will these companies be able to build their capacity on schedule, given the constraints on things like energy? If they do, will there really be demand for all the AI capacity? So far, there’s some cause for optimism we’ve seen significant adoption of Large Language Models (LLMs) across businesses and households. We think that can sustain for now, but we’ll continue to monitor the trends closely.