Use cases for AI investments already emerging despite investor capex concerns
Annabelle Miller
Despite investors starting to show some concerns about the share prices of tech companies such as Microsoft, Amazon, Google and Meta, these fears should not be overblown, says Annabelle Miller, principle – investments at ECP Asset Management.
“Earlier this year, investors showed some consternation following the release of DeepSeek and its impact on companies such as Nvidia and others in the “Magnificent Seven”. This coincided with a deceleration of the growth rates in these companies; however we don’t believe it is a sign of problems with the fundamentals of these businesses.
“Rather, what we are seeing is an increase in capital expenditure by these hyper scalers, who need to invest billions of dollars into the hardware used to power AI applications and the GPUs used to power AI computations.”
Ms Miller says the capital intensity measured by the capex/sales ratio of the hyper scalers has increased from roughly 11 per cent in 2020 to around 21 per cent in 2025.
“Across the big five hyperscalers, we’re seeing over US$300 billion being invested this year, rising to over US$500 billion by 2030. The increase in capital reflects the next evolution of technology as these big trillion-dollar companies invest in the hardware necessary to power AI applications,” she says.
Investors should be aware that these AI investments will likely take time to monetise, resulting in short-term margin pressure and impacting earnings in the coming quarters. However, Ms Miller says these investments are necessary to take costs out of the hyper-scalers and improve margins.
“We see this as one of the biggest long-term opportunities for investors. By 2030, we expect a significant amount of margin expansion as these businesses harvest the benefits of the investments made today.
“We are already seeing more use cases for AI within these businesses. Microsoft has saved up to US$500 million in call centre operating costs through AI applications. These applications are assisting with automating tasks like coding, content generation, office productivity and marketing.”
Other beneficiaries of this wave of capital investment include companies like Taiwan Semiconductor Manufacturing Company (TSMC), which has a monopoly on the manufacture of advanced chips for the likes of Nvidia which end up in the hyper scalers data centres, says Ms Miller.
“These advanced chips are manufactured using high-tech proprietary processes involving array of complex chemistry and physics. As these chips become more advanced, TSMC will be able to extract greater pricing power. TSMC has even taken share from its competitors Samsung and Intel, which have been unable to keep up with these advancements.
“This is a fast-moving technology space, so the winners and losers of the AI play can change within a year. In January this year, DeepSeek emerged from nowhere. It was on no one’s radar and all of a sudden there was a new competitor in town.
“We can’t be complacent about emerging winners. However, TSMC has proven that it has invested in numerous amounts of R&D processes to cement its position in the semiconductor supply chain,” says Ms Miller.
