Hyperscalers in forcus
Cloud growth is one key area Mersch sees continuing to drive returns. The so-called ‘hyperscalers’ like Google’s GCP, Amazon’s AWS, and Microsoft’s Azure, have all shown remarkable revenue growth. Where initial growth in these services was driven by organizations undergoing a digital transformation, the recent acceleration of AI workloads has appeared as a new growth area. Microsoft Azure, for example, saw a 30 per cent growth rate last quarter, and around eight per cent of that growth could be directly attributed to AI.
While cloud divisions are “rock solid” in terms of revenue, Mersch notes that they’re also ploughing billions into their capex to improve AI rollouts. Demand for semiconductors, and specifically Nvidia GPUs, is far outstripping supply. CEOs of these mega-caps are speaking more about the risk of underinvestment than the risk of overspending on AI. As these hyperscalers grow to the point where another acquisition will result in anti-trust action, their billions of dollars in free cash flow can go one of three places according to Mersch: share buybacks, dividends, or Nvidia GPUs.
“That does bring us to the next point of contention, which is that we need to see some ‘R’ from the ROI,” Mersch says. “We’re heavy on the ‘I’ with all the capex, but we’re not to heavy on the revenue return aspect.”
Can AI actually generate revenue?
Mersch says that the next fiscal year will be key to determining the amount of revenue tech companies are actually going to generate from their AI investments. He cites a recent Morgan Stanley chart that predicted outcomes for Microsoft’s AI-specific revenue. The base case was $40 billion, the bear case was $10 billion, and the bull case was $90 billion. With such a wide range, it’s difficult to know how the revenue side will play out.
Mersch notes that past tech innovations have been able to justify investor patience, but he says that patience may be required again. In looking for signs of AI revenue, Mersch notes that some companies are changing reporting around their specific business areas. He expects the cloud providers will continue to strip out their AI revenue because investors want to see that data.