Last week, stocks of major neoclouds, which are specialized providers of AI cloud computing infrastructure, faced significant pressure as investors raised concerns about their growth models in light of prolonged high interest rates. This uncertainty contributed to a 4.5% decline in the Nasdaq, marking its most substantial weekly loss since June. Additional concerns over competitive pressures from China and revenue growth at leading AI research labs further weakened investor confidence.
Read More
Despite these challenges for neoclouds, the overall AI investment landscape remains strong. While financing and execution risks are increasing, recent positive results in the semiconductor sector and rising GPU rental prices indicate sustained demand for AI computing resources. This suggests that the broader narrative surrounding AI growth is still intact.
Neoclouds operate under a different growth model compared to hyperscalers. Neoclouds, which focus on renting GPU compute capacity, typically require substantial upfront investments before generating cash flow from contracted revenues. In contrast, hyperscalers are diversifying their revenue streams with general-purpose cloud platforms and are increasingly dependent on debt markets for their AI investments. The cost of debt for hyperscalers remains manageable at approximately 6.5-7%, against a backdrop of expected returns on new data center investments of 25% or more.
The impact on overall AI-related expenditures is expected to be minimal, as compute supply remains constrained. Projections estimate that AI capital expenditures will grow by 33% to USD 1.2 trillion by 2027, with neoclouds comprising less than 10% of this investment. Although rising funding costs could pose risks for neoclouds, many of their customers demonstrate solid credit quality, and advance orders may provide some buffer. Moreover, GPU rental prices continue to escalate, reflecting strong demand amidst supply limitations. Recent data shows a 7% increase in rental rates for four-year-old GPUs in the last month, contributing to a nearly 30% year-to-date surge, and a 35% increase for two-year-old GPUs thus far this year.
Positive earnings reported by ASML and TSMC last week bolster confidence in AI fundamentals. TSMC indicated robust demand and anticipates that its capital expenditures for the period 2026-2028 will be significantly higher than those for 2024-2026. Similarly, ASML noted strong order inflow for 2028, extending projections into 2030. While this report does not evaluate individual companies, it reinforces the notion that demand for AI is both strong and continuing to rise. Upcoming results from hyperscalers are also expected to report accelerating growth in cloud revenue.
Thus, a positive outlook remains for the structural growth of AI, while recognizing that funding and execution risks could exacerbate across the AI infrastructure ecosystem. Tighter credit conditions may limit deployment, and potential delays in delivery might hinder the cash flow needed to support debt obligations.
For investors, this environment necessitates a selective approach. Focus will be shifted away from neoclouds due to funding and construction pressures, favoring instead well-funded hyperscalers that present a more favorable risk-reward profile. Additionally, high-quality semiconductor and hardware companies benefiting from AI spending are recommended, along with cautious investment in defensive technology sectors, including payment networks, select consumer electronics, and certain software firms.