GTHT has released a research report stating that AI computing power demand is driving the IDC industry into a new upcycle, and companies that have already completed resource and capacity deployment in advance will fully benefit.
AI training and inference demand continues to grow, and cloud providers are increasing capital expenditure, driving the release of demand for high-power-density data centers. The firm believes that the scale and sustainability of this cycle are expected to exceed the previous two cycles, and companies with resource reserves, delivery capabilities, and cost advantages will benefit more fully.
The main views of Guotai Haitong Securities Co., Ltd. (ASX: 601211) are as follows:
Reviewing the Two Previous Cycles of Cloud Computing and New Infrastructure, This Upcycle Will See Improvements in Both Scale and Sustainability
The cloud computing cycle (2013-2015) and the new infrastructure cycle (2018-2021) were both driven by short-term concentrated construction by CSPs, and after demand was satisfied, the industry shifted into inventory digestion. This round of demand, however, is driven by intelligent computing capital expenditure from major cloud providers, with inference computing power continuously generated as AI applications penetrate. The center of demand has shifted from large-scale training to a dual focus on training and inference, coupled with domestic cloud providers' Capex clearly tilting toward AI, and enterprise AI applications penetrating traditional industries further broadening demand sources.
On the supply side, high-power racks require synchronized adaptation of power supply, cooling, and interconnection, and existing low-density data centers cannot directly accommodate them. Old capacity digestion and new capacity shortages coexist, extending the duration of this cycle.
Price Recovery May Lag Behind Demand Recovery, Power Cost Advantages Support Profitability
Increased customer concentration strengthens buyer bargaining power, and combined with relatively loose supply in the earlier period, IDC prices are unlikely to rebound quickly and substantially in the short term. Profit improvement in this cycle relies more on the matching of orders, delivery, and rack-up rates, with the key to profitability being actual billed capacity and billing time.
IDC has a relatively high proportion of fixed costs, while variable costs are mainly electricity expenses, and PUE requirements continue to rise. Low electricity price regions such as Ulanqab and Qingyang in Gansu have prominent cost advantages, and the coordination of source-grid-load-storage with green power is expected to further lower electricity costs. IDC companies with high rack-up rates still have ample profit space.
Early-Stage Resource and Capacity Reserves Determine the Ability to Capture This Round of Demand, Q4 Valuation Switch Provides a Window for Positioning
Land, energy consumption, and external power infrastructure are difficult to supplement in the short term. Companies that completed resource reserves and main construction ahead of schedule at the bottom of the cycle can quickly advance electromechanical configuration after orders land, better matching customer delivery needs. As prefabricated construction shortens delivery cycles, early-stage deployment is expected to accelerate conversion into billed capacity and operating income.
IDC is a heavy-asset industry, and valuation can focus on EV/EBITDA, combined with project delivery, rack-up rates, and actual billing months to judge profit contribution. In Q4, market pricing gradually shifts toward the following year, and the delivery of projects under construction and rack-up of existing cabinets are expected to drive EBITDA growth, with performance realization and valuation switching jointly enhancing the sector's allocation value.
Risk warnings: 1) Downstream demand falls short of expectations, intensifying price competition; 2) Delivery billing and cost recovery efficiency fall short of expectations; 3) Capital衔接 and transaction integration benefits fall short of expectations.
Comments