As the gap between computing power supply and demand continues to widen, any slowdown in data center construction will create sharply divergent winners and losers across the entire AI supply chain.
Apollo Global Management Chief Economist Torsten Slok said in a recent analysis that computing demand currently far exceeds supply, making the pace of construction critically important: if construction timelines are delayed, assets that already control scarce physical capacity 鈥?including operating data centers, power suppliers and long-cycle equipment manufacturers 鈥?will gain pricing power and rent premiums; while high-valuation AI stocks priced on on-time delivery, as well as highly leveraged developer debt, will face the dual pressure of earnings downgrades and valuation compression.
Holders of scarce capacity reap the rewards
Slok argues that under a construction delay scenario, the most direct beneficiaries are the various parties that already own operating capacity.
Data centers that are already online can raise rents amid tight supply and fill every megawatt of capacity, because new competitors cannot enter quickly. Hyperscale cloud service providers that locked in power, land and chip resources early can also continue serving customers while rivals are stuck waiting, maintaining their competitive advantage.
Holders of land and grid interconnection resources are in a particularly favorable position. Sites that have already secured power guarantees, grid interconnection qualifications and related permits are the scarcest and hardest-to-replicate production factors in the entire supply chain, and their value will rise significantly as supply tightens.
Long-cycle equipment suppliers also benefit. Manufacturers of transformers, switchgear, generators and cooling systems already have large order backlogs and strong pricing power, with transformer delivery lead times reaching as long as 26 months. Where the grid becomes a bottleneck, the value of gas turbines, fuel cells and energy storage equipment will also climb.
High-valuation AI assets face earnings revisions
Slok warns that delays pose a direct drag on hyperscale cloud service providers' growth and profitability.
When capacity is constrained by supply rather than demand, the revenue growth ceiling is pushed lower, while already-committed capital expenditures continue to accumulate 鈥?even if production capacity sits idle temporarily. Pricing power on existing capacity can only partially offset this impact, and earnings growth will slow compared with original plans.
AI labs, startups and enterprise users will bear a more direct blow 鈥?computing rationing tightens, prices rise, and model training and product launch cycles lengthen, with smaller players being squeezed out of the market first.
Developers and investors in the construction phase are also harmed. The time to generate cash flow is pushed back, while interest and holding costs continue to accrue, reducing overall returns. Developers relying on leveraged financing or pre-leasing contracts will also face debt covenant triggers and refinancing pressure.
For equity investors, AI stocks priced on the premise of on-time delivery will be directly exposed to the dual risks of earnings downgrades and valuation multiple compression.
Chipmakers face timing mismatches, and macro effects cannot be ignored
The situation for chip manufacturers is relatively complex.
Chip demand itself remains strong, but if data center sites cannot be powered up and put into operation in time, chip deliveries may run ahead of actual installation progress, delaying the timing of revenue recognition and creating a timing mismatch risk rather than a shrinkage in demand itself.
From a macro perspective, the impact of construction delays should not be underestimated either. Torsten Slok pointed out that recent U.S. GDP growth has been highly dependent on AI investment, and a construction slowdown will depress short-term economic growth and postpone the realization of AI-driven productivity gains. In addition, residential and business users sharing the grid with data centers may also face pressure from rising electricity costs.
Delay versus cancellation: Investors need to clarify the key difference
Torsten Slok specifically emphasized that investors need to distinguish between two very different scenarios: "delay" and "cancellation." A delay lowers returns by postponing cash flow, but the asset itself still exists and may even appreciate due to a scarcity premium. The real risk lies in the breakdown of the scarcity logic 鈥?for example, a concentrated release of supply or a substantive softening of demand, at which point the premium on existing capacity would be eroded.
In terms of investment strategy implications, a delay scenario favors assets tied to scarce physical capacity, including operating data centers, power suppliers and long-cycle equipment manufacturers; it is unfavorable for assets priced on the basis of on-time delivery, especially high-valuation AI concept stocks and highly leveraged developer debt. The broader market will bear indirect transmission pressure through the weight of AI names in major indexes, as well as through overall slower growth.
Comments