goldman sachs’ latest research report points out that the market has significantly misjudged the pace of ai infrastructure expansion—far from peaking, the wave of ai‑driven capital spending is both stronger and more prolonged than widely anticipated. according to the firm’s estimates, global hyperscale data center capital expenditures in the ai sector will reach $1.1 trillion by 2027, nearly 20% higher than mainstream wall street expectations. moreover, factoring in accelerating model‑training demand, the rollout of multimodal applications, and the large‑scale deployment of enterprise‑level ai agents, this figure could climb as high as $1.4 trillion under an optimistic scenario.
the core driver behind this upward revision lies in the exponential surge in computing power demand. as ai agents deepen their penetration across critical sectors such as finance, manufacturing, and healthcare, global ai token consumption is projected to increase twenty-fourfold by 2030 compared with current levels. this trend is vigorously propelling an accelerated expansion across the entire upstream infrastructure value chain—from high‑performance computing chips, liquid‑cooled servers, and high‑speed optical interconnects, to supporting grid upgrades and the development of green‑energy supply systems—all now entering a phase of rapid investment. goldman sachs forecasts that the ai computing supply‑demand gap will persist at least through the end of 2027, with robust infrastructure growth likely translating into substantial revenue and profit growth for relevant companies.
however, the report also underscores significant structural constraints and valuation risks that should not be overlooked. on one hand, the actual roi of ai tools remains in the validation stage; memory bandwidth bottlenecks, regional electricity shortages, and a shortage of highly skilled operations and maintenance personnel have already caused delays in numerous large‑scale data center projects. on the other hand, valuations in the ai infrastructure sector are rising rapidly, with many stocks posting gains that far outpace current earnings realization. combined with uncertainties surrounding technological iteration and macroeconomic liquidity fluctuations, these factors could trigger temporary pullbacks. the analysis notes that strong long-term fundamentals do not guarantee short‑term stability; maintaining a disciplined approach and focusing on leading vendors with proven delivery capabilities and cost advantages remains the key to sustainably capturing the benefits of ai infrastructure growth.
