Goldman Sachs says AI data centers are short on workers, with hiring focused on power, cooling and commissioning rather than AI engineers

Goldman Sachs’ latest AI Adoption Tracker highlights a split in the labor market: AI is estimated to be eliminating about 11,000 U.S. jobs per month on a net basis, while data center construction has added 212,000 jobs since 2022 and is still creating roughly 9,000 roles each month. The openings are concentrated not in machine learning engineering, but in electrical contracting, HVAC, commissioning and critical facility operations. The report reviewed here breaks the market into two very different tracks: construction-phase jobs, which tend to pay well but are tied to projects, and operations-phase jobs, which are more permanent but often involve fixed-site shift work. It lists six major roles, from entry-level data center technicians to critical facilities engineers, commissioning engineers, electricians, HVAC technicians and operations managers, with U.S. 2026 pay ranges running from about $67,900 at the entry end to more than $215,000 for senior commissioning work on hyperscale sites. It also outlines three ways in: paid apprenticeships, certifications and lateral moves from adjacent fields such as electrical work, HVAC, security systems, military communications and IT operations. For Taiwan, the article points to hiring by Chunghwa Telecom, Delta Electronics and Google, as well as CDCP certification courses offered locally in Chinese. At the same time, Goldman’s figures suggest the biggest risk is not AI replacing these roles in the short term, but the temporary nature of construction work: about 4.7 million peak-period construction jobs versus about 697,000 permanent operations roles.