Siemens, the German industrial group, said on 7 August that it will invest more than $200 million in two factories in the United States building electrical equipment for data centres, adding more than 1,500 jobs. The company took around €6 billion of data-centre orders in nine months, which is the demand the plants are being built to serve.1
The equipment in question is switchgear, transformers and power distribution hardware, the machinery that turns grid supply into something a rack of accelerators can consume. A data centre cannot open without it, and the lead times on heavy electrical gear now run long enough that operators order the power infrastructure before the building exists. That bottleneck is why an industrial manufacturer, rather than a model developer, is the one hiring at scale.
Set against every other employment figure in this briefing, the Siemens number stands out for being firm, forward and countable, in the way IBM's July pledge to triple entry-level hiring did. Central banks are inferring job losses from registers and indices, employers are naming AI in announcements that no one can audit, and the hyperscalers publish headcounts on definitions that do not match each other. Siemens has committed to a number of jobs at named sites with capital already allocated.
Fifteen hundred manufacturing jobs will not offset the losses Korea counted in its pension records, and the two groups of workers share no skills, no sector and in most cases no country. An economy can gain electrical technicians and lose junior programmers at the same time, and no policy instrument currently in use moves anybody from the second group into the first.
