Three hundred jobs

This one happened in my own country. In October 2025, Google announced a €5 billion investment in AI infrastructure in Belgium: an expansion of its data-centre campus in Saint-Ghislain, in Hainaut, over two years. The announcement was made during Belgium's economic mission to the United States, in the presence of Princess Astrid, and the Belgian headlines, including the one in my own newspaper, carried the same second clause: three hundred new jobs. The framing was a triumph. A win for Belgium, a win for employment, a win for the future.

Divide the two numbers in the headline by each other. Five billion euros for three hundred full-time jobs is roughly €16.7 million of capital per job. The site's own history, as VRT reports it, says this is no anomaly: Google has put more than €11 billion into Saint-Ghislain since 2007, and the campus employs about six hundred people. Eleven billion, six hundred jobs, eighteen years. For comparison, the point needs no comparison: the number is its own argument. This is what frontier capital formation looks like when it lands in a small country, and it is the opposite of a jobs programme. It is a machine for turning electricity into computation, owned elsewhere, staffed by almost no one, celebrated by the people it will not employ.

The claim on the table

The frontier-ledger claim (chapters 2 and 6) says the AI build-out is history's largest private capital-formation event, that its gains are banked as assets owned by the frontier and that no claim on those assets is reserved for anyone else. Saint-Ghislain is the claim with a Belgian postcode. The asset sits on Belgian soil, draws Belgian power and enjoys a royal photograph; the asset itself belongs to Alphabet's shareholders. What Belgium receives is a payroll: three hundred salaries, plus construction work while it lasts. What Belgium does not receive is any share in what the machines earn, and the machines are the entire point of the €5 billion. The announcement even carries the book's consolation clause in miniature: Google will fund non-profits to give free AI training to low-skilled workers, which is to say, training for the people whose tasks this class of infrastructure exists to absorb.

The honest complication

None of this makes the investment bad for Belgium. Three hundred data-centre jobs are good jobs. The power-purchase agreements attached to the deal support new onshore wind farms, which Belgium's grid genuinely needs. Foreign direct investment on this scale is a vote of confidence a small economy cannot shrug at, and data centres are capital-intensive by nature: nobody, anywhere, builds one for the headcount. Google did nothing dishonest here. The wrong signal is not the investment. It is the framing that both the company and the government chose, and the press dutifully carried: measuring a capital event by its job count, the one metric on which it is almost irrelevant. That is the Horsey Horseless in press-release form, a machine dressed in the reassuring shape of the thing it replaces.

Verdict

Confirms. The ledger claim predicted exactly this shape: capital lands, the asset stays foreign-owned, the local share arrives as wages and goodwill rather than ownership and the institutions applaud the wrong line of the announcement. A country that wanted the right signal would ask the question the press conference skipped: not how many jobs the machines create, but who holds a claim on what the machines will earn standing on Belgian ground, drawing Belgian wind. Three hundred people will work in the building.

Who owns the machine?