Global Switch alumni launch Astor to sell grid queue position

Elliot Dittes, until recently chief executive of Global Switch, has launched Astor with a group of former Global Switch colleagues and the backing of Margaux Platforms, a London-based private-markets investor. The vehicle is both a company and a fund. It will develop powered land and multi-tenant edge colocation, metropolitan sites under 100MW that are offered to compute providers as shovel-ready and pre-equipped with flexible reference designs across modern workloads. Dittes takes the chief executive role.

The target geography is developed OECD countries, and the argument leans on Europe. The announcement cites cloud penetration, aging grid infrastructure and lengthening power connection queues as pressures on the continent's digital-sovereignty ambitions, and claims Europe runs three to five years behind the US in AI and cloud penetration. Dittes describes the demand in similar terms — infrastructure delivered quickly as access to grid power tightens — which is a design constraint being sold as an opportunity.

The size cap is the tell. Staying under 100MW keeps Astor out of the hyperscale campus contest, where the fight runs over gigawatts and years of position in the interconnection queue, and inside metro colocation, where capacity can sometimes be drawn from existing distribution. As this publication has argued, power rights now behave as a distinct asset class — the queue, the permit and the connection all price before the electron does. For a compute provider the product is a site that arrives with power and a fit-out plan rather than a parcel and a promise; for Astor, it is a land-and-design arbitrage on interconnection position, and the last-mile language is how that gets marketed.

Consent is the counterweight. A metro site trades an interconnection queue for a planning queue, and a portfolio of sub-100MW facilities implies many approvals in dense jurisdictions rather than one — so consent, not capital, decides what gets built. Astor has chosen the markets where that constraint binds hardest.

There is no fund size, no capital target, no first site, no named tenant; Margaux is described only as the backer. Astor will therefore underwrite its own leasing plan, taking on the risk familiar from merchant data-center debt: the forecast rather than the customer sets the price, except here the equity takes that side of the trade before a customer is named.

A first site and a named tenant would settle the question; until either appears, the pitch is doing the work of the portfolio.