Do German grid attacks force data centers to buy backup power?

On September 1st, attackers planted improvised explosive devices at the Turnow-Preilack substation in Brandenburg, a junction for power from the nearby Jänschwalde coal plant; within roughly a day, detectives were probing a suspected sabotage strike on an Amprion station in the west. Both cut lines that their grid operators restored within hours, and neither came close to a blackout. The events nonetheless re-opened a question that has nagged German industry since arson late last winter left some 50,000 Berlin homes without power for days: what should users of the electricity network now pay to insure against the next attempt? For the country's data centres, which can draw up to two-fifths of Frankfurt's electricity, the question is sharper still. Does elevated outage risk force them to buy more backup power, and does that squeeze margins? The near-term evidence, and the structure of their business, suggest the answer is mostly no.

The market test comes first. IONOS, the German hosting-and-cloud group that is the purest listed exposure, reported a record first half in August: revenue of €701.1m, up 6.9%, and an adjusted EBITDA margin that slipped to 35.0% from 36.1% a year earlier. Strikingly, the group's capital-expenditure ratio nearly doubled, to 7.5% of revenue from 3.5%. That looks like the threatened signal. But management attributed the jump not to hardening their power supplies but to "early procurement of hardware to mitigate rising prices" for equipment — a hedge against AI-driven cost inflation, not against sabotage. No mention of grid risk, of insurance, or of outage costs appeared in the call. Deutsche Telekom's T-Systems, which plies sovereign-cloud and colocation services from Frankfurt, likewise guides on demand and capacity, not on the threat.

That should not surprise. Redundancy is not a new cost that the attacks have suddenly forced upon operators; it is the baseline of their operating model. A colocation that could not promise uninterrupted power would not exist, and power supply is already the industry's biggest source of failures — the cause of 45% of data-centre outages, by the Uptime Institute's survey. Uninterruptible power supply, N+1 generators and fuel are built into the rent, sunk long before September 1st. The two attacks caused no data-centre outage and, as far as is disclosed, no business-interruption claim. Nothing in the operators' recent books records a grid-attack charge, because nothing was incurred.

The genuine cost channel is insurance, and it runs through the industry rather than through any single balance sheet. Global data-centre premiums are expected to roughly double, from close to $10.6bn to some $24.2bn, by 2030, in Swiss Re's telling. That repricing is real, and sabotage in Europe adds to a driver that already includes construction risk, cyber and, above all, power supply. But it is spread across thousands of sites worldwide, carried largely as a tacit rise in the price of scarce capacity. Even at double the premium, data-centre insurance remains a small line in the operating costs of a group like IONOS — visible in pricing, invisible in the EBITDA margin.

Which is the point. The attacks are best read not as a cost shock but as a small reinforcement of the one way they help operators: they confirm that reliable German power is scarce, and scarcity is what gives colocation pricing power. After fifteen years of falling rates, the industry's pricing has turned; the global average rate rose about 17% over the past half-decade, and brokers warn that capacity scarcity will push rents higher. In Frankfurt the constraint is acute: data centres already consume up to 40% of the city's power, and grid operators say no substantial new connections will be available until the mid-2030s, pushing builders to on-site gas generation. In a market where the next megawatt is years away, the handful of operators holding backed-up, grid-connected capacity — including Telekom and IONOS's newer Frankfurt sites — are the ones telling tenants what reliability costs, not the reverse.

The genuine risk sits in the tail, not in the trend. A sustained, multi-day outage at a single campus, of the sort that felled Berlin's households in January, would test redundancy at the limit and could throw up a real business-interruption loss. But that is precisely the event the existing architecture, and the elevated rents reliability already commands, are built to absorb — distributed sites, generators, and tenants who pay for the assurance. Assessed soberly, the case that grid attacks force data centres to buy backup power puts the cart before the horse: they already own it, they already charge for it, and the marginal sale is small. The measure of whether that changes is observational, not rhetorical. Watch the operators' next two quarterly calls for a reportable outage loss, and watch their disclosures on grid-outage insurance; look at whether Frankfurt utilisation and pricing keep rising as supply stays frozen. Until an outage that is actually borne by a data centre appears, the elevated risk is a reason scarcity stays expensive — an operator's friend more than its enemy.