Switzerland data localisation APAC

Switzerland's Three-Tier Cloud Tender Treats Data Localisation as a Risk Dial, Not a Blanket Rule

Switzerland's CHF 7bn Government Cloud tender keeps data in Switzerland only for the tier that needs it, and caps each bidder at one lot.

Swiss Government Cloud Tender People of Internet Research · Switzerland ~CHF 7bn Maximum tender volume Ceiling over the 14-year contract … 3 Lots in the tender Global public, Swiss-resident publ… CHF 246.9m Parliament-approved credit Commitment credit approved in Dece… 60 days Prequalification phase Runs from publication; awards expe… peopleofinternet.com
Swiss Government Cloud Tender People of Internet Research · Switzerland ~CHF 7bn Maximum tender volume 3 Lots in the tender CHF 246.9m Parliament-approv ed credit 60 days Prequalification phase peopleofinternet.com

Key Takeaways

On 20 October 2026, the Federal Office of Information Technology, Systems and Telecommunication (FOITT) published the tender for the Swiss Government Cloud (WTO 25002 609) on SIMAP. The headline number is a ceiling of about CHF 7 billion over 14 years. The more useful detail is the design. Switzerland did not pick one answer to the data-localisation question. It wrote three answers into three lots and matched each to a level of data sensitivity.

What the tender actually does

The federal press release sets out three lots, each with a binding maximum volume:

Each award runs seven years, and the market reopens every seven years, which gives the 14-year horizon. A bidder can win at most one lot. The FOITT says this reduces dependence on individual providers and jurisdictions. The Confederation is the main user, and cantons and communes can also take part. Private-sector use is excluded. Costs arise only for resources actually consumed. The 60-day prequalification phase runs from publication, and awards are expected in late summer 2027, provided no objections with suspensive effect are filed.

The strongest case for localisation

The argument for strict localisation deserves a fair statement. Governments hold data whose exposure would be costly and often irreversible, such as tax records, health data and security files. A provider headquartered abroad may be subject to foreign legal demands that Swiss law cannot override. Keeping data in the country and running it on open-source stacks the state can inspect, and if needed move, reduces that exposure. It also gives the state leverage in negotiations. The FOITT's own description says Level III offers "the highest degree of data sovereignty", and the tender's single-lot rule is a deliberate hedge against lock-in. These concerns are legitimate, and the design takes them seriously.

Why the tiering is the right call

The pro-innovation objection to blanket localisation is about cost and capability. Hyperscale services such as large-scale machine learning, global content delivery and rapid elasticity depend on scale that a national-only market rarely supports. A rule that all government workloads must sit in Swiss-only infrastructure would push agencies toward weaker tools or higher bills. It would also hand a few domestic or qualifying vendors a protected market.

The tender avoids that trap. Lot 1 lets agencies use global services for workloads where sensitivity is low and capability matters. Lot 2 gives a Swiss-resident option with open-source components for the middle band. Lot 3 keeps the most sensitive applications on federal premises. Data classification, not geography, decides where a workload goes. That is proportionate regulation in practice: the control is heaviest where the harm from exposure is greatest.

The FOITT's own programme page describes the aim as a hybrid multi-cloud that lets authorities use internal federal services alongside several external providers. It also lists access to innovation among its objectives, alongside data protection and information security. Those goals pull against each other, and the lot structure is how the tender manages the tension.

Where the risks sit

The design still leaves open questions. The first is cost. The Swiss Federal Audit Office (SFAO), in its first audit of the programme published on 3 December 2025, found that the SGC's economic benefit for the Confederation as a whole "has yet to be demonstrated". It recommended that the programme complete that calculation, which the programme planned for 2026. Parliament approved a commitment credit of CHF 246.9 million in December 2024, and total programme cost is CHF 319.4 million. Cloud usage and migration costs fall on the offices that use the service and are not in that figure. The CHF 7 billion ceiling is a maximum, not a forecast, but a ceiling that large invites scrutiny of actual demand.

The second risk is the single-lot rule. It diversifies suppliers, but it also forces bidders to choose a lot. The largest global providers will probably compete for Lot 1. Lot 2 may then draw a thinner field, because a Swiss-resident, open-source service has to be built and operated to a standard the state will accept. The tender's success will show in how many credible bids arrive after prequalification. If competition in Lot 2 is thin, the sovereignty benefit will cost more than it should.

The third is the lesson for other jurisdictions weighing localisation mandates. The recent security news shows the attack surface is not tied to a server's location. One investigation this month described a leaky AI supply chain exposing user data, and another reported that attackers stole credentials and tokens for all of Chainalysis's customers. Data held in-country can leak just as easily through weak credentials and poor supply-chain hygiene. Localisation reduces one class of legal and jurisdictional risk. It does not replace encryption, key control, access management and audits. Procurement that pairs locality with those controls will do better than a mandate that stops at geography.

What to watch

Three things will show whether this model works. First, how many bidders clear the 60-day prequalification, and for which lots. Second, whether the SFAO's call for a demonstrated economic benefit produces a published, testable business case. Third, how agencies classify workloads once the platform opens. If most migrate to Lot 1, the localised tiers become expensive insurance. If most sit in Lot 3, the Confederation will have built a private cloud and called it a multi-cloud strategy.

The tender is a credible template for governments that want sovereignty without isolation. Tiering by sensitivity, capping supplier concentration and reopening the market every seven years all preserve choice. The remaining test is whether Switzerland can show that the extra cost of its localised tiers buys measurable risk reduction.

Sources & Citations

  1. Swiss Federal Administration: SGC tender press release (20 Oct 2026)
  2. Swiss Federal Audit Office: first audit of the Swiss Government Cloud
  3. FOITT: Swiss Government Cloud programme page
  4. Swissinfo: Switzerland takes step towards sovereign cloud
  5. W.Media: Swiss government tenders three-tier sovereign cloud