This page could not be loaded. Check your connection, then reload the page.
The Workloads Are Not Leaving
Where production workloads run today and in ten years, and what buyers are changing about them. If sovereignty were moving workloads to sovereign alternatives, the ten-year picture would show sovereign and regional providers gaining share. The largest US cloud providers are the only environment that grows across the decade; every other environment (sovereign and in-country hyperscaler offerings, regional and local providers, on-premises infrastructure, private cloud platforms, and managed hosting or colocation) loses share or stays flat. The pattern is pervasively consistent across respondent roles, regions, industries, sizes, and sovereignty maturity. For an organization building a sovereignty plan, this settles where the workloads will sit in ten years. What remains open is the terms attached to that hosting: the contract and the architecture around it.
Where production workloads run, today and in three, five, and ten years All respondents
| Category | US-headquartered hyperscaler cloud | Sovereign / in-country hyperscaler offerings | Regional or local cloud providers | Private cloud | On-premises infrastructure | Managed hosting or colocation |
|---|---|---|---|---|---|---|
| Today | 17% | 16% | 17% | 18% | 17% | 16% |
| In 3 years | 23% | 15% | 15% | 16% | 17% | 15% |
| In 5 years | 25% | 16% | 15% | 15% | 16% | 14% |
| In 10 years | 27% | 16% | 15% | 14% | 15% | 13% |
Source: n=1,940 respondents, normalized by region
Share of production workloads that run on US hyperscalers today: 17.0%
Organizations that expect one environment to hold half or more of their workloads in ten years: 48.4%
Respondents that say it is extremely important that sovereignty comes from the vendors they already use: 41.3%
[We aren't moving] on-prem, but we are diversifying…we're working with another hosting company based in France for some of our workloads…it became a lot more important criteria [in RFIs and RFPs]. — CIO, enterprise retailer in EMEA
Organizations worried about depending on US providers are concentrating fastest
Concentration is rising alongside that shift: within ten years, many organizations expect a single environment to hold most of their production workloads, and for a little over half of them, that environment will be a US hyperscaler. Concentration compounds the readiness problem: the more an organization's workloads sit in one place, the larger any move would be and the fewer places it could go. Organizations most concerned about depending on US-headquartered providers, and organizations pulling work back in-house, expect a steeper rise in US hyperscaler share than organizations that are not concerned or not pulling work in-house. These buyers are staying with the largest providers because that is where new capabilities arrive first, treating a fallback plan as insurance while governing from inside an already concentrated position. Doing that safely requires a tested exit and a second source of support beforehand.
A US hyperscaler with a sovereign or in-country offering is the joint most-cited provider model at 30.4%, level with private cloud platforms and well ahead of regional, local, or sovereign providers at 23.4%. Buyers are not planning to leave the largest providers in order to become sovereign; they are planning to buy sovereignty from them, or to place a portable platform layer on top. In interviews, several executives described continuing with hyperscalers as a reluctant conclusion rather than a preference; the capability, the pace of new services, and the absence of a like-for-like alternative made the decision for them. That is acceptance, rather than enthusiasm, and it drives a different set of planning assumptions.
Notably, even holding data in-country in a facility owned by a company headquartered elsewhere does not put it outside the reach of that country's law, and encryption at rest does not change that as long as the provider holds the keys. The defensible test is what is contractually enforceable: who can be compelled to hand over the data, and what migration path exists before anyone needs it.
Workloads are concentrating, and the most worried organizations expect to concentrate fastest All respondents
| Category | A US hyperscaler is that environment | Another single environment |
|---|---|---|
| Today | 2% | 13% |
| In ten years | 26% | 22% |
Source: n=1,940 respondents, normalized by region
Workloads are concentrating, and the most worried organizations expect to concentrate fastest: ten-year change in US hyperscaler share, in points All respondents
| Category | Ten-year change in US hyperscaler share |
|---|---|
| Not concerned | 6% |
| Moderately concerned | 8% |
| Extremely concerned | 10% |
| Pulling work in-house | 10% |
Source: n=363 / 843 / 734 / 456 respondents
We have no concern using US cloud. Our group has bought AI companies in the US But regulations in India mandate that our data reside in our own data centers. — CTO, enterprise financial services organization in APAC
The money is moving between vendors, not back in-house
Most organizations are switching providers, rebalancing workload by workload, or outsourcing more; only a minority are pulling work back in-house. Most buyers also want that sovereignty delivered by the vendors that already run their cloud and infrastructure (see the callout). For a leader, the practical lever is the renewal conversation with an incumbent supplier, and the terms that matter are a documented migration path and a second source of support.
The money moves between vendors, not back in-house All respondents · 54.5% expect to outsource the same or more
| Category | All respondents |
|---|---|
| Switching to sovereignty-aligned providers (same ratio) | 30% |
| Re-balancing by workload | 23% |
| Pulling more work back in-house | 23% |
| Outsourcing more (vetted partners) | 19% |
| No change | 5% |
Source: n=1,940 respondents, normalized by region
The provider models buyers expect to engage with confirm it: as noted above, a US hyperscaler with a sovereign or in-country offering and a private cloud platform provider sit jointly at the top, ahead of regional, local, or sovereign cloud providers (23.4%). No single class of workload dominates what is moving: analytics, backup and disaster recovery, AI training and inference, and core transactional systems are all being moved by roughly a third of organizations.
Provider Models Buyers Expect to Engage All respondents (respondents could select more than one)
| Category | All respondents |
|---|---|
| US hyperscaler with sovereign or in-country offering | 30% |
| Private cloud platform providers | 30% |
| Regional/local/sovereign providers | 23% |
| On-premises with commercial open-source vendor | 19% |
| Telecom/managed hosting/colocation | 18% |
| Systems integrators or MSPs | 17% |
| Independent or specialized cloud providers | 17% |
| Build internally / on-premises | 14% |
Source: n=1,940 respondents, normalized by region
Respondents expect consolidation; Futurum expects fragmentation
The ten-year projection captures what respondents expect from the estate they know today: fewer environments and a larger share of production workloads on US-headquartered hyperscalers. That is a useful record of buyer intent. It is not necessarily a forecast of the market those buyers will face.
Futurum's three- to five-year outlook points in a different direction. Regulation is becoming more region-specific and more enforceable. Sovereign and in-country cloud offerings are becoming operationally distinct rather than merely positioning. Local and regional providers are gaining relevance, and telecommunications operators are moving from buyers of infrastructure to suppliers of sovereign cloud and AI services.
The result is likely to be a more fragmented supply landscape, not a simple continuation of today's concentration. Public-sector and regulated buyers will have more reasons — and more options — to separate sensitive data, AI, and critical workloads from dependencies they judge difficult to govern, support, or exit. The question for leaders is not whether US hyperscalers will remain central. They will. It is whether a strategy built around today's provider hierarchy remains appropriate as regulation, local capacity, and sovereignty requirements change around it.
Concentration is not a hyperscaler-specific problem and treating it as one would be a mistake. Of the organizations expected to hold half or more of their production workloads in a single environment within ten years, a little over half are concentrated on a US hyperscaler; the rest are concentrated on a regional provider, a private cloud platform, or their own infrastructure. The regulatory position reflects this: operational resilience rules such as DORA address systemic overreliance on any single provider, not on foreign ones. An organization that moves from one hyperscaler to one local cloud provider has changed its jurisdiction and kept its concentration risk. What reduces the risk is the ability to move, including a documented exit, a second source of support, and a platform layer whose tooling is not tied to the provider underneath it.
The direction of travel is clearer than its speed, and that uncertainty requires active management. The survey asked respondents to project their workload mix ten years ahead, yet the technology, regulatory, geopolitical, and supplier landscape has changed materially in only the past two or three years. A ten-year allocation should therefore be treated as a strategic assumption, not an operational plan.
Leaders should test that assumption regularly. They should identify which workloads require the newest cloud capabilities today, which could run on regional, sovereign, private, or on-premises platforms as those alternatives mature, and what events would change the preferred placement. The relevant decision may also change as AI and cloud capabilities become more widely available.
The practical conclusion is not to avoid consolidation. It is to avoid making an irreversible bet on it.
What to Do About It
Write down the assumptions behind your provider concentration and put a review date on them. A ten-year allocation is a strategic assumption, not an operational plan.