Scaleout Systems
Sovereign edge AI platform for federated learning in disconnected and contested environments.
Website: https://www.scaleoutsystems.com/
Cover Block
Public sources
| Field | Detail |
|---|---|
| Company | Scaleout Systems |
| Tagline | Sovereign edge AI platform for federated learning in disconnected and contested environments. |
| Headquarters | Uppsala, Sweden [Scaleout Systems] |
| Founded | 2018 [Scaleout Systems] |
| Stage | Seed [LinkedIn] |
| Business model | SaaS |
| Industry | Deeptech |
| Technology | AI / Machine Learning |
| Geography | Western Europe |
| Growth profile | Venture Scale |
| Founding team | Academic Spinout [Uppsala University, October 2021] |
| Funding label | Seed |
| Total disclosed | $1.5M in 2023 seed funding, with a further SEK 35m round announced in February 2025 [ArcticStartup] [Scaleout, February 2025] |
Links
Public sources
- Website: https://www.scaleoutsystems.com/
- LinkedIn: https://www.linkedin.com/company/scaleout-systems/
- GitHub: https://github.com/scaleoutsystems
Executive Summary
PUBLIC Scaleout Systems builds federated learning and edge AI infrastructure for organizations that need to train and run models across distributed environments without centralizing sensitive data, and the company merits attention now because it is starting to convert an academic architecture into a defense and industrial deployment story with fresh capital behind it [Uppsala University, October 2021] [Scaleout, February 2025] [RuntimeWire, September 2026]. Founded in 2018 as a spinout from Uppsala University in Sweden, the company emerged from research in distributed systems and scientific computing, with its early positioning centered on privacy-preserving machine learning for regulated and industrial settings [Uppsala University, October 2021] [Scaleout Systems].
The product stack appears to rest on FEDn, a federated learning framework, and what the company describes as a cloud-edge platform for deploying models closer to where data is generated; the practical distinction is that Scaleout is selling around data sovereignty, distributed training, and operation in environments where constant connectivity cannot be assumed [FEDn Documentation] [Scalable Federated Learning with FEDn: A Comprehensive Overview] [Scaleout Systems]. Public materials point to support for APIs, Kubernetes-based control-plane deployment, experiment orchestration, and collaboration features, which suggests a platform ambition rather than a point solution [FEDn Documentation].
The founding bench is unusually research-heavy for a seed-stage software company. Chief executive Andreas Hellander and chief technology officer Salman Toor are both affiliated with Uppsala University, and Toor is identified in public coverage as a lead architect of FEDn, which strengthens the technical provenance even if commercial scaling evidence is still limited [LinkedIn, Retrieved 2026] [RuntimeWire, September 2026].
On capitalization, the public record is directionally positive but not perfectly clean. A late-2023 seed round of about $1.5 million is reported by company-linked and secondary sources, and Scaleout announced a SEK 35 million financing in February 2025 led by Fairpoint Capital, while PitchBook reports $5.51 million in total funding and the company operates on a SaaS model [LinkedIn] [Scaleout, February 2025] [PitchBook].
The next 12 to 18 months matter less for category education than for proof of repeatability. Investors should watch for evidence that defense and industrial pilots convert into durable software deployments, that NATO DIANA participation and related non-dilutive support translate into procurement momentum, and that the company can establish a customer base broad enough to validate the platform beyond founder-led technical sales [NATO DIANA | Accelerator Programme] [RuntimeWire, September 2026] [Scaleout Systems].
Lightly corroborated -- Mix of company materials, university coverage, technical documentation, and selected third-party reporting; several core product and funding details are corroborated, but some operational claims remain company-sourced.
Taxonomy Snapshot
| Axis | Value |
|---|---|
| Stage | Seed |
| Business Model | SaaS |
| Industry / Vertical | Deeptech |
| Technology Type | AI / Machine Learning |
| Geography | Western Europe |
| Growth Profile | Venture Scale |
| Founding Team | Academic Spinout |
| Funding | Seed, total disclosed about $1.5M; additional SEK 35m round announced in February 2025 [LinkedIn] [Scaleout, February 2025] |
How the Company Got Here
PUBLIC
Scaleout Systems traces to Uppsala University research rather than a marketing-first software launch, which matters because the company still reads like an academic spinout commercializing a specific technical architecture. The company says it was founded in 2018 in Uppsala, Sweden, by researchers working on large-scale distributed systems and scientific computing, and its public materials frame the original problem as machine learning across sensitive, decentralized data without centralizing the underlying datasets [scaleoutsystems.com] [Crunchbase].
The public chronology is thin but directionally clear. Uppsala University profiled the company in October 2021 as a university-founded venture focused on federated learning for privacy-sensitive collaboration across organizations, which aligns with the company website's current positioning around sovereign edge AI and distributed model training [scaleoutsystems.com] [uu.se, October 2021]. By February 2025, Scaleout said it had raised a new investment round to accelerate cloud-edge AI development, suggesting a visible shift from a narrower federated-learning wedge toward broader deployment infrastructure in defense and other distributed environments [scaleoutsystems.com, February 2025].
Publicly named founders include Andreas Hellander, Salman Toor, Daniel Zakrisson, Jens Frid, Ebba Kraemer, Morgan Ekmefjord, and Ola Spjuth, although the exact founding roster varies somewhat by source and date [scaleoutsystems.com] [Crunchbase] [uu.se, October 2021]. No state filing or public legal-entity record was included in the materials reviewed for this section, so the cleaner read is that the core verified facts are the 2018 founding, Uppsala base, university-spinout origin, and the company's later expansion narrative around cloud-edge AI [scaleoutsystems.com] [Crunchbase].
Lightly corroborated -- Core founding and HQ details are corroborated by Crunchbase and the company website, but milestone chronology beyond 2025 relies partly on company materials and a university profile.
Product and Technology
Sources and analysis
Scaleout Systems is selling infrastructure for machine learning where the data and the compute are spread across many devices or organizational silos, and where centralizing that data is either impractical or undesirable [Uppsala University, October 2021] [Crunchbase]. In plain terms, the company’s public materials describe software for federated learning and edge AI, with the core promise that customers can train shared models across distributed environments while keeping underlying data with the original data controllers [Scaleout Systems] [Uppsala University, October 2021]. That positioning has remained consistent from the company’s earlier university spinout framing into its newer defense and cloud-edge language, although the more specific claims about sovereign AI, tactical computer vision, and autonomous systems are still primarily company-authored and should be read that way [Scaleout Systems] [Ars Technica, September 2026].
The named product surface is clearer than the broader platform narrative. Public documentation and coverage identify FEDn as a federated-learning framework, and third-party or technical sources describe a REST API for experiments and jobs, Kubernetes support for the server-side control plane, a dashboard-based UI, collaboration features for team workspaces, and administrative controls for managing network access and training progress [RuntimeWire, September 2026] [FEDn Documentation] [Scalable federated machine learning with FEDn] [Scalable Federated Learning with FEDn: A Comprehensive Overview]. GitHub materials also describe Scaleout Edge as an evolution of FEDn, while the company says customers can deploy its edge platform on their own infrastructure or on AWS and GCP using Terraform [GitHub] [Scaleout Systems]. The practical read is that Scaleout is not presenting as a single demo application vendor, but as an orchestration layer for training, deploying, and operating models across distributed sites.
What is verified in public goes furthest on architecture and least far on repeatable production outcomes. Sources support that FEDn is designed for horizontally scalable distributed deployments, can operate across both cross-device and cross-silo settings, and is intended for industrial as well as research use cases [Scalable federated machine learning with FEDn] [Scalable Federated Learning with FEDn: A Comprehensive Overview]. Separately, press coverage in 2026 points to a defense-oriented onboard AI demonstration involving drones and target selection, which is evidence that the company’s technology has been shown in a real-world military context, but it does not by itself establish broad deployment, standardization, or commercial scale across sectors [RuntimeWire, September 2026] [Ars Technica, September 2026].
Lightly corroborated -- Core product category and FEDn capabilities are supported by Uppsala University, technical documentation, GitHub, and press coverage, but several platform-level claims still rely on company materials.
Where the Demand Sits
Public sources This market matters now because Scaleout sits at the intersection of two budget lines that have both strengthened in Europe: practical AI deployment and data sovereignty, especially where data cannot easily be centralized across sites, devices, or jurisdictions [Uppsala University, October 2021] [Crunchbase].
The evidence base here is narrower than it would be for a mature software category, so the market read has to stay close to adjacent public markets rather than claim a precise federated-learning TAM. None of the captured sources provides a named third-party TAM, SAM, or SOM for federated learning or sovereign edge AI specific to Scaleout. What is public is the use case definition: the company and Uppsala University describe software that lets organizations train shared models across decentralized data sources while keeping underlying data with the original controllers, a pattern that fits regulated industry and distributed-asset environments rather than centralized internet software [Uppsala University, October 2021] [Scaleout Systems].
Demand signals in the source set point to a real problem statement, even if they do not yet size the category cleanly. Uppsala University said in 2021 that the company was already collaborating with automotive and aerospace companies and discussing work in insurance and finance, which suggests early relevance anywhere sensitive data is fragmented across business units or physical sites [Uppsala University, October 2021]. More recent public material shifts the emphasis toward cloud-edge AI, defense, computer vision, and autonomous systems, indicating that the addressable market may be broadening from privacy-preserving collaboration into operational AI at the edge [Scaleout Systems, February 2025] [RuntimeWire, September 2026].
Regulation and procurement patterns are also part of the tailwind. The company frames its product around retaining control of data, and that framing aligns with European enterprise and public-sector buying priorities around sovereignty, security, and trusted infrastructure, even where the sources stop short of naming a specific regulation [Scaleout Systems] [About the Company, Scaleout]. NATO DIANA adds a second layer of relevance: selection into the program and the associated contractual funding suggest that allied defense institutions see a live need for AI systems that can operate in disconnected or contested environments, not only in standard cloud settings [RuntimeWire, September 2026] [NATO DIANA | Accelerator Programme].
| Market slice | What the public record supports | Source basis |
|---|---|---|
| Federated learning | Decentralized model training without moving all source data to a central repository | [Uppsala University, October 2021] [Scaleout Systems] |
| Edge AI and cloud-edge AI | Deploying models close to where data is generated, including distributed devices and sites | [Scaleout Systems, February 2025] [RuntimeWire, September 2026] |
| Defense and autonomous systems | AI for disconnected and contested environments, including NATO-linked programs and drone-related demonstrations | [RuntimeWire, September 2026] [Ars Technica, September 2026] |
| Regulated and industrial data environments | Automotive, aerospace, healthcare, insurance, and finance appear as adjacent adoption zones in public sources | [Uppsala University, October 2021] [The Hub] |
The table shows a market better understood as an overlap of adjacent categories than as one clean software bucket. That usually helps early-stage companies explain urgency to buyers, but it can complicate category ownership because procurement, competition, and product requirements differ across industrial AI, privacy-preserving collaboration, and defense technology.
Adjacent and substitute markets matter because buyers may solve the same problem in several ways. In some accounts, the alternative is a conventional centralized MLOps stack that moves data into one repository, which works where privacy, bandwidth, or operational constraints are manageable [Crunchbase] [Scaleout Systems]. In other accounts, especially at the edge, the substitute is local inference and device management without collaborative training, or bespoke systems integration for defense and industrial fleets rather than a general-purpose federated platform, an important distinction because it affects both contract structure and sales cycle length [RuntimeWire, September 2026] [FEDn Documentation].
The public market read, then, is less about a single headline TAM and more about whether sovereignty, security, and edge deployment become durable buying criteria across Europe and allied defense ecosystems. The sources support that direction of travel, but they do not yet establish how much of the resulting spend accrues to horizontal software platforms versus primes, integrators, and application-specific vendors [RuntimeWire, September 2026] [Ars Technica, September 2026].
Lightly corroborated -- This section relies on one independent institutional source for early market context, plus company materials and later news coverage for demand signals; no named third-party TAM/SAM/SOM report was captured.
Competitive Landscape
MIXED Scaleout Systems is positioned less against a clearly disclosed peer set than against a stack of alternatives, namely centralized MLOps platforms, in-house federated learning builds, and defense or industrial integrators that can package edge AI into larger programs [Scaleout Systems, Unknown] [Uppsala University, October 2021] [Crunchbase].
The public record is unusually thin on named direct competitors, which matters in itself. The company is explicit about federated learning, cloud-edge AI, and sovereign deployment in distributed environments [Scaleout Systems, Unknown] [Uppsala University, October 2021], but the cited materials do not name rival startups or benchmarks. That leaves three practical comparison groups. First are incumbent cloud and MLOps approaches built around centralizing data for training, which are the default in many enterprises but fit poorly where data residency, latency, connectivity, or operational security constrain movement of data [Uppsala University, October 2021] [Scaleout Systems, Unknown]. Second are internal engineering teams at primes, OEMs, and regulated enterprises that can assemble their own federated learning workflows from open-source components such as Kubernetes-based control planes and APIs [FEDn Documentation] [Scalable federated machine learning with FEDn]. Third are adjacent defense and industrial solution providers, including systems integrators and platform partners such as Oracle or Akkodis, that may not sell federated learning as the product but can absorb it into a broader delivery motion [LinkedIn] [RuntimeWire, September 2026].
Where Scaleout appears strongest today is at the intersection of technical depth and deployment context. The founding team is tied to Uppsala University and scientific computing, and public materials connect the company to the FEDn framework, which has documented support for REST APIs, Kubernetes deployment, experiment dashboards, team collaboration, and both cross-device and cross-silo federated learning use cases [Uppsala University, October 2021] [FEDn Documentation] [Scalable Federated Learning with FEDn: A Comprehensive Overview] [RuntimeWire, September 2026]. That is a credible edge for buyers who need more than a research prototype but are not served by centralized AI stacks. The durability of that edge is mixed. The academic roots, documented framework, and alignment with sovereign-AI procurement themes are real assets [Uppsala University, October 2021] [About the Company, Scaleout] [RuntimeWire, September 2026]. Even so, software-led differentiation in orchestration and deployment can erode if larger vendors package similar functionality into broader platforms, or if defense and industrial customers prefer integrators that already control the procurement relationship.
The clearest exposure is distribution. Public sources name partner and customer relationships including Oracle, Saab, BAE Systems Bofors, Scania, BMW, FMV, and Akkodis [LinkedIn] [RuntimeWire, September 2026], which suggests access to demanding environments, but it does not establish that Scaleout owns the prime contractor role or the budget line. In practical terms, Oracle has the advantage if the purchase centers on cloud estate standardization, and major defense primes have the advantage if the requirement is wrapped inside a larger vehicle, autonomy, or command-and-control program [Oracle and Scaleout bring Federated Learning to the Tactical Edge] [RuntimeWire, September 2026]. By contrast, Scaleout's position is strongest where the buyer explicitly prioritizes distributed training under data-control and connectivity constraints. The company also looks less naturally positioned for mainstream enterprise AI categories where central data aggregation is acceptable and the deciding factor is enterprise sales scale rather than deployment architecture.
The most plausible 18-month competitive scenario is a bifurcation rather than a broad platform land grab. If European defense and sovereign-AI procurement continues to reward local control, edge deployment, and resilience in disconnected settings, Oracle looks like the winner if procurement consolidates around ecosystem partners, while Scaleout benefits as a specialist layer inside those programs [Oracle and Scaleout bring Federated Learning to the Tactical Edge] [RuntimeWire, September 2026]. If, instead, buyers treat federated learning as a feature to be folded into internal platforms or bundled cloud tooling, in-house teams are the likely winner if integration simplicity outweighs specialist capability, and Scaleout is the loser if its architecture is valued but not purchased as a standalone control point [FEDn Documentation] [Scalable federated machine learning with FEDn] [Scaleout Systems, Unknown]. The distinction to watch is whether procurement rewards the best federated-learning stack, or the vendor that already owns the surrounding workflow.
Lightly corroborated -- Competitive positioning is supported by company materials, Uppsala University coverage, FEDn documentation, and third-party reporting, but the sources do not name a verified direct competitor set or provide independent side-by-side market share evidence.
Opportunity
PUBLIC
The prize here is not a niche software outcome, it is the chance to become trusted infrastructure for training and operating AI in settings where data cannot be centralized and connectivity cannot be assumed, a problem that spans regulated industry and, increasingly, European defense programs [Uppsala University, October 2021] [Scaleout Systems, February 2025] [RuntimeWire, September 2026].
The headline opportunity is straightforward: if Scaleout Systems can turn its academic lead in federated learning into a repeatable deployment platform, it could become a default control layer for sovereign edge AI in Europe. That is a bigger claim than selling a federated-learning toolkit, and the public record gives at least a path to it. The company has operated since 2018 as a Uppsala University spinout focused on distributed machine learning [Uppsala University, October 2021] [LinkedIn]. Its named product base includes FEDn, which public documentation describes as a framework with orchestration, dashboards, team collaboration, REST APIs, and Kubernetes-based deployment support, while company materials position the broader platform around cloud-edge deployment and sensitive, distributed data environments [FEDn Documentation] [Scaleout Systems] [Scalable Federated Learning with FEDn: A Comprehensive Overview].
What makes that outcome reachable rather than purely aspirational is the combination of technical specificity and early institutional validation. The February 2025 financing round of SEK 35 million points to investor willingness to fund the cloud-edge AI expansion thesis [Scaleout Systems, February 2025] [AI training platform Scaleout secures SEK 35m in a new funding round, February 2025]. NATO DIANA participation adds a second signal, because the program is explicitly oriented around defense-relevant technology and includes contractual funding for selected companies [NATO DIANA | Accelerator Programme] [RuntimeWire, September 2026]. In a market where buyers care about data control, deployment flexibility, and local operation, those facts matter more than broad AI rhetoric.
| Scenario | What happens | Catalyst | Why it's plausible |
|---|---|---|---|
| European sovereign AI layer | Scaleout becomes a common infrastructure choice for defense, public-sector, and regulated industrial deployments that need model training across distributed data without centralization. | NATO DIANA participation converts into procurement-grade credibility and follow-on deployments with defense ecosystem partners [RuntimeWire, September 2026] [NATO DIANA | Accelerator Programme]. |
| Industrial fleet intelligence platform | The company expands from federated learning into full cloud-edge operations for vehicles, drones, and industrial fleets, becoming the software layer that updates and improves models across many sites. | Productization of Scaleout Edge as the evolution of FEDn, with easier deployment on customer infrastructure and cloud environments [GitHub] [Scaleout Systems] [FEDn Documentation]. | Public materials already describe deployment on customer-owned infrastructure, support for distributed environments, and an emphasis on operating models close to where data is generated [Scaleout Systems] [Scalable federated machine learning with FEDn] [Scalable Federated Learning with FEDn: A Comprehensive Overview]. |
| Trusted collaboration rail for sensitive data | Scaleout becomes the preferred platform for multi-party machine-learning collaboration in healthcare, automotive, and other sectors where participants need to retain control of underlying data. | A category shift toward privacy-preserving collaboration, pushed by regulatory pressure and enterprise demand for keeping data with the original controller [Uppsala University, October 2021]. | That was the company's original wedge, and the core technical proposition is well matched to sectors where data sharing is hard but model improvement still benefits from pooled learning [Uppsala University, October 2021] [LinkedIn]. |
These scenarios rest on the same underlying compounding logic. One successful deployment should make the next sale easier because the buyer is not only purchasing model training software, but a reference architecture for operating AI across disconnected endpoints, governed data silos, and mixed infrastructure. FEDn's public feature set, including orchestration, access management, shared workspaces, and flexible deployment from local to geographically distributed cloud environments, suggests the company is building for repeatability rather than one-off research projects [FEDn Documentation] [Scalable Federated Learning with FEDn: A Comprehensive Overview].
The flywheel, if it forms, would likely come from three places. First, each production deployment sharpens a library of patterns for cross-silo and cross-device training, which the literature around FEDn already describes as horizontally scalable and suitable for industrial applications [Scalable federated machine learning with FEDn] [Scalable Federated Learning with FEDn: A Comprehensive Overview]. Second, sectors that start with narrow pilots can broaden into adjacent use cases once the control plane is in place, because the underlying need, keeping data local while improving shared models, repeats across computer vision, anomaly detection, and other workloads [LinkedIn] [Scaleout Systems]. Third, defense and regulated-industry references tend to carry outsized signaling value, especially in Europe, where sovereign AI procurement is becoming a strategic rather than purely technical question [RuntimeWire, September 2026] [Scaleout Systems, February 2025].
The size of the win is hard to pin down from public comparables because the category is still forming, and no direct public peer was provided in the source set. Even so, one public anchor exists inside Scaleout's own financing record: the February 2025 round was reported at SEK 35 million with a SEK 132 million pre-money valuation, implying a post-money valuation of about SEK 167 million (estimated) based on simple addition [AI training platform Scaleout secures SEK 35m in a new funding round, February 2025]. If the company were to convert the "European sovereign AI layer" scenario into a scaled infrastructure position across defense and regulated industry, a 10x to 20x increase from that post-money baseline, or roughly SEK 1.7 billion to SEK 3.3 billion (scenario, not a forecast), is not difficult to imagine conceptually. That range is an analytical scenario built from the reported 2025 valuation anchor, not from disclosed revenue, and it turns entirely on whether Scaleout becomes embedded infrastructure rather than a specialist tooling vendor [AI training platform Scaleout secures SEK 35m in a new funding round, February 2025] [RuntimeWire, September 2026].
Lightly corroborated -- This section relies on a mix of company materials, product documentation, Uppsala University coverage, and one named-publisher defense report; the valuation anchor is supported by a single published report and the upside range is explicitly scenario-based.
Sources
Public sources
[Uppsala University, October 2021] Tech company founded at Uppsala University newcomer on Swedish 33-list. | https://www.uu.se/en/news/2021/2021-10-28-tech-company-founded-at-uppsala-university-newcomer-on-swedish-33-list
[Scaleout, February 2025] Scaleout secures new investment round to accelerate Cloud-Edge AI | https://www.scaleoutsystems.com/post/scaleout-secures-new-investment-round-to-accelerate-cloud-edge-ai
[RuntimeWire, September 2026] Scaleout’s edge AI picked and struck a target without a radio … | https://runtimewire.com/article/scaleout-edge-ai-autonomous-drone-targeting-alma
[LinkedIn, Retrieved 2026] Andreas Hellander - Scaleout | LinkedIn | https://www.linkedin.com/in/andreas-hellander-5025b189/
[ArcticStartup] Sweden's Scaleout Systems receives €1.5M seed funding for innovating federated machine learning, prioritizing data security and privacy in AI | https://arcticstartup.com/scaleout-raises-e1-5m-seed/
[scaleoutsystems.com, February 2025] Scaleout secures new investment round to accelerate Cloud-Edge AI | https://www.scaleoutsystems.com/post/scaleout-secures-new-investment-round-to-accelerate-cloud-edge-ai
[uu.se, October 2021] Tech company founded at Uppsala University newcomer on Swedish 33-list. | https://www.uu.se/en/news/2021/2021-10-28-tech-company-founded-at-uppsala-university-newcomer-on-swedish-33-list
[runtimewire.com, September 2026] Scaleout’s edge AI picked and struck a target without a radio … | https://runtimewire.com/article/scaleout-edge-ai-autonomous-drone-targeting-alma
[Ars Technica, September 2026] Small AI models let drones autonomously identify and attack… | https://arstechnica.com/ai/2026/09/nato-backed-startup-adapts-ai-for-autonomous-drone-recon-and-attack-missions/
[The Hub] The Hub | Scaleout Systems | https://thehub.io/startups/scaleout-systems
Articles about Scaleout Systems
- Scaleout Systems Trains AI Models Where Data Cannot Move — The Uppsala spinout, backed by Fairpoint Capital and NATO DIANA, is betting that federated learning is the key to AI in disconnected environments.