Recap of the 69th IT Press Tour in Ljubljana, Slovenia

The 69th edition of The IT Press Tour took place in Ljubljana, Slovenia, early September.Designed as a press-focused event, it brings together media representatives and a diverse range of enterprises and organizations for in-depth conversations spanning IT infrastructure, cloud computing, networking, cybersecurity, data management and storage, big data and analytics, and the growing role of AI across all these fields.

Six companies joined that edition, listed here in alphabetical order: Cybee, DataCore, ExpectedIT, HYCU, StratoFoundry and Veridion.

Cybee
Cybee, a French startup founded in Lyon in December 2022 and part of the family-owned ARYME Group, presented its cloud-native backup and cyber-resilience platform. Led by CEO Eric Deronzier and CTO Gabriel Saudin, the company builds on more than a decade of experience gained with Nuabee and earlier Ysosecure activities.

Cybee addresses four major trends: the shift of backup toward the cloud, European data sovereignty, the need for open and reversible backup formats, and growing performance requirements as data volumes increase. Based on the open-source Restic format, the platform combines immutable and encrypted backups, post-quantum resistance, fast recovery, intelligent orchestration and a unified Windows/Linux agent. Cybee claims restore performance of up to 4TB/hour per disk while avoiding the vendor lock-in associated with proprietary backup formats.

Click to enlarge

Its architecture connects protected servers through a gateway to a cloud stack integrating authentication, orchestration, storage and backup management, APIs and MCP capabilities. The platform can use multiple cloud storage providers and is designed around NIS2, DORA and ISO 27001 requirements. Early benchmarks show significant improvements, including reducing recovery of a 4TB server from 16 hours to four hours.

The go-to-market strategy starts with sister company Nuabee’s Managed DRaaS in H2 2026, covering more than 2,500 critical VMs across 150+ customers. Expansion is planned through Orange Business Flexible Recovery in H1 2027 and additional European MSPs later that year. Cybee targets organizations with more than 250 employees, particularly in regulated sectors such as healthcare, finance, transportation, industry and local government.

The roadmap includes database support in 2027, followed by XCP-ng and Proxmox hypervisor integration. Pricing starts at €5 per VM per month, with both à-la-carte S3 storage and bundled storage models.

DataCore
At the event, DataCore presented the next phase of its software-defined storage strategy through a session led by Said Boukhizou, recently promoted VP of Products. The company emphasized two major priorities: unifying its expanding product portfolio following multiple acquisitions and introducing agentic AI capabilities into storage operations.

DataCore reiterated its long-standing mission of giving customers freedom from infrastructure lock-in through four core principles: hardware freedom, data freedom, hypervisor freedom, and deployment freedom across core, edge, and cloud environments. The approach is particularly relevant as enterprises seek alternatives to VMware following Broadcom’s acquisition and need flexibility across hypervisors such as ESXi, Hyper-V, Proxmox, XCP-ng, and Xen.

Click to enlarge

The company also highlighted how infrastructure is being reshaped by four major trends: AI and HPC workloads that create data movement challenges, cyber resilience as a primary design requirement, stricter regulations including NIS2, DORA, and the Cyber Resilience Act, and the rapid expansion of distributed edge environments.

Since 2021, DataCore has significantly broadened its portfolio through acquisitions including Caringo, MayaData, Object Matrix, Pixitmedia, Arcastream, and StarWind. According to the company, the objective is no longer simply to own multiple technologies but to integrate them into a unified platform sharing common identity, telemetry, APIs, automation, lifecycle management, and support. The strategy follows a clear progression: acquire, integrate, unify, and ultimately autonomize.

A major focus is the addition of agentic AI for IT operations. Rather than providing dashboards alone, DataCore wants storage systems to observe infrastructure, reason about issues, recommend actions, and eventually execute approved operations while maintaining policy controls, auditability, rollback, and human oversight. Target use cases include capacity forecasting, performance optimization, cyber resilience validation, compliance monitoring, lifecycle automation, and support diagnostics.

The portfolio now aligns around complementary storage engines: SANsymphony for block storage, Swarm for S3 object storage and archive, StarWind for edge HCI, Nexus for parallel file storage and orchestration, and Puls8 for Kubernetes-native storage. More details are expected during IBC 2026 and the upcoming DataCore Days.

ExpectedIT
ExpectedIT, a German deep-tech startup, presented LPool during the tour. CEO Klas Moreau described a chip architecture designed to pool CPU, memory, storage, and networking resources across an entire server rack. The company compares its approach to Nvidia’s NVLink: while NVLink connects GPUs into a unified accelerator, LPool aims to eliminate boundaries between CPUs and make a rack behave as one coherent host system.

ExpectedIT sees a major opportunity emerging from agentic AI, where CPU-side processing, memory access, data movement, and orchestration increasingly become performance bottlenecks. LPool addresses stranded resources that remain underutilized on individual servers while workloads queue elsewhere. Its vendor-neutral architecture is designed to support Intel, AMD, and ARM processors.

Click to enlarge

The roadmap targets 10 sockets with its first generation, scheduled for tapeout in 2027, followed by 48 sockets with PCIe Gen 6 and eventually 96 sockets with PCIe Gen 7. Potential applications include large shared KV-cache head nodes, where ExpectedIT claims up to 7× higher throughput, as well as general-purpose scale-up servers.

The company estimates a $150B TAM by 2030, with a $30B serviceable market. Its business model combines IP licensing, fabless chip manufacturing, and potentially a longer-term platform and certification layer. ExpectedIT forecasts revenue growing from €5M in 2028 to €398M in 2032.

Technical progress includes completed memory-directory chip logic, a successful FPGA-based cross-CPU PCIe sharing demonstration, and Intel UPI licensing endorsement. The company has also signed LOIs with FractalBrain, Core42/G42, and Icelandic hosting provider 1984 ehf.

ExpectedIT currently has 16 engineers and €6.5M in seed funding, providing runway through September 2027. It plans to open a roughly €25M Series A in October 2026, targeting completion by March 2027. First CPU vendor licensing is expected in 2027, followed by production deployments in 2028.

HYCU
With a strong presence in Ljubljana, HYCU obviously presented its evolving strategy beyond SaaS backup and ransomware protection, positioning itself as an AI resilience company. CTO and co-founder Goran Garevski and Andy Fernandez, GM of AI & Cyber, explained how agentic AI is creating a new category of data protection risks.

HYCU argues that data protection has evolved through successive failure modes, from hardware failures and human errors to ransomware and now AI agents. Unlike previous threats, agents operate inside the perimeter with legitimate credentials and at machine speed. HYCU identifies three major risks: drift, where agents incorrectly interpret system state; misinstruction, including malicious prompt injection; and standing credentials that give agents excessive privileges. Traditional SaaS retention mechanisms were designed for human-scale mistakes and may prove inadequate when automated agents can modify or delete large amounts of data within seconds.

Click to enlarge

To address this challenge, HYCU promotes “protection at the speed of agents,” recommending a 30-minute RPO for critical repositories, immutable independent copies, comprehensive application coverage, and regularly tested recovery. HYCU also distinguishes recovery from continuity and introduced R-Serve, providing always-on, independent access to protected data, initially for iManage with Confluence planned.

A central element of the strategy is HYCU aiR, which transforms backup data into an organizational knowledge graph connecting applications, identities, activities, data, and time. Users can query historical backup information to investigate data exposure, user activity, security incidents, compliance issues, shadow AI, and agent behavior. This extends backup beyond recovery into security, governance, compliance, and operational intelligence.

HYCU R-Cloud now protects more than 100 workloads across hybrid infrastructure, cloud, SaaS, and AI/ML environments. New Azure DevOps support completes protection across GitHub, GitLab, Bitbucket, and Azure DevOps. HYCU says its technology is used by more than 4,800 organizations across 78 countries, reinforcing its ambition to make AI-driven resilience the next evolution of enterprise data protection.

StratoFoundry
StratoFoundry, founded in France in July 2026 by Emmanuel Forgues, presented its new data governance strategy during the 69th IT Press Tour in Ljubljana. Initially focused on S3 replication and resilience, the startup is repositioning itself as a “trusted data infrastructure” company with KARL, its new platform combining advanced object storage management, governance, and agentic AI.

KARL, standing for Knowledge, Analytics, Response & Learning, is described as a “Temporal Memory & Evidence Layer.” Rather than replacing repositories, backup products, or AI systems, it captures the governed context surrounding data and decisions. Its goal is to answer a critical question: what information, policies, AI processes, and human approvals were valid at a specific point in time? This becomes particularly important as data, policies, and AI-generated decisions continuously evolve.

Click to enlarge

Architecturally, KARL uses a governed, versioned, never-rewritten S3-compatible object store combined with signed journals, Merkle trees, RFC 3161 timestamps, content-addressed storage, indexing, and graph capabilities. Written in Rust, the platform supports multi-tenancy, RBAC/ABAC access controls, mTLS, and per-tenant metering.

StratoFoundry has developed several complementary modules. S3Guardian handles replication across S3, SaaS, and on-premises environments, while S3Proxy provides abstraction across multiple S3 endpoints. KARL Transit enables verifiable data portability; Forge tracks the evidence chain behind AI-generated software; Foundry creates sovereign local AI models while keeping customer data local; and Compass provides AI-assisted executive decision support. Additional Audit, Cyber, IA Studio, and Media modules remain largely at prototype or controlled-demo stages.

The go-to-market strategy starts with up to five paid founding design partners across Forge, Cyber, and Audit, followed potentially by MSP, compliance, AI integration, and DevSecOps channels. The roadmap moves from proving the core technology to industrializing the knowledge layer and eventually opening the ecosystem through APIs and OEM partnerships.

Importantly, KARL only reconstructs history that was captured and governed; it does not infer missing history or claim that recorded evidence is inherently true.

Veridion
Veridion presented itself as a data company rather than a software vendor, with its core asset being a continuously updated global company knowledge graph. The platform covers 693 million companies across more than 250 countries and territories, with 461+ attributes organized across 24 dimensions, including business activity, corporate structure, products, technologies, M&A, ESG, supply chains, risk signals and key executives.

Founded as Soleadify in 2019 and renamed Veridion in 2023, the company employs more than 60 people and has raised $7.5 million. Revenue is reportedly growing at more than 2× YoY, with customers including Experian, S&P Global, Moody’s, Morningstar, Dun & Bradstreet, FactSet and Hiscox.

Click to enlarge

Veridion differentiates itself from traditional registry-centric databases through automated, continuous data collection. Its infrastructure crawls approximately 1.4 billion web pages per month and generates around 100 million model predictions daily. Information from 119+ sources is collected, extracted, matched against existing entities and continuously refreshed, with a claimed match rate above 96%. Importantly, every attribute includes a confidence score and source trail, providing traceability behind the generated information.

Supporting this operation is a substantial infrastructure distributed across three replicated data centers. It includes thousands of processing and crawling cores, around 5.3TB of GPU VRAM, 100TB of hot NVMe object storage, 640TB of HDD object storage, 4PB of cold HDFS capacity and a 252TB Cassandra database.

Customers access the same knowledge graph through Match, Enrich, Search, Location, ESG and Corporate Groups APIs, batch feeds, or native integrations with Snowflake, BigQuery and Databricks. An MCP interface is also planned for AI agents.

Veridion’s business model is based on enterprise data access rather than user seats. Its ambition is to replace periodically refreshed company databases with a live, machine-built, evidence-backed representation of the global business landscape serving enterprise applications and, increasingly, AI agents.