Sidetrade launches sovereign AI stack to lock CFOs out of hyperscalers

Sidetrade's SAFE framework runs agentic finance AI entirely on owned GPUs and data centres, targeting a 30–35% EBITDA margin by 2030.

By Disrupts Editorial Team Tue, 15 September 2026 AI level 2 of 5

AI level 2 of 5: drafted by our AI editorial assistant from the company's announcement; fact-checked, edited and signed off by Mark Walker, Editorial Director. What the levels mean

A brightly lit, modern control room features a multi-panel video wall displaying abstract blue shapes, a row of light wood desks with control panels, keyboards, mice, and chairs, and large windows revealing a city skyline.

Sidetrade, the Euronext Growth-listed Order-to-Cash (O2C) software company, has launched SAFE, the Sidetrade Agentic Framework for Enterprise, a fully owned AI stack designed to let corporate finance functions deploy autonomous agents without routing sensitive receivables data through third-party hyperscalers or frontier model providers. The move reframes a debate that has dominated enterprise AI adoption: not which model is most powerful, but who controls the infrastructure, the data, and ultimately the bill.

The launch lands against a stark industry backdrop. McKinsey's State of Organizations 2026 report, cited by Sidetrade, found that while 88% of organisations have deployed AI, 81% have yet to see significant financial impact. The gap is sharpest in finance, where agents touching receivables, disputes and credit risk carry data-governance obligations that generic cloud deployments struggle to satisfy.

A full-stack answer to the sovereignty problem

SAFE consolidates five layers of AI infrastructure, data centres, GPU compute, fine-tuned open-weight language models, an agent framework, and a cross-network domain dataset, under Sidetrade's own ownership. The company says it operates private data centres on both sides of the Atlantic, runs its own token production, and holds ISO 27001 and SOC Type II certification. Agents are validated against OWASP's Agentic Top 10 before they can be published to SAFE's internal Agent Registry.

The proprietary data layer is where Sidetrade makes its most assertive competitive claim. Its O2C Data Lake aggregates close to $10 trillion in B2B transactions across nearly 45 million buying companies, accumulated over more than a decade. "This choice is ten years old, not ten months old," said CEO Olivier Novasque. "Sidetrade began building its O2C data asset in 2015 and moved its models onto its own GPUs before the agentic wave made compute scarce. SAFE completes that stack."

SAFE's model-routing architecture directs each task to the most cost-efficient model available. Deterministic code handles event routing and policy enforcement; proprietary predictive models score payment risk; only tasks requiring language reasoning, such as drafting dispute responses, reach the fine-tuned LLM layer. The design is also a cost argument. As agentic AI pushes vendors from subscriptions toward consumption-based pricing, Sidetrade is offering multi-year contracts with locked compute pricing, selling capacity in O2C IQ units rather than exposing customers to third-party per-million-token volatility.

SAFE is already in production under the Aimie Cash Collection Agent, deployed with multinationals including Securitas, Accor and Sodexo. A wider rollout of Aimie IQ and Agent Builder Studio to beta customers is scheduled for Q4 2026.

The convergence angle: sovereignty as a capital thesis

The SAFE launch sits at a convergence point between enterprise software, private AI infrastructure, and a rapidly evolving CFO technology market. Sidetrade is, in effect, making the same architectural argument that is reshaping the hyperscaler relationship across multiple industries: that for mission-critical, data-sensitive workflows, the economics and governance of rented public compute are becoming untenable.

This dynamic is not unique to finance. Defence primes, healthcare networks, and industrial manufacturers are all facing the same question of whether frontier AI labs and hyperscalers are appropriate custodians for their most sensitive operational data. Sidetrade's O2C specialisation gives it a narrow but deep domain dataset that a general-purpose model provider cannot replicate quickly, and the company is now packaging that moat into a contractually predictable commercial model.

For investors watching the enterprise AI landscape, SAFE's launch signals a maturing segment: the era of "which model wins" is giving way to "who owns the stack." Sidetrade's 2030 plan targets more than 50% of revenue from AI-native products and an EBITDA margin of 30% to 35%, up from the company's current profile. Whether owned-infrastructure unit economics can scale to support those targets is the open question heading into the company's first-half 2026 results, due 22 September.