Three days after Pope Leo XIV’s encyclical Magnifica Humanitas called for AI to be “disarmed,” Mistral CEO Arthur Mensch mounted the most direct corporate rebuttal yet to what has become the Catholic Church’s most consequential intervention on AI regulation. The venue was Mistral’s inaugural AI NOW Summit in Paris on 28 May 2026. The timing was not incidental.
Mensch’s position was blunt: “Choices about deployment and usage are not our business,” he told reporters, as the company disclosed that defence activities account for between 10 and 15 percent of revenue, with active contracts for the French, Singaporean, and Luxembourg armed forces. That revenue share, while modest, anchors a strategic posture that distinguishes Mistral from most of its American peers.
Anthropic had attempted to bar the US government from using its systems for mass surveillance or fully autonomous weapons, a stance that led to a legal confrontation with the Pentagon — one that has since partially thawed. Meanwhile, staff at Google DeepMind protested the sale of their technology to Israeli and US forces. Mensch’s formulation cuts through that paralysis: “Europe needed its own AI tools because rivals and adversaries were already using them.”
France’s Ministry of the Armed Forces had already formalised a three-year framework deal with Mistral, notified on 16 December 2025 and steered by AMIAD, the Agence ministérielle pour l’intelligence artificielle de défense; access extends to the CEA, ONERA, and the Navy’s Hydrographic Service. The ministry’s defence AI strategy, launched in March 2024, allocated €130 million in the 2024 budget, with that funding set to double by the end of the current military programming law. Mistral is not a speculative beneficiary of that envelope — it is now its primary private-sector contractor.
In February 2025, Mistral and German defence-tech firm Helsing announced a strategic partnership at the AI Action Summit in Paris, overseen by AMIAD with an annual budget of approximately €300 million. The collaboration centres on Vision-Language-Action models that combine language understanding with visual perception and decision outputs — designed to interpret sensor data from military platforms and support real-time tactical decisions. The layering of a ministry-wide software contract on top of an active VLA development programme with a drone manufacturer constitutes a defence-AI stack, not a set of discrete pilots.
Against that military backdrop, Airbus signed a partnership agreement with Mistral on 28 May 2026, supporting Airbus’s ambition to place cutting-edge AI at the core of its operations from initial design to on-board capabilities, spanning commercial aircraft, helicopter, defence, and space activities, under strict security and sovereignty requirements including for military aerospace applications. Under the agreement, Airbus will acquire licences for the full Mistral product suite, enabling deployment on-premises, in trusted clouds, or wherever it makes sense for Airbus and its customers.
Collaboration areas include AI systems for aircraft and spacecraft, automation of technical documentation, AI-driven engineering simulations, edge AI for object recognition to support flight safety, and defence applications such as cyber investigations and coding support in highly secure environments. The partnership also guarantees Airbus access to Mistral’s leading researchers and influence over the AI product roadmap.
The mechanism linking defence positioning to the Airbus deal is data sovereignty, not model performance alone. The deal comes as European companies grow increasingly uneasy about dependence on US-based AI providers, amid concerns over data sovereignty, security exposure, and the legal reach of American legislation over data stored on US-owned infrastructure. Mistral is a French entity without direct exposure to the US Cloud Act; that legal status, combined with its own operation of the Les Ulis data centre, constitutes a justifiable regulatory advantage for clients handling data subject to GDPR or NIS2 obligations, transposed in France in October 2024.
On 28 May, Mistral announced a new 10 MW facility at Les Ulis in the Essonne department, south of Paris, dedicated to inference operations and scheduled to open in Q3 2026. It follows the existing 40 MW training site at Bruyères-le-Châtel and forms part of a wider €4 billion compute infrastructure programme spanning France and Sweden, with planned capacity of 200 MW by 2027 and 1 GW by 2030. The March 2026 debt financing round of $830 million, raised from a consortium of seven banks, is funding data centre construction.
The infrastructure build matters because it resolves a long-standing tension in Mistral’s sovereignty pitch. Mistral used Microsoft Azure for a significant share of model training, with Mistral Large hosted on Azure AI Studio — a deployment that, even on European servers, remains subject to US legal jurisdiction, a risk the French government treated as intolerable for defence data; the AMIAD agreement circumvents this by requiring physical isolation from Azure. Own infrastructure closes that gap for classified workloads, though the commercial dual-track structure — sovereign for government, Azure for enterprise — introduces operational complexity that investors should monitor.
The second-order implication most observers are missing is the competitive disruption to established European defence and industrial software vendors. The announcement of the Airbus-Mistral partnership sent Dassault Systèmes shares reeling during the trading session — a reaction reflecting broader concern about the threat posed by AI to providers of industry-specific software solutions. The fear is clearly identified: generative AI models, capable of producing code, technical documentation, and design elements, will erode the added value and margins of legacy software suites; even if Dassault’s core PLM expertise remains unchallenged, clients may reallocate portions of their budgets toward third-party AI components.
Mistral’s May 2026 acquisition of Vienna-based Emmi AI, reported by Reuters on 19 May, sharpens that competitive threat. Emmi builds foundational models trained with the laws of physics — large engineering models that simulate complex processes such as fluid flow around an aircraft wing — which Emmi claims can run in real time versus the multi-day runtimes typical of traditional numerical solvers. More than 30 Emmi researchers and engineers joined Mistral’s Science and Applied AI teams in May 2026; financial terms were not disclosed. Those simulation capabilities land directly inside the Airbus engineering and design use cases named in the partnership.
Mistral now employs approximately 1,000 people and is targeting €1 billion in revenue for 2026. Sacra estimated the company’s annualised recurring revenue exceeded $400 million in January 2026, up from approximately $20 million in January 2025 — roughly a 20-fold expansion over twelve months. The September 2025 Series C, led by ASML with €1.3 billion for an approximately 11% stake, brought Mistral’s post-money valuation to €11.7 billion.
The revenue trajectory is real; the gap between €400 million ARR and the €1 billion target is also real, and the capex burden is intensifying simultaneously. What remains to be tested is whether the customer commitments translate into meaningful revenue — Mistral has not disclosed contract values, deployment scope, or revenue targets for its new industrial product line, and Airbus, BMW, and EDF each have substantial internal AI programmes whose relationship to Mistral’s offering has not been defined publicly.
The capital-allocation implication runs in two directions. For growth-equity and late-stage venture capital positions in European AI, Mistral’s dual industrial-plus-defence franchise — with sovereign infrastructure as the enabling layer — is the clearest European moat story in the sector, and the Airbus partnership anchors the industrial thesis with a named, credible prime. The probability that Mistral converts its current positioning into a defensible €1 billion revenue base by end-2026 would probably be higher if contract values were disclosed; absent that transparency, the primary risk to underwriting the €11.7 billion valuation is execution against capex that could rival projected top-line gains. For defensive software vendors serving European aerospace and defence — notably Dassault Systèmes — the Airbus signal raises the probability of budget reallocation toward Mistral’s stack, making short or underweight positions in legacy PLM names a plausible hedge for those already long the French AI trade.
