How Mistral built a billing stack that keeps pace with frontier AI

From token-based APIs to subscriptions and enterprise contracts, Mistral uses Lago at the center of a billing environment spanning payments, CRM, analytics and accounting.

Company background

Mistral AI builds frontier models, products and infrastructure for millions of users and more than 100 large enterprises. Customers including ASML, Airbus, Cisco, HSBC, Stellantis and CMA CGM use its technology across semiconductors, aerospace, financial services, automotive and global logistics.

That adoption has translated into rapid commercial growth. In February 2026, CEO Arthur Mensch said Mistral's annualized revenue run rate had surpassed $400 million, up from $20 million a year earlier.

In September 2026, Mistral announced a €3 billion Series D led by Samsung Electronics, with Scaleup Europe Fund and PSG Equity as co-leads, at a post-money valuation of more than €21 billion. Its ambition extends from software to infrastructure: Mistral says Compute will offer access to tens of thousands of GPUs, while its longer-term European roadmap targets up to one gigawatt of AI capacity by 2030.

The billing challenge

Before launching its paid API, Mistral needed a billing system that could match the way AI products are actually consumed. Conventional seat-based subscriptions were not enough: usage and price can vary by model, token type and customer agreement.

Mistral's engineers were focused on building models and products. Building a billing platform alongside them would have meant maintaining metering, pricing, credits, invoices, payments, tax logic and finance integrations internally.

The team chose Lago for an API-first, open-source billing foundation that could support:

• Token-based and usage-based pricing
• Usage aggregation by model and token type
• Prepaid credits and automatic wallet top-ups
• Payment collection through multiple Stripe accounts
• European VAT management
• Self-serve and negotiated customer contracts

Turning AI usage into billable charges

Mistral sends usage events to Lago. Lago's billable metrics and filters then aggregate that usage using dimensions such as the model and whether tokens are inputs or outputs.

This keeps product usage separate from pricing. Mistral can introduce a model or change how a product is priced without rebuilding the underlying billing workflow for every release.

Lago calculates the resulting charges and generates invoices. The detailed usage behind those charges remains available to the teams operating and reviewing billing.

Combining prepaid usage, payments and tax

Prepaid wallets give customers a way to fund usage in advance. Lago tracks the credit balance and supports automatic top-ups when configured, while Mistral remains responsible for the product experience and access controls around that balance.

Payments remain modular. Lago connects billing and invoicing to Stripe, including Mistral's use of multiple Stripe accounts, without making the payment processor the source of billing logic.

Lago also supports the European VAT requirements attached to those transactions, keeping pricing, invoicing and tax in the same billing workflow.

Supporting self-serve and enterprise billing

Mistral can manage standard billing flows through Lago's API while using the interface for custom or negotiated arrangements. Its Salesforce integration connects billing with the commercial context used for larger customers.

This division matters as Mistral's business expands. Product and engineering teams can keep shipping, while finance and operations retain control over customer-specific billing and invoices.

A billing stack built for product and finance

Mistral's billing environment spans the systems its product, commercial and finance teams rely on:

• Mistral's own systems remain the source of product usage.
• Lago aggregates usage, applies pricing, manages subscriptions and prepaid credits, and generates invoices and credit notes.
• Stripe processes payments through Mistral's connected payment accounts.
• Salesforce is used for customer and contract context.
• SAP S/4HANA supports accounting and financial close.
• BigQuery is used for billing analysis and revenue reporting.

This architecture gives each system a clear job. Mistral keeps control of the product experience, Lago provides the shared billing layer, and the wider finance stack can consume its outputs without forcing product engineers to maintain a second billing platform.

More than software: an engineering and financial operations partnership

Lago's delivery model is forward-deployed: Mistral works with a dedicated Solutions Engineer who owns the account and coordinates directly with Lago's product, engineering and operations teams.

On the engineering side, the teams work together on the details that determine whether billing works in production: pricing configurations, payment flows, data integrations and billing migrations.

On the financial operations side, Lago works with Mistral on invoicing and credit-note workflows, reporting requirements, enterprise billing and the handoff into accounting systems.

This partnership creates a direct feedback loop between Mistral's product roadmap, finance requirements and Lago's own roadmap. Instead of handing Mistral a generic support queue, Lago brings the people who design and operate the platform into the room.

Billing that can change with the product

Mistral's portfolio now extends well beyond the paid API for which it first selected Lago. The company offers AI products for developers and enterprises as well as a growing compute platform.

The original architectural choice remains relevant: keep billing programmable and independent from product code, so new models, pricing structures and commercial motions do not require a new billing system—or another internal team to maintain one.

“Lago has been able to follow the pace of our releases and has allowed us to focus on what we do best.”

— Timothée Lacroix, CTO, Mistral AI

Sources

• Mistral AI: €3 billion Series D at a post-money valuation of more than €21 billion
• Mistral AI: Mistral Compute and a platform serving millions of users
• Mistral AI: regional inference and European compute roadmap
• Mistral AI customer stories
• Financial Times interview: Mistral's annualized revenue run rate exceeds $400 million