Mistral AI Statistics 2026: Valuation, Revenue, Users & Pricing

Last Updated: September 9, 2026

Mistral AI closed a €3 billion Series D on September 8, 2026, pushing its valuation to roughly €21 billion (about $24 billion) in a round led by Samsung Electronics, EQT's Scaleup Europe Fund, and PSG Equity. That's nearly double the €11.7 billion mark set a year earlier. Annualized revenue has grown from roughly $100 million in 2025 to more than $400 million in 2026, with CFO Johan Bergqvist telling Reuters the company is targeting $1 billion in ARR by year-end.

Mistral's valuation history tracks one of the steepest re-rating curves of any AI lab outside the US.

The March 2026 debt round wasn't equity. It was $830 million raised specifically to buy 13,800 Nvidia chips for a new data center near Paris, separate from the company's software business entirely.

That distinction matters. Mistral is now funding two businesses at once: a model and API business, and a physical compute business. The company has publicly described this as a shift toward becoming a "neocloud" provider, not just a model vendor, with a stated goal of bringing up to one gigawatt of data center capacity online by 2030.

For context on where that valuation sits against the rest of the frontier lab field, Anthropic's statistics and OpenAI's statistics both show valuations well into the hundreds of billions, meaning Mistral remains a fraction of either company's size despite its rapid re-rating.

Mistral's revenue trajectory is the clearest evidence its business is not just a funding story.

A jump from $100 million to a targeted $1 billion ARR inside two years is an unusually steep curve, even by frontier-lab standards, and it's happening while the company simultaneously funds a hardware buildout. That combination, revenue growth plus infrastructure investment, is closer to how Nvidia's own statistics describe compute-side AI spending than how a typical software company scales.

Le Chat is Mistral's consumer and business-facing assistant, and its adoption curve is the fastest-growing part of the company's public numbers.

Le Chat passed 1 million app downloads within 14 days of its February 2025 mobile launch, briefly overtaking ChatGPT on the French App Store. Estimates compiled from Similarweb-sourced third-party data put Le Chat at roughly 5 million monthly users as of early 2026, with the mistral.ai website itself pulling an estimated 9-11 million desktop visits a month.

The open-weight side of the business tells a separate adoption story. Mistral's models have been downloaded more than 100 million times combined, with Mixtral 8x7B alone reaching 20 million downloads in its first month after release. That kind of open-weight pull is closer to how what is generative AI describes developer-driven adoption than how closed-model vendors typically report usage.

Mistral's lineup spans five active tiers, each aimed at a different workload rather than one model trying to do everything.

Mistral Large 3 is the primary general-purpose model, handling most reasoning, writing, and analysis tasks. Mistral Medium 3.5 sits above it on price and is positioned as the highest-capability tier for harder requests. Mistral Small 4 targets everyday production traffic where speed and cost matter more than maximum reasoning depth. Codestral is tuned specifically for code generation and repository-level completion. The Ministral 3 family, spanning 3B to 14B parameters, targets edge and on-device deployments where models run locally rather than through the API.

A separate Magistral line focuses on step-by-step reasoning tasks, priced at the higher end of Mistral's catalog. Several of these models, including the Medium 3.5 flagship, ship as open weights under permissive licenses, a meaningfully different distribution strategy than the closed-weight approach used by OpenAI and Anthropic.

Mistral runs two separate pricing systems: pay-per-token API access through "la Plateforme," and flat monthly subscriptions for Le Chat.

Le Chat's consumer plans run entirely separately: free, $14.99/month for Pro, and $24.99 per user/month for Team, according to pricing data compiled by CloudZero.

The gap versus US rivals on the flagship tier is the headline number. Mistral Large 3, at $0.50/$1.50 per million tokens, runs roughly 80% cheaper on input and 90% cheaper on output than GPT-5.4's $2.50/$15.00 rate, and 83% cheaper on input against Claude Sonnet 4.6's $3.00/$15.00 pricing. That gap is a large part of why Mistral has become the reference point for cost-sensitive workloads in AI Business Weekly's pricing guide.

Several Mistral models, including the Medium 3.5 flagship, ship as open weights, meaning teams processing tens of millions of tokens a day can self-host instead of paying per token at all, converting the cost into GPU compute rather than API fees.

Enterprise adoption is the clearest signal that Mistral's growth isn't purely retail download numbers.

Mistral's enterprise pitch centers on European data sovereignty: GDPR-aligned handling, open-weight models regulated buyers can inspect, and the option to run entirely on EU infrastructure. That pitch is the same wedge described in how to implement AI in business for organizations weighing vendor lock-in against regulatory exposure. ASML's position as Mistral's largest outside shareholder, taken in the September 2025 round, reinforces that this is treated as much like industrial policy in Europe as it is a venture bet.

Geography data shows Mistral's usage is still heavily concentrated at home. France accounts for roughly 35% of website traffic, close to 3.3 million visits a month, well ahead of any other single market, based on Semrush data cited in third-party analysis. Desktop traffic outweighs mobile by a wide margin, at roughly 79% to 21%, and direct visits (people typing the URL or using a bookmark rather than arriving from search or social) account for about 73% of total traffic.

Demographic estimates from Similarweb put Mistral's user base at roughly 59% male and 41% female as of February 2026, with the 25-34 age group representing the largest single segment at nearly 34%. That skews slightly younger than the typical enterprise software buyer, consistent with Le Chat's consumer app downloads driving a meaningful share of overall traffic alongside the enterprise API business.

Mistral's $830 million debt-financed purchase of 13,800 Nvidia chips signals a deliberate move away from renting compute from hyperscalers and toward owning it outright.

That's a meaningfully different capital structure than pure-software AI labs. Combined with the neocloud pivot announced alongside the Series D, Mistral is now competing on two fronts simultaneously: model quality against OpenAI, Anthropic, and Google, and infrastructure ownership against the cloud providers those same labs currently rent from. For a full breakdown of how compute spending is trending across the sector, AI Business Weekly's AI spending statistics tracks the broader infrastructure buildout Mistral's bet sits inside.

What Is Mistral AI?
The complete explainer on Mistral's model lineup, founding story, and open-weight strategy.

AI Market Share 2026
Where Mistral ranks against OpenAI, Anthropic, Google, and Chinese labs in overall market position.

AI Pricing Guide 2026
Full per-token cost comparison across every major model provider, including Mistral's full lineup.

Anthropic Statistics
How Anthropic's valuation, revenue, and enterprise adoption compare to Mistral's.

Generative AI Market Statistics
The market-wide numbers behind the growth Mistral is capturing a share of.

How much is Mistral AI worth in 2026?
Mistral AI's most recent valuation is approximately €21 billion (around $24 billion), set by a €3 billion Series D that closed on September 8, 2026 and was led by Samsung Electronics alongside EQT's Scaleup Europe Fund and PSG Equity. That's roughly double the €11.7 billion valuation Mistral carried after its September 2025 Series C, which was led by ASML.

Who owns Mistral AI?
Mistral AI is a private company founded by CEO Arthur Mensch and headquartered in Paris. Its investor base includes ASML, which took a roughly 11% stake and became the largest outside shareholder in the 2025 Series C, along with General Catalyst, Andreessen Horowitz, Lightspeed Venture Partners, and, as of the September 2026 round, Samsung Electronics.

How much revenue does Mistral AI make?
Mistral's annualized revenue grew from roughly $100 million in 2025 to more than $400 million in 2026. CFO Johan Bergqvist told Reuters the company is targeting $1 billion in annual recurring revenue by the end of 2026, a target that would roughly 2.5x the company's reported mid-year run rate in a single year.

How many users does Le Chat have?
Le Chat reached 1 million downloads within 14 days of its February 2025 launch, and third-party estimates put monthly active usage at roughly 5 million users by early 2026. The mistral.ai website separately draws an estimated 9 to 11 million desktop visits per month.

Is Mistral AI cheaper than ChatGPT and Claude?
Yes, on a per-token basis Mistral's flagship Large 3 model is substantially cheaper than comparable models from OpenAI and Anthropic. At $0.50 input and $1.50 output per million tokens, Large 3 runs about 80% cheaper on input than GPT-5.4 and roughly 83% cheaper on input than Claude Sonnet 4.6, based on pricing data tracked by CloudZero's Mistral pricing breakdown.

What is Mistral AI's business model?
Mistral generates revenue from three sources: a pay-per-token developer API through "la Plateforme," Le Chat's consumer and team subscriptions (free, $14.99/month Pro, and $24.99 per user/month Team tiers), and enterprise or sovereign on-prem deployments. Many of its models ship as open weights specifically to drive adoption, with monetization concentrated in the paid API and enterprise contracts layered on top.

How many enterprise customers does Mistral AI have?
Mistral reported more than 125 enterprise customers as of its September 2026 Series D announcement, building on an earlier July 2025 disclosure of 1,031 high-value customers and roughly 450,000 total customers across all tiers. The Financial Times has reported that roughly 60% of Mistral's revenue comes from European customers specifically.

Why is Mistral AI buying its own data centers instead of renting compute?
Mistral raised $830 million in debt financing in March 2026 specifically to purchase 13,800 Nvidia chips for a new data center near Paris, part of a broader "neocloud" strategy announced alongside its September 2026 Series D. The company has stated a goal of bringing up to one gigawatt of data center capacity online by 2030, positioning itself to compete on infrastructure ownership, not just model quality, against labs that currently rent compute from major cloud providers.

Mistral's September 2026 numbers describe a company scaling on two fronts at once: software revenue growing roughly 4x in a year, and a hardware buildout large enough to require its own debt financing. The valuation jump to roughly $24 billion reflects investor confidence in that dual bet, not just model performance. Whether Mistral hits its stated $1 billion ARR target by year-end will be the clearest test yet of whether Europe's frontier AI champion can convert open-weight adoption and aggressive pricing into revenue at the pace its capital raises assume.