By the end of this article, you’ll have a formula to measure the strength of your moat and the competitors’. 

As an example, I’m looking at two French giants. They target the same market, but their trajectories have diverged. One is fighting through a massive restructuring; the other is valued at $13.8 billion.

Here is the formula that explains why one is collapsing while the other is winning:

Moat = (Your Unique Output) × Speed / Cost

Context

Atos is the incumbent. They provide enterprise IT for European governments and banks. Their world is built on massive physical infrastructure and long-term contracts.

Mistral is the new company, founded in April 2023, they provide enterprise AI for European governments and banks.

Not every company can apply the formula. This is made for the industries that are going through a renaissance. The examples below clarify how to apply the formula to your case.  

1. Atos

Atos operates on a linear model. To get more, you must spend more.

• Insight (Static): istheir unique output.  A server stores data but doesn’t interpret it. To get an “insight,” Atos has to send a human consultant.

• Speed (Low): Building infrastructure takes months.

• Cost (Extreme): In 2023, Atos spent €562 million just on capex and leases.

The Result: Atos’s moat is 

1(static data is equivalent to a constant number)×Low/Very High. 

The numerator is stuck while the denominator compounds. Their share price reflects this, having dropped over 90% before the 2025 reverse split.

2. Mistral: The Ratio Shift

Mistral operates on an exponential model. Every unit of progress makes the next unit cheaper.

• Insight (Active): Their models reason (their unique output). They find patterns in seconds that would take an Atos consulting team weeks to find.

• Speed (Instant): Mistral Large 3 is live and callable via API. A client goes from “Hello” to “Inference” in days.

• Cost (Falling): Mistral Large 3 output costs roughly $7 per million tokens, significantly lower than GPT-5 or Claude Opus. As models get more efficient, their margins expand automatically.

The Result: Mistral’s moat is High×High/Low. By early 2026, their annualized revenue run-rate went past $400 million. They captured the very clients Atos used to own.

The Shift Across Industries

This is going beyond the data companies here are different scenarios

Deep-Sea Mining

• The Old Moat: volume / very high capital.

• Incumbents focus on tonnage. To grow, they build more ships and hire more crew.

• The Shift: precision /Cost (low and improving)

• New players use Autonomous Underwater Vehicles (AUVs) to map the seafloor.

Earth Observation

• The Old Moat: coverage / very high launch cost.

• Incumbents sell pixels. You buy the image, then you hire someone to look at it.

• The Shift: Marginal CostDetected Events×Real-time

• New entrants run AI directly on the satellite. They don’t sell an image of a port; they sell the “Decision”—a ship has changed course. The client pays for the answer, not the photography.

• detected events × real-time / marginal cost per inference.

Why it matters for your pitch

This is a useful way to communicate your competitive advantage to investors, a new way that beats the classic comparison table

Reference List

Atos. (2024, March 26). Atos reports full year 2023 results. [Press release]. https://www.atosgroup.com/en/press/atos-reports-full-year-2023-results

FutureSearch AI. (2025). DeepResearch bench: A comprehensive benchmark for deep research agents. https://futuresearch.ai/docs/case-studies/deep-research-bench-pareto-analysis/

Gans, J. S. (2025). Growth in AI knowledge (Working Paper No. 33907). National Bureau of Economic Research. https://www.nber.org/system/files/working_papers/w33907/w33907.pdf

Mistral AI. (n.d.). Pricing. Retrieved April 25, 2026, from https://mistral.ai/pricing/