Licensing the Moat: How 104 Filed Patents Map to 196 Potential Licensees in Sovereign AI
Mickai's filed patent estate maps to 196 companies and 311 patent-company pairs, framed strictly as potential-licensee sizing and prior-art leverage across the largest names in compute.
The Question Behind the Patents
Most patent stories get told as defence. This one is told as economics. The question I am asked is simple: what is a sovereign AI patent estate actually worth, and to whom? Worth is a function of who needs the ground you already stand on. So this piece is not about litigation or fees. It is about mapping. Specifically, how 104 filed UK patent applications, carrying roughly 2,340 claims, map onto 196 companies and 311 patent-company pairs as potential licensees.
Those numbers are sizing, not contracts. No deal is signed and no fee is named. What I am describing is the shape of an addressable surface: the set of companies whose products plausibly intersect with claims Mickai has filed, owned by Mickai LTD, with myself as named inventor. Filed, not granted. That distinction matters, and I will come back to it, because the moat is built on priority and prior art, not on a monopoly we do not yet hold.
What Mickai Is, So the Estate Makes Sense
Mickai is a sovereign AI operating system. A SIOS. It is AI that regulated businesses own and run inside their own walls, on-premise and air-gapped, with every action written to a tamper-evident, post-quantum-signed audit record we call the OAR. It is built and it is live.
The estate is not abstract IP filed in the hope of a future product. It describes the architecture of a system that exists, runs, and audits itself. That is why it maps so widely. Sovereign deployment, tamper-evident audit, post-quantum signing of agent actions, and on-premise model orchestration are not niche concerns. They are the exact capabilities the largest AI platforms are now racing to add for regulated customers they currently cannot serve.
How 104 Patents Become 311 Pairs
A patent-company pair is a single relationship: one filed application whose claims read onto one company's product or roadmap. A single patent can map to several companies, and a single company can intersect with several patents. That is how 104 applications expand into 311 pairs across 196 distinct companies.
The clustering is not random. The mapping concentrates around six names that already define the compute and enterprise-AI stack: Microsoft, AWS, NVIDIA, Google, Adobe and IBM. These are the platforms building agentic systems, audit tooling, on-premise inference, and compliance layers for regulated buyers. Where their roadmaps move toward sovereign, auditable, air-gapped AI, they move toward ground Mickai has already filed on.
To be precise about the framing: this is potential-licensee sizing. It estimates how many companies operate in the claim space, not how many have agreed to anything. The value of that estimate is strategic. It tells you the surface area of the moat.
Why Filed, Not Granted, Is the Point
"Filed not granted" is sometimes read as weakness. In a fast-moving field it is the opposite. A filing date establishes priority. From the moment of filing, the disclosed material becomes prior art against everyone who comes later. So even before grant, 104 applications across roughly 2,340 claims do two things at once. They reserve a path to enforceable rights, and they raise the prior-art floor for every competitor working the same problems.
For a hyperscaler building sovereign and auditable AI features, that prior-art floor is the consideration. It shapes what they can freely build, what they may need to design around, and where a licensing relationship is the cleaner path. None of that requires a granted patent today. It requires a credible, dated, broad disclosure, which is exactly what an estate of this size and specificity provides.
The Wedge That Makes the Moat Bankable
A patent estate is only as valuable as the market it gates. Here the market is sovereign AI, and the wedge is regulation. Roughly 0.85 million UK businesses, around 15 percent of the total, and an estimated 5 million across the EU are legally constrained from putting sensitive workloads on public-cloud AI. The constraints are not preferences. They are PRA SS2/21, UK GDPR special-category handling, the NHS DSP Toolkit, the EU AI Act high-risk regime, ITAR and EAR, the NIS Regulations, and the CLOUD Act.
These rules describe the buyer that public-cloud AI structurally cannot reach without an architecture like Mickai's. The sovereign AI market sits at roughly USD 40 billion in 2025 and is projected toward USD 148 billion by 2032. The patent estate maps onto the precise mechanisms that let a regulated enterprise adopt AI at all inside that market. That is what makes the licensing surface bankable rather than theoretical.
The Dual-Buyer Thesis
Posture matters here, because it is the strategic heart of this. Mickai is an ally, not an OpenAI-killer. The estate supports a dual-buyer thesis. On one side, regulated enterprises license and deploy the SIOS directly to serve customers the public cloud cannot. On the other, the platform companies themselves are potential licensees of the underlying IP, because the capabilities they want to ship to regulated buyers intersect with what Mickai has filed.
Both buyers are real, and they reinforce each other. Every regulated deployment proves the architecture in production. Every proof point sharpens the relevance of the claims to a platform that wants the same capability at scale. The estate is the connective tissue between a live product and a category the largest companies in the world have an obvious reason to want inside their walls.
The Studios That Operationalise It
The IP is not a filing cabinet. It runs through Greek-named Studios that turn the architecture into regulated workflows: Nemesis for fraud and AML, Plutus for finance, Tyche for underwriting, Prometheus for forecasting, Iris for customer service, Nomos for compliance, Astraea for legal, Panacea for clinical, Pythia for business intelligence, and Aletheia for audit. Around them sit Trust Agent, the AMT agentic marketing team, Vinis voice, OAR-as-a-Service, and HELIOS hardware. Each Studio is a concrete instance of the claims operating against a real regulatory surface, which is what makes the patent-to-product mapping legible rather than aspirational.
Momentum and Where This Sits
Substance comes first, but momentum is a signal worth one line. As of June 2026, I was ranked number four on Crunchbase, with the Mickai company in the top one to two percent globally. I treat that as a dated, third-party read on attention, nothing more. The work underneath it is the point: a live SIOS, a 104-application estate, Birmingham manufacturing secured, and a revenue path that scales to billions at high gross margin as the regulated wedge is converted.
The economics close the loop. A live product, a defensible and broad IP position, and a dual-buyer market combine into a category a hyperscaler would have a rational reason to want to own. The IP estate and the dual-buyer thesis are what underwrite the enterprise value. We are building to scale and heading for the top.
Where This Sits
We are choosing partners rather than chasing them. The selection criterion is strategic fit with a sovereign, auditable, regulated-AI category that is still early and moving fast. If you operate in compute, enterprise AI, regulated industries, or the IP that gates them, and the mapping in this article matches your roadmap, the right conversation is a direct one.
Reach me at micky@mickai.co.uk.
By Micky Irons, founder and CEO of Mickai.
Frequently asked questions
What does it mean that the patents are filed, not granted?
Filing establishes a priority date. From that date the disclosed material becomes prior art against later entrants, reserving a path to enforceable rights and raising the prior-art floor for competitors working the same problems. It is leverage that exists before grant.
What are 311 patent-company pairs?
A patent-company pair is one filed application whose claims read onto one company's product or roadmap. Because a single patent can intersect several companies and a single company several patents, 104 applications expand to 311 pairs across 196 distinct companies. This is potential-licensee sizing, not signed agreements.
Why are Microsoft, AWS, NVIDIA, Google, Adobe and IBM named?
They define the compute and enterprise-AI stack and are building agentic systems, audit tooling, on-premise inference, and compliance layers for regulated buyers. Where their roadmaps move toward sovereign, auditable, air-gapped AI, they move toward ground Mickai has already filed on.
What is the dual-buyer thesis?
Regulated enterprises license and deploy the SIOS directly to serve customers the public cloud cannot, while platform companies are potential licensees of the underlying IP. Each reinforces the other: deployments prove the architecture, and proof points sharpen the relevance of the claims at platform scale.
How big is the sovereign AI market?
It sits at roughly USD 40 billion in 2025 and is projected toward USD 148 billion by 2032. The regulated wedge spans roughly 0.85 million UK businesses and an estimated 5 million across the EU constrained from public-cloud AI by rules such as PRA SS2/21, UK GDPR, the NHS DSP Toolkit, the EU AI Act, ITAR and EAR, the NIS Regulations, and the CLOUD Act.