Mickai is a sovereign intelligence operating system that regulated businesses own and run inside their own walls. It runs on the customer's own hardware, on premises and air gapped, with zero data egress. That single design decision is the reason we can pursue two buyers at once, and the reason those two buyers do not compete for the same dollar.
Two buyers, one stack
The dual-buyer thesis is straightforward. We sell sovereign AI to the regulated firms that the public cloud cannot lawfully reach. And we license the patented stack to the platforms that want to reach those same firms. The first motion is a direct sale of a system a bank or a hospital owns outright. The second is a licence to a large platform that wants a sovereign, on premises layer it does not have today.
These motions do not cannibalise each other because they solve different problems for different parties. The regulated buyer needs a system that keeps data inside the building and produces evidence a supervisor will accept. The platform needs a way to serve that buyer without asking them to send a single byte to a shared cloud. We supply both from the same architecture.
Why the public cloud cannot follow
A large share of the economy is not choosing to avoid public cloud AI. It is legally prevented from using it. Around 0.85 million UK businesses, roughly 15 per cent, and around 5 million across the EU cannot place regulated workloads on shared public infrastructure. The constraints are specific and they are enforced: PRA SS1/23 on outsourcing and third party risk, UK GDPR special category data, the NHS Data Security and Protection Toolkit, EU AI Act high risk obligations, ITAR and EAR export controls, the NIS Regulations, and the extraterritorial reach of the US CLOUD Act.
Each of these is a hard boundary, not a preference to be negotiated. A hospital cannot move special category patient data into a shared environment because a model is convenient. A defence supplier cannot let export controlled designs cross a jurisdiction because an API is fast. The public cloud model, where data travels to the model, is structurally on the wrong side of these rules. No amount of encryption in transit changes where the data comes to rest and who can compel access to it.
So there is a market defined by exclusion. Sovereign AI is estimated at around USD 40 billion in 2025, rising to around USD 148 billion by 2032. That growth is not driven by novelty. It is driven by firms that need modern AI and cannot lawfully consume it the way it is sold today.
What we actually ship
The system runs where the data already lives. Around fifty specialist models sit under a deterministic arbiter, so the routing of a request and the selection of a model follow fixed rules rather than probabilistic guesses. That determinism matters to a regulator, because a system whose behaviour can be reproduced is a system that can be audited.
Every action is written to the Open Audit Record. The OAR signs each action under FIPS 204 ML-DSA-65, a post-quantum signature scheme, and hash-chains it into a tamper-evident ledger that can be verified offline. A supervisor does not have to trust our word or our uptime. They can take the ledger, verify the chain on their own equipment with no connection to us, and see that nothing was altered after the fact.
Pantheon, a post-quantum Layer 1 currently on testnet, extends this to multi-node attestation, so that a fleet of sovereign installations can agree on a shared record without any of them surrendering control of their data. The whole design assumes the network is hostile and the auditor is sceptical, because in regulated environments both assumptions are correct.
An ally to the platforms, not a rival
It would be easy to frame this as a fight with the AI majors. It is not, and framing it that way would misread the opportunity. A platform that adds a sovereign, on premises layer instantly reaches a regulated market it cannot serve today. We are the layer that makes that reach possible, which makes us an ally rather than a competitor.
Our internal analysis maps 196 companies and 311 patent-company pairs as potential licensees. The names in that set include Microsoft, AWS, NVIDIA, Google, Adobe and IBM. We want to be clear about what that figure is and is not. It is potential-licensee sizing based on where our patented methods intersect with published product surfaces. It is not a signed book of business, and it is not an accusation of infringement. It is a map of where a licensing conversation would create value for both sides.
For a platform, the logic is simple. The regulated market is real, it is large, and it is currently unreachable through a shared cloud. Building a sovereign, offline, auditable layer from scratch is a multi year undertaking against a patent position that already exists. Licensing a layer that is already built, and already carries the evidence a supervisor expects, turns an unreachable market into a served one. The value of that is implicit in every one of those 311 pairs.
The patent position
The stack is protected by 104 filed UK patent applications, roughly 2,340 claims across 13 invention families, owned by Mickai LTD, with the named inventor Mickarle Sean Junior Wagstaff-Irons. These are filed applications, not granted patents, and we describe them as filed throughout. The families cover the arbiter, the audit record, the attestation layer and the methods that hold them together as a system rather than a collection of parts.
A patent position is what lets both halves of the thesis stand up. It gives the regulated buyer confidence that the system they own is built on defensible ground. And it gives the platform a reason to license rather than reimplement, because the map of 311 patent-company pairs describes exactly where reimplementation runs into prior work.
Where we are as a company
Mickai LTD is a UK company, Companies House number 17166618, with manufacturing secured in Birmingham and led by founder and chief executive Micky Irons. The system is built and live, running on customer hardware rather than promised for a future release. The Birmingham manufacturing base means the hardware and the software ship together as an owned appliance, which is what a firm under SS1/23 or ITAR actually needs, not a subscription to someone else's estate.
We are speaking with a small number of aligned partners as we scale. The point that matters for a buyer, and for a platform reading over that buyer's shoulder, is that the capability exists now and the market it serves is defined by law rather than by fashion. The regulated firms that cannot use public cloud AI are not waiting for permission to want modern systems. They are waiting for a system they are allowed to run.
Frequently asked questions
Does selling to regulated firms and licensing to platforms create a conflict?
No. The regulated firm buys a system it owns and runs inside its own walls. The platform licenses the patented stack to reach that same regulated market through its own products. The two motions serve different needs from the same architecture, and neither undercuts the other. A platform licensing the layer expands the total reach of sovereign AI rather than competing for our direct customers.
How can a regulator trust the system if it runs offline and disconnected?
Because the trust does not depend on a connection to us. Every action is signed under FIPS 204 ML-DSA-65 and hash-chained into the Open Audit Record, a tamper-evident ledger. A supervisor can take that ledger and verify it offline on their own equipment. Determinism in the arbiter means the same input produces the same, reproducible behaviour, which is the property an audit relies on.
What does the figure of 196 companies and 311 patent-company pairs represent?
It is potential-licensee sizing produced by our internal analysis, mapping where our patented methods intersect with published platform products. It is not a signed book of business and it is not a claim that any named company infringes. It is a view of where a licensing conversation would add value, given a patent position of 104 filed UK applications and roughly 2,340 claims.
Top comments (0)