TL;DR
Thorsten Meyer AI has challenged its earlier support for owning the full AI stack, arguing that most companies gain more from using the strongest available model with vendor fallbacks. It says sovereign infrastructure remains justified for organizations facing binding legal, security or data-residency restrictions.
Thorsten Meyer AI has reversed much of its recent case for owning AI infrastructure, arguing that most organizations should use the best-performing model available and rely on routing and fallback systems for resilience. The July 16 analysis limits the case for full AI sovereignty largely to buyers constrained by law, classified workloads or strict data rules.
The publication said five weeks of reporting had repeatedly favored model ownership, domestic infrastructure and protection against foreign control. Its new analysis deliberately tests that position, concluding that sovereignty can impose a large capability penalty, higher operating costs and delays that prevent companies from releasing products while competitors move ahead.
Its comparison cites self-reported benchmark results showing Inkling at 77.6% on SWE-bench against 95.0% for Fable 5, with a wider gap on Terminal-Bench. The publication cautions that those figures, drawn from vendor tables and Artificial Analysis, are awaiting independent replication. They support its argument but do not establish that every sovereign model will underperform on every workload.
The analysis also points to the cost of certification, staffing and idle computing capacity. It says SecNumCloud qualification can cost far more than ISO 27001 compliance and cites estimates of $75,000 to $100,000 per year for specialist staff. Those figures come from the publication’s earlier reporting and named industry sources; the supplied material does not include the underlying calculations.
Against sovereignty: the strongest case for just using the best model
This publication has spent five weeks arguing one thing — and every piece converged. That should bother you. It bothers me. When eight analyses reach the same verdict, you’re not running an analysis. You’re running a thesis, and the evidence has started arriving pre-sorted.
So here’s the case against — argued properly, with the same evidence, turned around. Not a strawman erected to be knocked down. The version a smart CTO would put to me across a table, and which I have not yet answered in public. The claim: for almost everyone, sovereignty is an expensive hedge against a risk they’ve mispriced — and the rational move is to use the best model and get on with it.
Defence · classified · national health data · DORA-bound finance. The foreign-legal-order risk isn’t theoretical and isn’t insurable by other means — it’s a legal gate. No benchmark opens it. Your alternative isn’t a worse model; it’s no deployment at all.
Statistically, you are. You have a reasonable, politically legible, entirely unbudgeted feeling — and an industry built to monetize it. The capability compounds, the tax is real, the opportunity cost is brutal, and 18 days is survivable.
I’ve spent five weeks arguing you should own your stack. The strongest case against says: for most of you, that’s an expensive way to be worse, sold by people whose real product is a feeling. And that case is mostly right. What survives is smaller and sharper — everything above the router line (the qualification programme, the owned cluster, the custom pre-training run, the €11B data centre) you should buy only if a law requires it, never because a narrative does. A router is the sovereignty most people actually need. 90% of the resilience for ~2% of the cost — and it would have made 12 June a non-event. So run the honest test: are you bound, or are you performing?
Model Quality Versus Legal Control
The change matters because companies are deciding whether to buy leading commercial models, host open models themselves or fund domestic alternatives. Choosing sovereignty without a binding requirement can mean lower model performance, slower product delivery and higher costs. Choosing a foreign service can leave an organization exposed to vendor withdrawal, policy changes or another country’s legal authority.
Thorsten Meyer AI draws a line between organizations that are legally bound and those seeking general resilience. It places defence, classified work, national health data and some DORA-covered finance in the first group. For these users, the publication argues that sovereignty is a deployment requirement, not an optional insurance policy.
As an affiliate, we earn on qualifying purchases.
Five Weeks of Sovereignty Arguments
The reassessment follows eight analyses in which the publication examined model ownership, European providers, infrastructure spending, shareholder control and the risk that a foreign government or supplier could withdraw access. The author said the repeated result had begun to resemble a fixed thesis rather than an open evaluation of the evidence.
The strongest counterexample, according to the publication, was a service restriction that allegedly ran from June 12 to July 1. It says a Commerce directive removed access to Fable 5 and Mythos 5 for 18 days, while alternatives remained available. The publication interprets that episode as evidence that multi-model routing may address most companies’ continuity risks at a fraction of the cost of owning infrastructure.
“For almost everyone, sovereignty is an expensive hedge against a risk they have mispriced.”
— Thorsten Meyer AI, Reality Check AI Dispatch
Cost and Benchmark Evidence Unverified
Several central claims remain unconfirmed outside the supplied analysis. The benchmark results are self-reported, the cost comparisons rely on earlier reporting, and the circumstances behind the cited 18-day service restriction are not documented in the source material provided here. It is also unclear whether fallback models delivered comparable accuracy, security and capacity during that period.
The line between regulated and unregulated buyers may also be less precise than the analysis suggests. Contract terms, sector rules, customer promises and future export controls can create different exposure by country. A routing service reduces dependence on one vendor, but it does not remove every legal or supply-chain risk.
Buyers Must Test Their Constraints
Organizations now face a workload-by-workload decision: identify any binding legal restrictions, measure the performance gap between available models and test whether provider switching preserves service during an outage. Buyers without a legal barrier can compare a multi-provider router against the cost of self-hosting.
Independent replication of the benchmark figures and fuller documentation of the June service restriction would help test the publication’s conclusion. Regulated buyers will also need to watch European certification rules, export controls and vendor ownership changes before committing to long-term infrastructure.
Key Questions
Is Thorsten Meyer AI abandoning AI sovereignty?
No. It now supports sovereign systems for legally constrained workloads while arguing that most other organizations should prioritize model capability and provider redundancy.
What does owning the AI stack mean?
It generally means controlling the model weights, computing infrastructure and deployment environment instead of depending entirely on an external API provider.
What is the proposed alternative to full sovereignty?
The analysis recommends a routing layer with multiple model providers. If one service becomes unavailable, traffic can be directed to another model.
Which organizations may still need sovereign AI?
The publication identifies defence, classified systems, national health data and some regulated financial workloads as cases where foreign control or data handling rules can block deployment.
Are the performance and cost figures independently confirmed?
Not from the supplied material. The benchmarks are described as self-reported and awaiting replication, while the cost figures come from the publication’s prior reporting and cited industry sources.
Source: Thorsten Meyer AI
