The Silent Economic War nobody has declared

There is a contest underway that does not appear on the front pages of the financial press, is not named in most national AI strategies, and is absent from the boardroom conversations of the majority of enterprises now deploying AI at scale. It has no declaration, no visible front line, and no formal combatants. It is being fought through investment flows, platform architecture decisions, data ownership structures, and the design of the AI systems through which an increasing proportion of global economic value will be created and distributed.

Bloor Research frames this as the Silent Economic War. The prize is global GDP — currently in the range of $105 to $117 trillion and growing. The contest is not about who builds the most capable AI tools. It is about who owns the productive digital infrastructure through which future value flows, who sets the terms on which that infrastructure is accessed, and who captures the economic output it generates.

The United States is competing through platform dominance and foundation model ownership. The major large language models, cloud infrastructure layers, and agentic AI platforms are overwhelmingly American. Access is licensed. The pricing terms are set by the owner. Every enterprise deploying these systems is renting productive infrastructure it does not own and cannot replicate. Every AI efficiency gain achieved on a licensed American platform flows, at least partially, back to the capital that owns that platform.

China is competing through sovereign stack development and parallel ecosystem export. DeepSeek, Huawei’s AI infrastructure, and state-backed model development are not purely domestic plays. They are the opening architecture of a parallel digital economic system designed to offer an alternative to US-owned infrastructure, particularly to nations who recognise the dependency risk of the current dominant model. The UK, the EU, and most enterprise organisations sit in neither of these positions: currently net consumers of AI infrastructure they do not own. The Saudi Public Investment Fund provides the clearest illustration of what a deliberate sovereign response looks like. With over $40 billion committed to AI infrastructure investment — data centres, model development, and national AI capability — Saudi Arabia is not adopting AI. It is competing for ownership of the infrastructure through which its future economic output will flow. The logic is explicit: a nation that licenses its AI capability from foreign platforms is a net importer of digital value regardless of how rapidly its GDP grows. The PIF’s AI programme is a sovereign capital deployment decision, not a technology procurement decision. That distinction is the one most national strategies and most enterprise boards have not yet made.

The three sovereign levels and why all three matter

The sovereign response to the Silent Economic War operates at three levels simultaneously, and the organisations and nations that understand all three levels will be structurally better positioned than those that address only one.

At the national level, Sovereign AI requires deliberate decisions about who owns the infrastructure through which the nation’s AI ambitions will be realised, what the fiscal architecture looks like when productive output is generated by systems in low-tax jurisdictions, and how the nation positions itself as a net exporter rather than a net consumer of digital value. Most current national AI strategies, including the UK’s AI Opportunities Action Plan, do not grapple with these questions at the level of architectural decision.

At the organisational level, sovereign intelligence architecture means owning the AI capability that creates productive output, governing the data that trains the organisation’s systems, designing the human-digital workforce architecture that makes the organisation a net contributor of digital value, and building the operating model that captures rather than transfers the gains from AI investment. Sovereign AI is not a data centre decision. It is an architecture decision — the complete intelligence architecture of the organisation across its operating model, its workforce design, its knowledge governance, and its value capture structure.

At the individual level, the sovereign digital asset is Digital Me: the body of original knowledge and professional judgement built into a form that is owned by the individual, portable across organisations, and capable of generating value independently of any single employer. Every individual whose intelligence is being extracted into AI systems they do not own is participating in the Silent Economic War on the wrong side of the ownership line.

What the Silent Economic War means for the 2028 window

The tax base consequence is the argument most national governments have not yet confronted directly. When productive output is generated by AI systems owned by entities in low-tax jurisdictions, when the gains from workforce reduction do not recirculate through human wages into consumption and payroll tax, and when the capital that owns the infrastructure is concentrated in a small number of entities and geographies, the tax base that funds public services is structurally eroded. This is not a future risk. It is a present dynamic that is accelerating.

Bloor Research identifies 2028 as the inflection threshold at which the cost of catching up begins to exceed the practical capacity to do so for most organisations and many nations. The design decisions being made now — about infrastructure ownership, knowledge architecture, workforce design, and value capture — will determine the economic landscape of the decade that follows. The window is not indefinite.

The contest for the productive digital future is underway. The organisations that name it, design their sovereign intelligence architecture, and position themselves as net contributors rather than net consumers will determine the economic landscape of the next decade. The ones that do not will find the terms of their participation set by someone else.

Methodology note

Two data points in this newsletter carry the most analytical weight and are the most subject to challenge. Their methodology is set out here for readers who wish to stress-test the numbers.

1. The 1.3 million structural employment gap (UK)
   Source data: ONS Annual Survey of Hours and Earnings (ASHE), HMRC PAYE Real Time Information data series, ONS Labour Force Survey, cross-referenced against ONS GDP series. Baseline period: 2000–2019 (pre-COVID). Methodology: Bloor Research modelling applies the historical GDP-to-employment ratio from the baseline period to current GDP levels to derive the predicted headcount, then subtracts actual payroll employment to identify the structural gap. The COVID year (2020–2021) is excluded from the structural gap calculation to isolate the technology and structural variable from the cyclical distortion. The same methodology applied to comparable EU data produces a near-identical gap-to-GDP ratio, isolating technology deployment and offshoring as the structural variables rather than UK-specific policy factors. The 1.3 million figure represents the central estimate. Bloor Research acknowledges a range of approximately 900,000 to 1.6 million depending on baseline period selection and sectoral decomposition assumptions. The structural direction of the gap is consistent across all reasonable methodology variants. The detailed working is available to Bloor Research subscribers on request.
2. The 26–65% layoff concentration in knowledge-intensive roles
   Source data: Bloor Research observation across organisations with advanced AI maturity, defined as organisations that have deployed AI at scale alongside significant offshoring and technology-led operating model transformation. Observation period: 2024–2026. Salary range in scope: £50,270 to £125,140 and above (broadly, higher-rate income tax threshold to additional-rate threshold, UK). Methodology: cross-functional analysis of published redundancy programmes (public announcements, regulatory filings, and published workforce disclosures) combined with Bloor Research primary research interviews. The 26–65% range reflects variation across sectors and organisation types rather than a single point estimate: the lower bound is observed in organisations with partial AI deployment; the upper bound in organisations where AI transformation is most advanced. The concentration direction is consistent across all organisations in the observation set. This is Bloor Research primary research. The underlying interview data is held under standard research confidentiality. The methodology is available for peer review through the Bloor Research subscription programme.

Print Article for this blog Download here