The Big Shift, Part II: How Plutonomy Is Building Its Own OS

Part I, The Big Shift: Wealth On Its Own Terms, began with a fairly simple observation: wealth behaves differently once it becomes large enough to build its own infrastructure.

Plutonomy, a term coined by then-Citi strategist Ajay Kapur in the mid-2000s, describes an economic system in which growth, consumption and asset prices are disproportionately shaped by a small concentration of wealth. Its investment logic is straightforward: when the wealthiest households account for an outsized share of spending and asset ownership, markets increasingly take their direction from those balance sheets rather than the income of the average consumer. Part I used the concept to explore what happens when that concentrated wealth also begins constructing its own financial architecture.

For much of the past two decades, wealth management followed a stable division of labour in which institutions supplied custody, liquidity, intelligence, products and access, while clients selected from the menu. The largest family offices are now redrawing that division by bringing investment capability in-house, selecting external expertise more deliberately and treating geography, currency exposure, liquidity and deal access as configurable parts of a privately directed system. At sufficient scale, access becomes something to engineer.

Part I also identified a tension inside the AI investment cycle. Software promises abundance through falling production costs, faster delivery and near-zero marginal cost of replication, while the physical system beneath it remains governed by scarcity across advanced chips, compute, grid capacity, power, land, water and capital. The model may capture the headlines, but the bottleneck has a persuasive claim on the economics.

On the day Part I was published, NVIDIA announced memoranda of understanding with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR to create financing platforms capable of mobilising more than US$500 billion into AI infrastructure over time. The capital remains to be raised and the partnerships are subject to final agreements, yet the announcement offered an unusually timely demonstration of how an investment thesis becomes an asset class: once the scarcity is widely recognised, someone begins building the financial machinery around it.

This edition follows that process. Direct ownership, co-investment and institutional platforms may all provide exposure to AI infrastructure, although they distribute information, control and economics very differently. The same pattern is also appearing in prime property and the emerging transformation economy, where value increasingly depends on the operating and access systems surrounding the physical asset.

The central question is therefore how much an investment thesis is worth once everybody shares it, and whether the greater advantage has already moved to those designing the route into the opportunity.

A Half-Trillion-Dollar Ambition

An MOU is an agreement to work towards an agreement, and the distinction between ambition and committed capital matters when the headline contains twelve zeros. Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR sit at the centre of global private capital, credit and infrastructure finance. Alongside NVIDIA, they have proposed independent platforms capable of mobilising more than US$500 billion from third-party investors over time. The capital remains to be raised and final agreements remain to be signed. The coalition itself is the signal: some of the world’s most influential capital providers are beginning to design the financing architecture around AI compute.

The more consequential development lies in what these platforms are seeking to make financeable. NVIDIA describes compute as productive infrastructure that can be deployed across customers and workloads, generate usage-linked revenue and draw demand from a broad ecosystem of developers and offtakers. Goldman Sachs extended the argument into capital markets by identifying the opportunity to create a credit market backed by NVIDIA compute.

Chips have traditionally entered the investment equation as equipment purchased by an operator. NVIDIA’s framing invites investors to treat compute capacity as an asset with utilisation, revenue, useful life and transferable value, thereby giving the physical bottleneck beneath AI a financing language of its own.

Knight Frank’s 2026 Family Office Survey had already identified data centres among the operational real-estate sectors attracting family-office interest. The report is discussing the property and physical infrastructure that house compute; NVIDIA is seeking to financialise the compute capacity operating within and across those facilities. They occupy different layers of the same system, which makes the comparison more revealing than if they were interchangeable.

A family may own the building, an infrastructure provider may control the power, NVIDIA supplies the compute architecture, customers enter usage agreements, and a financing platform packages the resulting economics for investors. Every participant can claim exposure to the AI buildout, although each is underwriting a different risk and receiving a different portion of the return. Beneath the broad label of AI infrastructure sits a capital stack whose layers are becoming investable one by one.

The Capability Question

The case for family-office participation remains strong. Knight Frank found that direct ownership appeals to families because it offers greater influence over development, risk and upside, while operational sectors such as data centres, logistics, healthcare and student accommodation can reward capital able to hold through several cycles without optimising around a conventional fund exit.

A single-family office has no external limited partners requiring distributions on a predetermined schedule, which gives it room to move across asset classes, accept a longer payback period and make decisions around one balance sheet. That duration is valuable, although its investment advantage depends on the capabilities built around it.

A family may have the capital and patience to own a data centre while lacking the technical team needed to assess power availability, cooling design, hardware obsolescence, customer concentration or development risk. Liquidity may allow a rapid commitment, yet governance may still delay approval; the investment thesis may be understood perfectly while the route to the asset remains absent. Permanent capital becomes prepared capital only when duration is matched by expertise, relationships and an executable process.

Institutional scale becomes difficult to compete with at precisely this point. Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR bring specialist underwriting, distribution, operating relationships and access to deep pools of third-party capital. Their advantage extends beyond purchasing infrastructure because they can create repeatable channels through which infrastructure is financed.

Single-family offices retain a different set of strengths through a shorter chain between conviction and commitment, alignment around a single pool of capital and the freedom to negotiate structures that fall outside a standard fund mandate. Offices that have developed genuine sector expertise can underwrite complexity directly and use external partners selectively. Where that expertise is missing, agility introduces speed into expensive mistakes just as efficiently as it accelerates good decisions.

The family-office shift described in Part I encompasses a redistribution of authorship across wealth management. Banks and asset managers continue to provide essential custody, liquidity, expertise and products, while sophisticated families are becoming more deliberate about which parts of the investment process they direct themselves.

The offices best positioned for the next phase will understand the distinction between ownership and control, exposure and access, duration and readiness. Building the largest internal team is unlikely to be the universal answer; building enough capability to negotiate from knowledge, and enough judgement to identify the right operating partners, is far more important. An investment thesis that cannot be translated into origination, underwriting and governance remains an opinion, however sophisticated the analysis behind it.

Co-investments And Selective Control

Direct ownership is often presented as a binary choice between holding the asset and holding a fund that owns it, although real markets distribute ownership and control across a much wider spectrum.

A family office might develop a project outright, form a joint venture with an operator, co-invest alongside a specialist manager, commit to a financing platform or acquire exposure through a fund. Each route reaches the same broad investment theme while allocating governance, information, fees and operating responsibility in a different way.

The label matters less than the underlying rights. An allocator needs to know who originated the opportunity, who underwrote the technology and operating risks, who controls the financing and customer contracts, who determines leverage and exit, and how much of the return remains after each layer has taken its share. Access to timely operating data may ultimately matter as much as the percentage recorded on the ownership schedule.

This is where co-investment earns its place in the family-office model. Part I noted that many offices prefer to invest alongside trusted families or specialist managers instead of leading every transaction themselves. Such arrangements can preserve direct economic participation and selected governance rights while drawing on operating capability that would be uneconomic to recreate for a single investment.

The objective is selective control. Family offices gain little from supervising a cooling system; the advantage lies in recognising which capabilities generate the return, which rights protect it, and which functions can be bought safely as a service. Being early to a theme has practical value when it improves those eventual terms. Once a packaged institutional product becomes the only realistic route, the investor may still receive exposure while the allocation, information and fee stack are controlled elsewhere.

The Same Pattern In Property

AI provides a live example because the financing machinery is being announced in real time, while prime property shows a more mature version of the same pattern.

Knight Frank describes increasingly mobile UHNWIs organising their lives across several jurisdictions and using established cities on a “dip-in, dip-out” basis. This behaviour is often compressed into a supposed 90-day tax rule, presumably because complicated legal systems become more appealing when reduced to a round number. In practice, no universal 90-day safe harbour exists.

From 6 April 2025, the UK replaced the domicile-based remittance basis with a residence-based system and introduced a four-year foreign income and gains regime for qualifying new arrivals following at least ten years of non-UK residence. UK tax residence can depend on accommodation, work, family and other ties alongside day count, while every other jurisdiction applies its own tests.

A London apartment, Milan residence or Dubai base may support a mobile lifestyle, although ownership can also create legal and fiscal connections that must be coordinated with the principal’s wider arrangements. The family office’s task is to make the collection of homes, companies, investments and personal ties function as one coherent structure across several authorities.

That operating model is changing the property product itself. Knight Frank reports that some clients who once sought US$30 million to US$50 million primary London residences are now looking closer to US$15 million for more practical boltholes, while others increasingly rent. Over the five years covered by the report, top-end rents rose by 63% in New York, 53% in London and 48% in Singapore.

Demand is also moving towards turnkey, managed and branded residences. Knight Frank forecasts more than 1,000 branded-residence schemes operating worldwide by 2030, with the category expanding beyond hotel groups into fashion and wellness brands. A principal moving between three continents has limited enthusiasm for a two-year correspondence about unfinished Mayfair plumbing, so convenience commands a premium; underneath that convenience sits something more valuable, which is confidence that security, staffing, privacy, maintenance and service will function in the owner’s absence.

The residence is consequently becoming part property and part operating platform. The brand occupies the layer between ownership and use, earning its premium by making the asset reliably accessible wherever the owner happens to be.

The Operator Premium

Across both parts of this series, the assets attracting patient private capital are consistently those that require someone to run them. Data centres, healthcare property, student accommodation, logistics, managed residences and wellness-led hospitality derive value from the interaction between physical assets, specialised services, customer relationships and recurring revenue.

Knight Frank describes a broader “transformation economy” emerging around health, longevity, belonging and purpose. Status consumption remains very much alive, the superyacht industry can relax, while younger UHNWIs increasingly expect a purchase to offer some improvement in the person consuming it, whether through a longer life, better health, privileged access or stronger belonging.

For an allocator, the relevant question concerns which businesses can translate that desire into durable economics. A longevity clinic requires clinical capability, trusted data, specialist staff and repeat engagement; a managed residence needs an operating system behind the concierge desk; a data centre depends on power, cooling, customers and continuous reinvestment. In each case, the physical asset provides the foundation while the operator converts it into a functioning business.

This makes direct ownership more demanding than the phrase suggests. Holding the property secures exposure to the physical asset, whereas capturing the wider economics requires integration across operations, service and the customer relationship. The premium accrues to those capable of maintaining that integration over time.

AI infrastructure follows the same logic. Compute may become an investable asset, yet its value will continue to depend on utilisation, power, software and customers. Financial engineering can distribute the exposure and expand the capital available to build it; the operator remains responsible for turning that capital into productive capacity.

Food for Thought

Once an investment thesis reaches consensus, recognition ceases to be its scarce component.

Part I appeared just as NVIDIA and six of the world’s leading capital providers announced plans to build financing channels through which more than half a trillion dollars of third-party capital might eventually flow. The timing was useful. It showed, in real time, how financial markets organise themselves around a newly recognised bottleneck.

First the scarcity is identified; then its risks are described, its revenues contracted, its ownership divided and its access distributed. Each stage allows more capital to reach the opportunity, while creating a new position from which someone can influence allocation, information and economics.

For family offices, the strategic choice lies in deciding which part of that stack they need to control before the structure takes shape around them. The most effective model will rarely eliminate intermediation altogether. It will retain authority over the capabilities that create advantage, combine them with specialist operating partners and purchase the remaining functions with clear knowledge of their cost.

As private and institutional capital converge on the same scarce assets, the route into the opportunity becomes part of the opportunity itself.

Source Note

This edition draws on Knight Frank, The Wealth Report 2026 (20th edition), including the 2026 Family Office Survey; NVIDIA’s announcement of 10 August 2026 concerning memoranda of understanding with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR; and current HMRC guidance on the UK’s four-year foreign income and gains regime and Statutory Residence Test. NVIDIA’s proposed financing partnerships remain subject to final agreements, and the US$500 billion figure is an ambition to mobilise third-party capital over time rather than capital already committed. Knight Frank’s forecasts are estimates rather than guarantees. Tax commentary is general information only.

DISCLAIMER Constellation is published by Nevv Capital for general information and discussion only. It is not investment, legal, tax or financial advice, nor a research report, and it is not an offer or solicitation to buy or sell any security, instrument or entitlement. References to companies, funds, technologies and projects do not constitute recommendations. Readers should conduct their own legal, regulatory, technical and investment due diligence.

© 2026 Nevv Capital. All rights reserved.