DALIO 2.0

Executive summary

The work in one minute: Dalio, Tetlock, Marcucci

The starting point — Ray Dalio. In his 2021 book, Dalio reads the rise and fragility of great powers through debt, productive capacity, innovation, wealth distribution and conflict. The US–China rivalry makes his "Big Cycle" topical: world orders change when these forces shift. The model does not imply an inevitable succession.^[1][51]

The method — Philip Tetlock. A grand historical explanation becomes more useful when it is broken down into verifiable questions, probabilities and conditions that could refute it. Judgements should be recorded before events and revised as new evidence arrives.^[2]

The extension — Gianpaolo Marcucci. This work separates monetary, industrial, technological, energy and military power; it considers multipolarity, AI as a modifier of the cycle, and digital finance. From these it derives five scenarios, implications for Italy and for investment, instruments and an update register. The aim is to understand which forces change the future and how to adapt decisions accordingly.

The report sets out the initial framework; the Scenario Observatory preserves the reconstruction from 2022 onwards and adds the subsequent quarterly revisions. The reference to 2021 indicates the year of the book, whereas the register of our model begins on 12 September 2026.

The strategic conclusion

The central conclusion is a persistent but less exclusive American leadership within a system of distributed power. China can expand its industrial and technological role without becoming the principal issuer of international financial assets. Europe can increase its autonomy in some segments without achieving full self-sufficiency.

The most rigid reading of the Big Cycle overlooks this decomposition. It would be wrong, however, to attribute to Dalio an automatism whereby China must inevitably replace the United States: in his framework, education, innovation and political decisions are already decisive. The proposed extension concerns how those forces are measured and turned into decisions that can be updated.^[1]

Three pieces of evidence fix the starting point. In 2025 foreign-exchange reserves, the dollar remains close to 57%, the euro at 20% and the renminbi at 2%. The US lead in private AI capital is wide, while the proximity of some Chinese models to the best American models narrows the gap at the frontier. Neither measure, on its own, establishes who will monetise the technology best.^[9][16]

The economic research justifies a wide range of outcomes. AI improves results in specific tasks, but transferring those gains to the whole economy requires organisational investment, skills and demand. Productivity gains can coexist with difficulties in entering the labour market and with a larger share of income absorbed by capital. The relationship between innovation and political cohesion remains open.^[4][6][7]

For Italy the decisive step is diffusion across firms. High debt makes every durable acceleration in growth valuable, but technological catch-up is not automatic: a small firm with disorganised data, few technicians and limited financial capacity may have much to gain and yet remain unable to adopt the tools.

Implications for investment decisions

Research should focus on three complementary groups: assets that generate already observable cash flows; suppliers of capacity that is hard to replace; hedges consistent with different shocks. Electricity grids, automation and semiconductors deserve attention, but they are not universal safe havens. Strategic usefulness, corporate profitability and the return on the security must each pass separate tests.

The entry price remains a decisive variable. Rising demand can be accompanied by overcapacity, competition or the regulation of margins. Even a correct technological forecast can therefore produce financial losses. Portfolio construction must limit the common implicit bets: low real rates, continuity of Asian supply, availability of capital and the persistence of exceptional margins.

Bitcoin, stablecoins and tokenisation add a cross-cutting monetary dimension: non-sovereign assets, the diffusion of currencies and the transformation of settlement can alter control over financial networks. These channels have different implications for the dollar and for intermediaries.^[17][46]

The operational part pairs the scenarios with a shortlist of 15 ETFs and ETCs and one Bitcoin ETN, with ISINs, documented costs, function and exclusion conditions. Selection proceeds from exposure to vehicle and from vehicle to price, with an explicit check on overlaps.

Map of the report and reading criteria

Three levels of statement

Documented evidence. Official statistics, statistical compilations by specialised organisations and research findings. They are accompanied by a source number, with a link and a final bibliography. Data from one survey are not extended to populations it does not cover.

Analytical inference. Plausible causal links derived from comparing the evidence. For example, a grid-connection constraint can slow the expansion of computing even when financial capital is available. The direction of the mechanism is more robust than its quantitative intensity.

Scenario hypothesis. Configurations built to organise uncertainty and analyse the sensitivity of investments. Weights, monitoring thresholds, scores and debt/GDP simulations are explicitly the report's own elaborations. They are not provided by the institutions cited.

Timing matters

Statistics on reserves, robots and corporate adoption arrive with different lags. Markets, by contrast, anticipate future results and can move well before the final figures. For this reason the report distinguishes the economic signal from the price signal. An improvement in the data may already be incorporated in prices; a stock-market correction may reflect nothing more than earlier excessive expectations.

The market valuations reported are dated snapshots, with no buy recommendations based on geographic or sector membership alone. The dashboard does not update its sources automatically: it preserves a reproducible documentary base. The supporting spreadsheet contains the numbers and formulas needed to check the charts and revise the assumptions.

1. Dalio: a causal model to be made verifiable

Dalio's contribution lies in relating productive capacity, finance and political order. Eight recurring measures of power cover education, innovation, competitiveness, military strength, trade, production, financial centre and reserve currency. Debt and internal and external conflicts interact with these dimensions. The expanded list in this report is an analytical reorganisation, not a literal taxonomy of the book.^[1]

The methodological difficulty arises from the limited number of observable great historical transitions. The Dutch, British and American experiences are not independent trials run under the same conditions. The monetary system, institutions, technology, demographic structure and the destructive capacity of war all change. A historically recognisable sequence can direct attention without establishing the date or outcome of the next transition.

What needs to be measured

The term decline can denote at least four phenomena: a falling share of world GDP, a slowdown in domestic welfare, a loss of coercive capacity and reduced financial centrality. A country can exhibit the first and retain the others. The rise of initially poorer economies mechanically reduces the relative weight of the dominant power even as it continues to grow richer.

For an investor the distinctions become even more relevant. Listed companies can generate revenue outside their home country, and their profitability can diverge from national growth. Conversely, a rapidly developing economy may require continual capital increases or pass most of the benefits on to consumers and workers. The share of geopolitical power and the return per share are variables linked only through intermediate steps.

The proposed revision

Dalio 2.0 retains the circuit between economic capacity, debt and political legitimacy. It introduces a separate measure for each function of power, an assessment of the dependencies between nodes and a set of conditions that could invalidate the initial diagnosis. The question becomes: which capabilities are improving, in which domain and at what cost?

A hypothesis of American decline would lose force if productivity, the tax base and the ability to attract capital improved steadily without an equivalent rise in financing costs and internal conflict. A hypothesis of Chinese ascent would instead require separating productive success, the sustainability of domestic demand and the attractiveness of financial instruments. These tests reduce reliance on suggestive analogies and make the comparison between theses genuinely useful.

2. Tetlock: from narrative to testable forecasting

The contrast between foxes and hedgehogs is useful as a warning against reliance on a single explanation. It does not license classifying someone as a poor forecaster on the basis of their public style. Assessing forecasting ability requires precise questions, probabilities recorded before events and resolution criteria independent of the desired outcome.

Chang, Chen, Mellers and Tetlock document that brief training in probabilistic reasoning improved accuracy in geopolitical tournaments. The study reports Brier score improvements of the order of 6-11% relative to the control group. The experiment concerns well-defined events and bounded horizons: it does not directly validate a probability about the hegemonic country in 2035.^[2]

Questions that can be resolved

The sentence "AI will renew American power" must be broken down. One can test whether productivity growth exceeds a preset threshold, whether tax revenue grows faster than interest spending, or whether the dollar's share of reserves stays above a stated level. The answers do not settle the whole geopolitical problem, but they narrow the room to reinterpret every event as confirmation of the thesis.

The proposed protocol records for each question: wording, source, deadline, initial probability, rationale, updates and resolution rule. The probability is changed when the relative evidence between hypotheses changes, not every time a news item consistent with an already preferred story appears. One anomalous quarter should not carry the same weight as a persistent shift across several series.

Two separate disciplines

The first concerns calibration: among events assigned 70%, roughly seven in ten should occur over the long run. The second concerns usefulness to the portfolio. A correct forecast may yield no profit if the market was already pricing it in; a relatively imprecise forecast can be useful when the price offers a sufficient margin of safety.

The committee must therefore keep two registers. The register of forecasts measures the quality of judgements. The register of decisions measures how much was paid, what loss was bearable and why the exposure was compatible with the rest of the portfolio. Conflating the two leads one to treat every profit as proof of a good theory and every loss as proof of the opposite.

As an illustrative threshold for US acceleration, we propose productivity growth above 2.5% on average over three years, accompanied by sectoral diffusion. This is a rule of the report, not an economic law.

In this report the five scenario weights are a first working discipline. They are not a demonstration of forecasting skill: that can be assessed only through future updates and verifications.

3. Multipolarity: functions of power that come apart

Multipolarity does not require all powers to have equivalent strength. It can emerge when no single actor is able to control financing, production, technology, energy and security at the same time. The distribution is asymmetric: some countries exert influence across many domains, others hold a capability that is hard to replace in a narrow segment.

Farrell and Newman show how the structure of economic networks can confer capacities for surveillance and coercion on those who control central nodes. Mere interdependence does not automatically produce balance: jurisdiction, institutional tools and the concrete ability to exclude other actors from the network are also required.^[3]

From size to position

A large market can depend on a few equipment suppliers; a financial centre can depend on energy and data flows; a dominant producer can depend on access to foreign demand. The measure of power must therefore include the time needed to replace a function, the cost of replacement and the node's capacity to withstand retaliation.

The investment conclusion is that a company's nationality is not enough. A portfolio of securities listed in many countries can remain exposed to the same interruption in chip production or to the same cost of capital in dollars. It is the map of operational and financial dependencies that shows whether geographic diversification is also economic diversification.

There is, moreover, a limit to the coercive use of power. Making access to an infrastructure unstable gives customers an incentive to finance alternatives. The supplier can increase its influence in the short term and reduce its own centrality in the long term. This tension links industrial policy and valuations: a high rent today does not prove that the position will remain irreplaceable in 2035.

4. The map of poles and dependencies

The table offers a functional reading of the main actors. It indicates advantages and constraints relevant to the scenarios; it is not a quantitative ranking and does not assign any country a technological monopoly.

The evidence on AI capital, robotics and reserves confirms that these domains do not overlap perfectly. The observed industrial and monetary specialisation supports the hypothesis of a system with several centres, but does not prove that they will form a stable order.^[10][15][16]

The connector powers

Other countries can benefit from the shift of activity between blocs. The advantage depends on being able to offer market access, logistics capacity and credible rules. Being a commercial transit point is different from building an autonomous production base; part of the advantage can disappear if rules of origin become more restrictive.

For India and the Gulf it is particularly important to avoid mechanical equivalences. A working-age population must become productive labour; available capital must become operating plants and revenues. Investment research should give priority to verifying these steps rather than to indicators of scale with no corresponding economic return.

Europe also contains nodes of influence outside the EU. The United Kingdom and Switzerland retain financial and scientific functions that cannot automatically be counted as capabilities of the Union. In the report, the EU, geographic Europe and the euro area remain distinct concepts.

5. Selective globalisation and the cost of resilience

The working scenario envisages a combination of international trade and duplication of strategic capabilities. Regionalisation may accelerate in sensitive technologies and remain limited in other goods. There is no need to imagine the complete closure of borders to obtain significant effects on margins, prices and investment.

The ECB and the ESRB regard fragmentation as a source of macro-financial risk: a geopolitical shock can propagate through asset prices, trade, funding and credit conditions. The Financial Stability Review of May 2026 also notes that the conflict in the Middle East has already introduced a shock to energy supply and to growth prospects.^[24][25]

A hedge has a cost

Two suppliers, more inventory or a plant close to the final market reduce some operational risks. They can, however, increase working capital, lower plant utilisation and worsen the return on capital. The benefit must be measured against the expected cost of a disruption, not treated as a free gain.

Resilience is also shock-specific. Two factories located in different countries but dependent on the same material, software or logistics hub do not offer complete protection. An effective sourcing policy requires knowing the dependencies of suppliers, including those not directly under contract.

Inflation: level, persistence and policy response

A tariff or an energy price rise can lift the price level. Generating persistent inflation requires repeated shocks, transmission to wages and services, less well-anchored expectations or an accommodative economic policy response. The equation “fragmentation equals permanently rising inflation” is too simple.

The duplication of investment can also produce overcapacity in some sectors. As a result, higher energy prices can coexist with deflation in industrial goods. The aggregate outcome depends on the weight of the components and on monetary and fiscal reactions.

For the portfolio, what matters is identifying who can pass costs on to the customer, with what lag and for how long. A company with a high market share may have little pricing power if its main buyer is the public sector or if a regulator sets the permitted return. The growth of spending on economic security opens opportunities, but also changes the distribution of risks among taxpayer, consumer and shareholder.

6. AI as a general-purpose technology

A general-purpose technology changes different activities and makes complementary innovations possible. In the case of AI the effect can run through research, administration, design, customer service and production. Progress in a model is only the first step: usable data, redesigned processes, verification capacity and people able to manage errors are also needed.

The literature on the productivity J-curve describes precisely the role of complementary investments, often intangible and imperfectly measured. In the early phases the cost of transformation can precede the benefit recorded by the statistics; later, the results can become more visible. It is a plausible mechanism, not a guarantee that any AI spending will be rewarded.^[4]

Five economic steps

The speed at which these steps are crossed may differ by sector. Generating a draft text is simpler than guaranteeing a production cycle with responsibility for safety and reliability. Industrial applications often require hardware, installation, maintenance and integration with existing plants.

For a user, what matters is the cost per verified result: price, number of attempts, human review and reliability. Models that are cheaper per query can be more expensive across the full process. Competition between models can transfer value to those who control data, distribution and applications.

From substitution to creation

Reducing the time a task requires produces a saving. Creating a product that was previously impossible can expand demand. The two modes have different implications: the first can compress employment if demand does not respond; the second can generate new activities and investment. Their relative weight cannot be inferred from the number of subscriptions sold.

The thesis of a geopolitical renewal requires that the benefit exceed the cost of capital and spread beyond technology suppliers. If almost all the spending remains inside the AI supply chain, revenue growth among its components may overstate the increase in final demand. This is why contracts between companies in the same ecosystem must be distinguished from revenues coming from independent customers.

7. The economic evidence: strong results, open generalisation

Estimates on AI differ because they measure different objects. An experiment on one task, a macroeconomic model and a national statistic cannot be placed on the same scale without correcting for population, horizon and the definition of productivity.

Sources: ^[5][6][11]. The magnitudes are deliberately presented in a table and not as comparable bars.

Acemoglu's caution stems from the step from tasks to GDP: a large saving on a small portion of the economy can produce a limited aggregate increase. The estimate consulted concerns a cumulative level, not 0.66 percentage points of additional growth every year. New activities or subsequent progress can change its calibration, but do not remove the need to make the link between micro and macro explicit.^[5]

How opposing readings can coexist

An optimistic reading gives weight to the creation of new capabilities, accelerated research and lower coordination costs. A cautious reading highlights reliability, accountability, data scarcity and the difficulty of changing complex organisations. Both identify possible mechanisms; they diverge mainly on speed, breadth and complementarity.

An efficiency gain can also be passed on to consumers through lower prices. In that case the social benefit is high and the shareholders' rent may be modest. Productivity measurement, welfare and company valuation remain three distinct levels.

The operating criterion is to look for converging signals: repeated adoption, lower total costs, quality maintained, higher final demand and self-financing capacity. User enthusiasm is a first signal; a persistent improvement in cash flows and productivity outside the technology supply chain is a far more demanding proof.

8. Productivity and debt: the correct relationship

GDP growth helps to reduce the debt/GDP ratio, but debt does not stand still. Interest accrues, the public budget may be in deficit and financial operations can change the stock. The simplified relationship useful for the scenarios is:

d(t) = [(1 + i) / (1 + g)] × d(t-1) - p + a

Here d is debt as a percentage of GDP, i the effective average nominal cost, g nominal growth, p the primary balance (positive when in surplus) and a the stock-flow adjustment. The yield on a new ten-year bond does not coincide with the cost of the entire debt. The speed of transmission depends on maturities and refinancing.

A sensitivity exercise

The chart starts from 138.2%, the IMF forecast for Italian debt in 2026. From that point it assumes an average nominal cost of 3.5%, a primary surplus of 1.5% of GDP, zero adjustments and constant nominal growth of between 2% and 4%. Only the starting point comes from the IMF: the trajectories to 2035 are the report's own mechanical simulations.^[19]

The exercise makes the cumulative effect of growth visible, but does not prove that AI will produce any of the paths. Higher productivity can raise revenue, but can also induce more public spending or be associated with higher real rates. Part of the benefit on the denominator can be offset by the numerator.

The condition g > i is not sufficient to guarantee stability in the presence of large primary deficits. Conversely, a primary surplus can stabilise debt even when i > g. Judging sustainability therefore requires following growth, the budget, average cost, maturity and the credibility of institutions together. It is this set, rather than a single inequality, that connects AI to Dalio's cycle.

9. United States: the possibility of a productive renewal

The American advantage combines the availability of capital, the ability to finance loss-making companies through their early phase and a technology ecosystem that can distribute new tools rapidly. This combination makes a productivity renewal plausible, without implying that it is already measurable on the scale needed to change the fiscal path.

The AI Index reports $285.9 billion of private AI investment in the United States in 2025, against $12.4 billion in China, noting that the metric does not adequately capture all Chinese public support. The OECD instead measures venture capital: $194 billion to the United States, about 75% of the global value of deals. The two totals have different scopes and should not be added together.^[9][10]

How the financial advantage can become an economic one

Capital allows parallel experimentation, purchases of computing capacity and absorption of initial losses. To turn it into a lasting advantage, however, at least some of the experiments must generate solvent demand. Otherwise the abundance of funding can delay the weeding-out of less efficient projects.

The favourable case requires productivity growth extending to healthcare, business services, logistics, manufacturing and administration. The fiscal effect would be more robust if incomes and the tax base rose broadly. High profits concentrated in a few companies do not, on their own, guarantee a proportional growth in public revenue.

The advantage can also be limited by domestic constraints: permits, grid connections, training and access to talent. Leadership in models does not automatically solve these problems. The test of the American renewal will be the ability to grow the AI-using economy alongside the AI-producing one, while maintaining an adequate return on invested capital.

10. The American fiscal constraint remains open

Financial leadership makes the American problem different from that of a country that borrows in foreign currency, but does not eliminate it. Borrowing in one's own currency reduces some liquidity risks and introduces other adjustment channels: inflation, changes in real yields and redistribution between creditors and taxpayers.

In the Budget and Economic Outlook published in February 2026, the CBO projects federal debt held by the public at 101% of GDP in 2026 and 120% in 2036. For 2026 the projected total deficit is 5.8% and the primary deficit 2.6%. This is a baseline founded on January legislation, not an outturn nor an unconditional forecast updated to September. Its scope is not directly comparable with the gross general government debt of Italy.^[22]

Three possible transmissions of AI

The first runs through revenue: more output and income can broaden the tax base. The second runs through spending: better administration can make service delivery more efficient, but the savings require organisational change and political choices. The third runs through rates: greater investment opportunities can increase the demand for capital and the required real return.

These channels can offset one another. It is not correct to extrapolate higher nominal growth while holding rates and fiscal behaviour unchanged as if it were a forecast. That assumption is useful only as an exercise to isolate a mechanism.

The test for the favourable thesis

The American renewal becomes more convincing if cyclically adjusted revenue improves, productivity remains strong for several years and the premium required to hold long-dated debt does not rise in a destabilising way. It becomes less convincing if technological progress is accompanied by growing structural deficits and greater uncertainty about monetary institutions.

For the portfolio, the separation between corporate success and fiscal discipline rules out a common shortcut: treating any dollar-denominated asset as an equivalent bet. Equities, short-dated bonds, long-dated debt and cash respond differently to real growth, inflation and the term premium. Conviction about American leadership does not, on its own, establish which of these exposures is best rewarded.

11. Labour, distribution and political stability

AI can raise output per hour while simultaneously reducing demand for certain tasks. The net effect on employment depends on the creation of new activities, on the response of demand and on how long it takes people to be reallocated. A profession's exposure to a technology is not the same as its elimination. The IMF draws precisely this distinction between exposure, substitutability and complementarity.^[8]

The August 2026 revision of the Canaries in the Coal Mine study finds no evidence of a generalised expulsion from the workforce. It does, however, report that employment among 22-25 year olds in exposed professions is 19% below the counterfactual path built on less exposed professions. The adjustment comes mainly through reduced hiring. This is observational evidence on administrative data, not a causal measure of AI that is immune to confounding factors.^[7]

The entry-level problem

If junior tasks also serve to learn the trade, their reduction may alter how future professionals are trained. A company that saves on entry-level labour may find itself with fewer experienced staff in later years. The cost does not necessarily show up in the first year's income statement.

Conversely, tools that make once-rare skills accessible can ease the entry of new operators, increase competition and lower prices. The productivity gain among less experienced workers observed in customer support shows that complementarity is possible, but it does not determine the aggregate balance on employment.^[6]

From the labour market to politics

The transition becomes politically fragile if costs are concentrated and immediate while benefits are diffuse or delayed. A rise in average income can coexist with losses of status and insecurity for significant groups. There is, however, no stable function that converts an employment figure into a probability of institutional crisis.

The variables to watch are real incomes, the duration of unemployment, occupational mobility, access to training and the distribution of capital ownership. The design of institutions shapes how progress translates into welfare and consent.

For investments, the concentration of margins may initially support valuations and later raise regulatory risk, fiscal pressure or distributional conflict. Social sustainability therefore enters the analysis of cash flows as a possible condition for the durability of returns, without being reduced to a generic reputational judgement.

12. China: industrial scale as an option on AI

China's most relevant advantage for this analysis lies in the ability to connect digital tools to an extensive industrial base. Improving quality control, maintenance, logistics and design can generate value even without permanently holding the top-ranked language model.

According to the IFR, China installed around 295,000 industrial robots in 2024, 54% of the world total. Among the markets shown in the chart, Japan and the United States follow. Italy installed 8,783 units, remaining the second-largest European market. These are installation flows of industrial robots, not a measure of robots equipped with advanced AI or of the total stock.^[15]

The AI Index records, as of March 2026, a gap of about 2.7% in the cited comparison between the best US and Chinese models. It is a specific measure, not a universal indicator of technological capability. The distinction between American digital AI and Chinese physical AI describes relative advantages, not an exclusive division of labour.^[9]

From productive capacity to profitability

Scale enables learning, supplier density and the diffusion of solutions across plants. But high productive capacity must meet remunerative demand. If firms expand supply simultaneously in saturated markets, the result can be falling prices and margins, even as the technology improves.

The industrial advantage does not automatically resolve property-sector fragilities, demand imbalances or the quality of capital allocation. The Chinese recovery hypothesis requires an improvement in economic efficiency, not merely an increase in the number of plants. It is a condition of the scenario, not an outcome already secured.

The investor must also distinguish between the benefits to the Chinese economy and the economic rights of the accessible shareholder. Governance rules, corporate structure, the risk of restrictions and the actual remuneration of capital all bear on the final result. A positive thesis on robotisation is therefore not the same as indiscriminately buying the Chinese equity index.

13. Electricity: capacity available where it is needed

The expansion of AI makes the availability of reliable electricity more important, but energy is not a homogeneous market. Generation capacity, grid, connections, flexibility and cooling must be available in the same place and within the required timeframe. The number of projects announced is not the same as the capacity that can actually be powered.

The IEA estimates data-centre electricity consumption at 485 TWh in 2025 and, in its updated central projection, around 950 TWh in 2030. The second figure is a conditional scenario. It covers all data centres: it should not be attributed entirely to AI.^[12]

The rent can stop at different points

An equipment manufacturer can benefit from long delivery times; a grid operator can expand its remunerated capital base; a generator can enter into long-term contracts. Each model has a different exposure to construction costs, interest rates, energy prices and regulation.

For a utility, demand growth and earnings growth are not synonyms. The return on investment may arrive with a lag and require external capital. For a data centre, a contract with a large customer can reduce commercial risk but increase concentration, and it does not eliminate the technological risk of the hardware.

The IEA also includes electrification and other industrial and residential uses in the expansion of electricity demand. This plurality of drivers makes the theme less dependent on AI, but it does not make it immune to recession, the cost of capital or overinvestment.^[13]

Selection should favour capacity that is already authorised, intelligible contracts, a sustainable financial structure and return on capital. The strategic theme is the availability of a reliable service, not the mere possession of an energy label.

14. Strategic scarcity and investable rent

Strategic scarcity concerns an input whose unavailability prevents an important activity from being carried out. For the investor, however, at least three further requirements apply: economic rights over the scarce capacity, the ability to retain its margins, and a share price compatible with the risk. An indispensable service can be a mediocre investment if its return is regulated or competition absorbs the rent.

The IEA documents a persistent concentration in the refining of critical minerals. In its pipeline scenario, the projected gap for copper in 2035 narrows relative to the previous year but remains significant. It is a conditional comparison between demand and projects, not a certainty of physical shortage or a forecast of the copper price.^[14]

A more complete economic criterion

The formula «scarcity × relevance × pricing power» is a good research question, but not a valuation model. To arrive at value one needs volumes, margins, reinvestment, the durability of the advantage and the cost of capital. Additional profitability must be measured against the capital required to sustain it.

A company can increase operating profit by 10% and have to raise investment by 30%. If the additional capital earns less than its cost, growth destroys value. Conversely, a supplier with capacity already installed can monetise an expansion of demand without the same financing requirement.

The most useful test is therefore the incremental return on capital, accompanied by an estimate of how long the scarcity will last. The term «structural» does not mean permanent: high prices finance precisely the solutions that can eliminate it. Research must look for the winners and for the mechanisms through which their advantage could be arbitraged away.

15. Dollar, euro, renminbi and gold

Monetary centrality encompasses reserves, funding, invoicing and payments. The shares are not interchangeable. A currency's weight in payments can change without an equivalent change in reserves, while the dollar value of holdings is affected by exchange rates.

The chart shows the rounded shares cited by the ECB for 2025: 57% dollar, 20% euro and about 2% renminbi. The 21% residual is calculated on rounded values. Gold is excluded from the denominator of foreign-exchange reserves; it cannot be added as a fourth slice of the same distribution.^[16]

Why the alternative is not automatic

An international currency requires investment opportunities, liquidity, legal trust and markets capable of offering assets usable as reserves. Increasing a share of world trade does not immediately generate all these conditions. A diversification scenario can therefore reduce the dollar's relative role without producing a single substitute.

A possible move in the share from 57% to 50% is treated only as a hypothetical example. It is not a forecast derived from the data. To assess a trend one must distinguish between exchange-rate movements, transactions and the composition of the statistical sample.

Technology may also strengthen the dollar

The BIS observes that stablecoins are predominantly pegged to the dollar and sets out the limits and risks of their monetary functioning. An analytical possibility follows: new payment channels can extend the use of the American currency beyond traditional circuits. Payments innovation and de-dollarisation are not synonyms.^[17]

Physical gold is not the liability of an issuer. This distinguishes it from credit assets, but it does not make it free of volatility, costs or custody risk. The form of holding also introduces characteristics of its own. It can perform a diversification function in crises of confidence, while it does not guarantee a perfect inflation hedge over every horizon.

15a. Bitcoin, stablecoins and tokenisation: the digital monetary dimension

The monetary extension completes the framework: AI changes productive capacity and distribution; digital finance changes the forms of value, settlement and the points through which control is exercised. It is a force that cuts across the five scenarios. It does not require assuming a macroeconomic impact identical to that of AI, nor adding a sixth regime.

Three objects, three functions

Bitcoin's original design proposes peer-to-peer transfers with a distributed consensus mechanism. The protocol's scarcity does not by itself determine the price: for the investor, demand, access and security of holding remain central.^[50]

The alternative can coexist with a stronger dollar

Stablecoins are predominantly referenced to the dollar. Their spread can widen access to the American currency. Tokenisation, by contrast, can be adopted by regulated infrastructures, even without a public blockchain. The choice of platform does not coincide with the choice of currency.^[17]

From this distinction follows one of the report's inferences: a more digital world can preserve, or strengthen, certain forms of American centrality. Bitcoin introduces a different channel, but its growth does not imply that payments, credit and reserves will migrate together towards a single alternative standard.

The entry of states and the new intermediaries

The US executive order of 6 March 2025 establishes a strategic reserve initially funded with definitively forfeited Bitcoin and provides for the study of additional budget-neutral acquisitions. The document evidences a policy direction; it does not certify net purchases on the market, nor the volume actually held in the reserve today.^[44]

The IMF regards tokenisation as a transformation of the financial architecture: integrated settlement can reduce some frictions, while the need for liquidity and the transmission of errors become more immediate. Custody, code and platforms acquire systemic relevance.^[46]

The Dalio question thus extends to new points of control: issuers, reserves, access, standards and infrastructure. A system that is distributed at the technical level can have concentrated on-ramp and custody services. For the investor, value may be captured by intermediaries or users without accruing to a network's token.

15b. Digital finance: scenarios and the choice of exposure

The working probabilities remain 35%, 25%, 20%, 10% and 10%. The digital dimension changes the mechanisms to watch; it does not authorise a recalibration without new evidence.

This analytical row remains separate from the numerical scale of the matrix. We do not assign Bitcoin a constant correlation, an expected return or a universal protective function. The IMF documented a rise in correlations with equities in 2020-2021 relative to 2017-2019: it is historical evidence of the instability of the relationship, not an estimate that automatically holds in 2026.^[45]

An identifiable candidate, to be compared

VanEck Bitcoin ETN (VBTC), ISIN DE000A28M8D0. Documented TER 1.00% per annum; physical replication via Bitcoin. It is an exchange-traded note, not a UCITS fund. The monthly fact sheet is dated 31 August 2026.^[48]

The investor buys the security and remains exposed to the issuer and to custody as well; they do not receive the private keys to the BTC. Spreads, market hours and the risk of price moves while the exchange is closed must be considered. Its presence in the shortlist identifies an operational example: it does not demonstrate that it is the cheapest vehicle or preferable to every competitor.^[49]

Sizing the loss and measuring adoption

Before selecting the ETN, compare other instruments on the same underlying, guarantees, rights over collateral and total cost. A thesis on bank tokenisation is not enough to justify a purchase of BTC. The position, if admitted, requires a risk budget separate from gold, cash and technology equities.

Monitoring adds: the size and quality of stablecoin reserves; the currencies actually used; transfers attributable to payments, stripped as far as possible of technical movements; tokenised securities actually traded; interoperability; flows into BTC products; concentration of custody and rolling correlations during shocks. None of these indicators is presented here as live data.

The IMF stresses that platforms can make the system more efficient or fragment it. The operational test concerns transactions, liquidity and the ability to resolve incidents, beyond announcements.^[47]

16. Europe: a collective capacity still to be built

Europe has scientific, industrial and financial resources; the decisive question is its capacity to use them together. The Draghi report organises its diagnosis around innovation, competitive decarbonisation, security and financing. It also highlights the need for additional investment of the order of €750–800 billion a year. That figure refers to the overall requirement identified in 2024, not to AI alone, nor to a programme that is already fully funded.^[18]

The problem of scale

A company may have good technology and still struggle to raise capital, grow beyond its home market or sell to public-sector customers with differing procedures. The single market offers potential scale; how far that potential becomes productivity depends on the ability to operate as a genuinely integrated market.

In AI there is no need to replicate every segment of the supply chain. A credible strategy can combine competitive access to models and cloud, control of critical data, the ability to switch suppliers and specialisation in industrial applications. Economic sovereignty need not mean autarky, which can raise costs without producing a real alternative.

Two trajectories

In the favourable case, capital, energy networks and public demand become better coordinated. Specialised firms can scale, dependencies are managed and a larger supply of liquid financial assets strengthens the role of the euro. The outcome would be a third pole with selective capabilities, not necessarily a global hegemon.

In the weak case, programmes remain fragmented and resources fund inefficient national duplicates. The best talent may continue to generate innovation, but the rewards of distribution and scale shift elsewhere. The risk is a loss of bargaining autonomy rather than a disappearance of productive capacity.

Implications for selection

Not all European stocks benefit equally from integration. One must identify the concrete obstacle a reform would remove: the cost of energy, access to capital, aggregate demand or permitting times. A political announcement is a preliminary signal; appropriations, orders, operational projects and economic returns are the subsequent tests.

Europe therefore remains an option conditional on execution. Buying a valuation discount indiscriminately, without checking why it exists, can expose the investor to weak growth, low profitability or an unfavourable sector structure rather than to a mere market inefficiency.

17. Italy: little room for weak growth

Italy combines high debt with modest growth. In the IMF's July 2026 framework, real GDP grows by 0.5% in 2025 and is projected to grow at the same pace in 2026 and 2027. Debt rises from 137.1% in 2025 to a projected 138.2% in the two following years. This trajectory incorporates assumptions about prices, the budget and external conditions; it is not an inevitable outcome.^[19]

The constraint is dynamic

High debt increases sensitivity to shocks even when its average maturity dilutes the immediate transmission. The cost of new funding can feed progressively into the budget; a slowdown in growth can reduce revenue and increase the weight of relatively rigid spending. The interaction matters more than an isolated reading of a percentage ratio.

Private savings and household wealth contribute to the system's resilience, but they are not automatically resources available to repay public debt. Moving from one to the other requires economic or fiscal decisions with distributional and behavioural effects.

AI changes the potential; it does not remove the constraints

A technology that makes labour more efficient can help a country with unfavourable demographics. However, if lower costs come with fewer hours worked, the gain in productivity may exceed the gain in GDP. Labour participation, skills and demand remain decisive.

The Italian investor, too, must avoid overlapping professional, property, banking and bond exposures. Holding BTPs, working in Italy and owning Italian real estate creates a common sensitivity to incomes, credit and public finances. The robustness of one's wealth requires looking at the whole, not merely at the composition of the securities account.

18. The potential for diffusion in Italian firms

The European comparison offers a more direct indicator than generic digital intensity. ISTAT records the use of at least one AI technology by 16.4% of Italian firms with at least ten employees in 2025, up from 8.2% in 2024. Eurostat reports 20.0% and 13.5% respectively for the EU. The comparison covers the population captured by the surveys, not all Italian micro-enterprises.^[20][21]

A lag is not an expected return

A lower starting level leaves room for improvement, but it can also signal persistent obstacles. The availability of cheap software does not remove data shortages, managerial difficulties, limited capital or a weak ability to assess suppliers. Catching up becomes a credible thesis when these obstacles are addressed.

For Italian manufacturing, the link with AI can run through machinery, design, machine vision, maintenance and order management. Knowledge of the production process can become an advantage complementary to purchased technology. Not every firm needs to develop its own frontier model.

The industrial North, including Friuli-Venezia Giulia, is an area of interest for this hypothesis. This is a reading based on productive specialisation, not a regional estimate of AI's contribution. Measuring it would require data on adoption, investment and productivity at the level of firms and territories.

What to look for in suppliers

A company selling integration, components or automation should demonstrate sustainable installation times, replicable results and customers who renew their contracts. Many pilot projects without a move to production suggest that the economic value is not yet established. The most important evidence is an improvement in the customer's process, accompanied by a satisfactory margin for the party that makes it possible.

19. From demonstration to the factory floor: an illustrative case

Italian productivity must also be analysed from the bottom up. A firm can start from an economically relevant, measurable and well-defined problem: defects, downtime, slow quoting or manual document handling. Choosing the technology before the problem increases the risk of funding an experiment with no usable result.

A numerical example, entirely hypothetical

Consider a manufacturer with an annual cost of defects of €400,000. A quality-control system reduces that cost by 15%, after verification on real production. The gross saving would be €60,000 a year. With €20,000 in recurring costs and €100,000 of initial investment, the net pre-tax benefit would be €40,000 and the simple payback period 2.5 years.

This example is not a market benchmark. It shows why a technical saving must become an economic saving. If the reduction in defects were only 5%, the gross saving would equal the recurring cost and the project would not recover the investment. If the system shortened lead times but neither cut costs nor increased saleable output, the monetary benefit could be smaller than the operational one.

Four steps of execution

First, build a baseline with full costs and quality indicators. Next, test on a portion of the process, monitoring errors and the effect of learning. Then decide whether to extend use on the basis of the observed return. Finally, verify that the benefit persists after maintenance, upgrades and product changes.

For those investing in industrial companies, this discipline makes it possible to distinguish AI announcements from economic improvements. Value lies not in the number of applications introduced, but in a sustainable increase in cash flows, after all the costs needed to maintain it.

20. Five scenarios: definition and working weights

The scenarios describe the prevailing regime in the second half of the 2026–2035 horizon. They are not five fully independent events: a country can improve in AI during a phase of fragmentation. To make the matrix usable, the report classifies the future according to the mechanism that dominates growth, earnings and risk premia.

The weights take up the initial assumption of the thesis and are maintained as an editorial distribution open to discussion. They have not been estimated on a historical sample or derived from institutional sources. Their sum of 100% guarantees accounting consistency, not accuracy. The dashboard serves above all to test how much the conclusions change when the weights change.

Classification rule

If persistent supply shocks prevail, the regime is S3. If a disinflationary contraction tied to disappointment over investment prevails, it is S5. Among outcomes with positive productive diffusion, S2 denotes a strengthening relative American leadership, S4 an Asian recovery that drives the economic surprise, and S1 a more widely distributed benefit. This rule is a convention and leaves intermediate cases.

The shocks of 2026 make some components of S3 already observable. The 20% weight concerns the persistence of the regime over the medium term, not the probability of geopolitical tension arising from today onwards. Confusing the two horizons would make the probabilities look inconsistent with events that have already occurred.

Extreme events, including a serious disruption of Asian semiconductors or a sovereign crisis, must also be analysed as cross-cutting stresses. No second numerical distribution on "who will be the hegemon" is produced: without an explicit mapping, aggregating these scenarios into different percentages would create spurious precision and overlaps.

21. S1: managed multipolarity and productive diffusion

Initial weight: 35%. The US retains financial and technological centrality; China and other poles strengthen specific functions. AI diffusion improves productivity without making social conflicts and the costs of fragmentation unmanageable.

Mechanism and verification

Organisational investment, skills and infrastructure carry advances in models through to production. Supply chains adapt and growth spreads beyond AI suppliers alone. The scenario gains strength with widespread productivity, rising returns on capital and rising incomes; it loses credibility if the benefits remain concentrated and energy constrains adoption.

Three instruments, three functions

Documented annual cost: OCF/TER; it does not include all implementation costs. Dates and documents in the bibliography.

Operational preference. VWCE is the shortlist's base candidate when global exposure is required. GRID adds a specific thesis on grids and electrification. RBOT should be chosen only if its composition matches the desired exposure: as at 9 September 2026, technology accounts for 72.07% and industrials for 22.46%.^[31]

The robotics label therefore does not guarantee diversification away from software and semiconductors. For factory automation alone, a different industrial selection may be needed. GRID, for its part, invests in equities: it does not offer the profile of private infrastructure with contracted revenues.

When to wait. If the price already discounts exceptional diffusion, or the portfolio is concentrated in the same supply chain, the theme may be right and the purchase ill-timed. Check the margins and investment of the underlying issuers, as well as flows into the ETFs. Bonds and liquid reserves remain functions distinct from the equity base.

Digital monetary channel. Interoperability and institutional adoption may accompany S1. Bitcoin remains a distinct exposure, to be assessed against demand and liquidity; see chapter 15b.

22. S2: AI-driven American renewal

Initial weight: 25%. The US converts capital, models and infrastructure into stronger and more persistent productivity than the other major economies. The benefits pass through to earnings, incomes and the tax base; fiscal sustainability improves only if spending does not absorb the entire dividend.

Mechanism and verification

The discriminating signal is growth in final demand and cash flows, together with productivity over several years. Better benchmarks and rising capex are not enough on their own. High real yields can accompany economic success while compressing equity multiples or penalising long duration.

Three levels of exposure

Annual costs are TERs. All three share classes are denominated in USD and do not hedge the exchange rate into euro; buying a EUR listing does not change this.

Operational preference. CSPX expresses American renewal more broadly than a basket of AI themes alone. SMH is the specific choice when the conviction concerns semiconductors; LOCK when it concerns spending on digital security. The ISIN identifies the European product: the ticker SMH also identifies a different US fund.

CSPX added to VWCE deliberately increases the US weight. SMH and LOCK should be funded from a thematic budget: adding them together without checking the underlying issuers can amplify the same bet on available capital, high margins and digital demand.

When to discard or reduce. An invalidated thesis, weak orders, investment not covered by external revenues, or a price incompatible with realistic margins. The alternative to an overpriced thematic fund may be to stay with the broad base, without necessarily replacing it with another ETF of a similar name.

Digital monetary channel. USD stablecoins and tokenised infrastructure can extend America's financial advantage. BTC adoption and its valuation do not necessarily follow the same trajectory.

23. S3: persistent fragmentation and stagflation

Initial weight: 20%. Restrictions, duplication and supply shocks outweigh the productivity dividend. Energy and security absorb more resources; deficits may widen while central banks confront the conflict between price stability and support for growth.

Mechanism and verification

Energy-importing and energy-exporting countries react differently. Shortages may be regional and temporary. The scenario strengthens with persistent price increases, long delivery times and less stable inflation expectations; it weakens with supply normalisation and broader diffusion of production.

Instruments linked to different shocks

Annual costs are TERs. The gold ETC is not a unit in a UCITS fund: its legal and custody structure should be read in the prospectus.^[38]

Operational preference. SGLN is consistent with a gold reserve, IBCI with sensitivity to euro-area inflation, DFNS with growth in defence demand. They are different responses. DFNS holds global assets and does not, on its own, express a thesis of exclusively European military autonomy.

IBCI can lose value if real rates rise, even with high inflation. Gold may be sold in the first phase of a liquidity crisis. Larger military budgets may already be priced into producers' valuations.

Reasoned exclusions. The shortlist does not include an oil ETC chosen solely on a forecast of higher prices: the futures curve, contract rolls and the price path require separate analysis. Likewise, miners and energy producers add company-specific risks on top of the commodity. Strategic relevance does not make protection universal.

Digital monetary channel. Regional payment networks and restrictions may increase fragmentation. Bitcoin may attract non-sovereign demand while at the same time suffering access restrictions or deleveraging.

24. S4: Chinese industrial recovery and Asian rebalancing

Initial weight: 10%. Industrial AI and robotisation contribute to a Chinese recovery underpinned by productivity, domestic demand and returns on capital. The rebalancing may be commercial and industrial without displacing the dollar's monetary centrality.

Mechanism and verification

Distinguishing recovery from overcapacity requires sustainable earnings and demand. More output financed with poorly remunerated capital does not confirm the thesis. Governance and the distribution of benefits to shareholders remain decisive: an industrial advantage may be passed on to customers through lower prices.

Choosing the actual geography

Annual costs are TERs. Accumulating USD share classes, with currency exposures not hedged to EUR.

Operational preference. EIMI is the candidate for a broad emerging-market rebalancing; FLXC for a deliberate overweight in China. FLXI covers India and can sit alongside a multipolar thesis, but it is neither an automatic hedge for China nor a necessary consequence of S4.

FLXC invests in the Chinese equity market as represented by its index, not only in robotised factories. If the conviction is strictly industrial, one must check the constituents' revenues or build research on specific companies. RBOT, presented under S1, offers another exposure to automation, with the technology composition already noted.

When to wait. If the only justification is the discount to the US, the essential step is missing: earnings quality, protection of minority shareholders and access to cash flows. For India too, favourable demographics do not make the multiple paid irrelevant. China's foreign suppliers remain exposed to restrictions and local substitution.

Digital monetary channel. A larger role for Asian payment circuits does not equal greater demand for Bitcoin. Distinguish currency, platform and investable asset.

25. S5: AI disappointment and disinflationary contraction

Initial weight: 10%. Revenues and savings do not justify the pace of investment. The revision of expectations reduces capex, demand and valuations. A technology can keep improving while those who financed capacity at excessive prices lose capital.

Separating recession from fiscal crisis

In S5A inflation and yields fall while issuers' credibility holds. This is the sub-case used in the matrix. In S5B the contraction combines with a fiscal or monetary crisis: risk premia may rise and the long-dated debt of the country concerned may lose value. S5B is a cross-cutting stress, with no additional probability on top of the 100%.

Three instruments with different duration

Documented annual costs. XEON is accumulating; the SEGA and DTLE share classes shown distribute income. The cost of DTLE's currency hedge is not the same as its TER.

Operational preference. XEON is the candidate for a reserve with low rate sensitivity; it is not a guaranteed deposit. SEGA offers euro-area government exposure, including the sovereign risk of several countries: it is not equivalent to a Bund-only portfolio. DTLE expresses a strong thesis of falling US yields with the currency hedged.^[39][40][42]

Under S5A, DTLE can be a tactical protective component. Under a US-centred S5B it may be precisely the exposure to avoid. Its effective duration in the August fact sheet is 14.94: a one-percentage-point move in yields implies a first-order price sensitivity close to 15%, before coupons and convexity.^[42]

Revision signals. Cancelled orders, weak utilisation and financing difficulties must appear together. Liquidity serves to avoid forced sales; gold may help in a crisis of confidence but may lose in the first phase. None of the three bond instruments replaces a definition of future expenses and maturities.

Digital monetary channel. Under S5A Bitcoin may suffer from the selling of risk assets; under S5B any alternative demand does not remove volatility and liquidation risk. Its hedging function must be verified.

26. Scenario sensitivity matrix

The matrix summarises qualitative sensitivities, other things being equal, over a multi-year horizon. The scale runs from -2, strongly unfavourable exposure, to +2, favourable; 0 indicates an ambiguous or balanced effect. It does not measure absolute returns, percentage outperformance or estimated correlations. Valuations, selection and the path of interest rates can change every result.

The more cautious assessment of infrastructure and automation under S3 acknowledges debt, cyclicality and execution risk. Inflation-linked bonds do not carry a universally favourable sign in recession, because they depend on both inflation and real yields. Gold retains a potentially useful sensitivity, with high uncertainty about the path.

Bitcoin: a conditional row

The monetary extension adds the reading of Bitcoin described in chapter 15b. Under S1 and S2 adoption may help, under S3 and S5 liquidity and access matter, while S4 implies no automatic benefit. The row remains outside the dashboard's scores and averages.

How to use the colours

A row that is positive in several scenarios is a reason to dig deeper, not to raise its weight automatically. The same factors can support many rows: hyperscaler capex, industrial demand or the cost of capital. Adding up themes with different names can amplify a common risk.

The dashboard allows the weights to be changed and computes an average of the scores. It is a heuristic exercise: the scale is ordinal, so the distances between -2 and -1 are not estimated economically. Readings of the kind “score 1 equals 10% return”, or the use of the result to optimise allocation automatically, are not permitted.

Before any real decision, the matrix must be set alongside prices, underlying exposures, risk of loss and liquidity needs. The value of the exercise lies in making the thesis's dependencies explicit and showing which conclusions change quickly when the assumptions are modified.

27. Valuations: how much future is already in the price

The market snapshot serves to put the thesis in context, without replacing the analysis of individual securities. The MSCI fact sheet as at 31 August 2026 reports the following multiples. The forward P/E uses expected earnings and is exposed to revisions. Differing sector composition means that not every gap can be read as a pure geographic discount.^[26]

An example of the risk of overpaying

Suppose initial earnings of 1 per share, a price of 40 and earnings growth of 15% a year for five years. With a terminal multiple of 20, the final price would be about 40.2: the annualised return from price alone would be almost nil. The company would have doubled its earnings and the investor would have done little more than recover the nominal capital.

The example is hypothetical and excludes dividends, costs, taxes and changes in the share count. It does not describe a listed company. Its purpose is to isolate the interaction between growth and the terminal multiple, also reported in the spreadsheet.

Valuation work must then include competition, returns on capital, investment intensity and the durability of margins. A comparison of P/E ratios can be misleading if earnings are temporarily elevated, depreciation does not reflect obsolescence, or growth requires continual share issuance. The technology thesis becomes investable only after these steps have been cleared.

28. From scenario to instrument selection

Selection takes place in two steps: identify the useful exposure and verify which vehicle delivers it at an acceptable cost and risk. Pages 26–30 propose 15 ETFs and ETCs identified by ISIN; chapter 15b adds a Bitcoin ETN, for a total of 16 candidates. They are a targeted working universe, not an exhaustive ranking of every product available.

Criteria that really decide

Shortlist preferences. Global: VWCE; US overweight: CSPX. Grids: GRID; semiconductors: SMH; broad automation: RBOT. Emerging markets: EIMI, reserving FLXC and FLXI for explicit geographic convictions. Protections should be selected for the shock, as on page 34.

Checking for overlaps

Exposure to an issuer is the sum of each fund's weight multiplied by the issuer's weight within it. The global candidate's FTSE All-World already shows about 61.6% US in the July 2026 fact sheet. Adding CSPX deliberately changes that balance; adding thematic funds can increase the same risks further.^[28]

The selection rule. For the same desired exposure, compare effective cost and replication quality. Across different exposures, the TER is not the dominant criterion. A very cheap ETF can be the wrong instrument. If the vehicle passes the technical checks but a sustainable valuation of the underlying assets is lacking, it stays under observation.

In the HTML version, the 'Instruments' command opens the selector by scenario, currency and cost. The Bitcoin group is kept separate from the five scenarios. The cards link to the issuers' documents; the five scenario pages and chapter 15b set out the function and risk of each candidate.

29. Protections: choosing instrument and size

The shortlist alternatives are not interchangeable. The choice depends on the shock to be hedged and on the loss one is willing to bear along the way. A fund's risk may differ from what its name suggests.

ISINs and costs are on pages 28 and 30; issuer documentation is in the bibliography.^{[38][39][40][41][42]}

Sizing to the shock

The first-order relationship is price change ≈ -duration × change in yield. For DTLE the documented duration is 14.94; for IBCI it is 7.20 as at 10 September 2026. A parallel shock of +1 percentage point implies roughly -14.9% and -7.2% respectively, before coupons, convexity, indexation and other changes. For IBCI the relevant reference is the real yield.^[41][42]

An example of a limit, not a recommended weight. If a position with an assumed shock of -15% may contribute at most 1% to the portfolio's loss, the arithmetic ceiling is 1/15 = 6.7%. Correlations, gaps and liquidity may call for a lower weight. This calculation makes the risk explicit instead of deriving the weight from enthusiasm for the thesis.

Currency, maturities and cost

A EUR listing does not hedge the currencies of the underlying assets. A EUR-hedged share class reduces currency risk but involves rate differentials and costs: the portfolio's USD return is not the net return in euro.

For an expense with a fixed date, compare the fund with individual bonds or defined maturities consistent with that commitment. An ordinary bond ETF rolls its portfolio: it does not promise redemption at par on a date chosen by the investor. Always keep the spending reserve separate from the tactical position.

30. 2045: structural options and the limits of forecasting

Beyond 2035 the risk grows that not only the variables change, but the relationships between them. Projecting the same growth rates for another ten years produces precise numbers without making the analysis any more reliable. 2045 is therefore treated through structural possibilities, without a numerical probability distribution.

Acceleration of research and the physical economy

If AI sharply reduced the cost of research and design, the benefits could extend to materials, pharmaceuticals, energy and automation. Faster growth in supply could ease some of the scarcities considered decisive today. The investor should avoid turning the bottlenecks of 2026 into perpetual rents.

In the opposite case, the difficulty of operating in complex physical environments could limit diffusion. The technology would remain very useful in some tasks, but with macro effects below expectations. Safety, reliability and liability requirements can slow the passage from demonstration to production.

Ownership, institutions and distribution

Greater automation capacity makes the distribution of capital ownership and data rights more important. Labour income may evolve differently from aggregate output. Fiscal, educational and competition institutions will shape how economies absorb the transition.

There is no single conclusion on concentration. The technology can strengthen platforms with scale and distribution, or lower barriers to entry and allow small operators to compete. Which dynamic prevails will also depend on interoperability, access to data and the cost of computing capacity.

International order and the physical environment

Digital infrastructure will continue to depend on territory, energy and materials. Climate change, water availability, grid security and the reliability of trade routes can alter the geography of investment. At the same time, energy innovations could reduce dependencies that are central today.

The portfolio consequence is to preserve options for adaptation: adequate liquidity, spread-out exposures, attention to obsolescence and the capacity to revise the thesis. The long term does not justify indifference to price. It justifies broader research into the durability of advantages and the alternatives that could destroy them.

The conclusion remains open but operational: use Dalio to identify the forces, Tetlock to discipline judgements, and allocation to limit the cost of error. Robustness comes from the ability to pass through different futures without being forced to stake everything on a single winner.

31. Indicator register: AI, the United States and China

The register separates the series to be monitored from the operational thresholds proposed by the report. The thresholds are not estimates of the institutions cited. A signal requires confirmation across several observations and comparison with data revisions; it does not automatically trigger a change to the portfolio.

Reference sources for the register: BLS, IFR, CBO and the New York Fed.^{[11][15][22][27]}

32. Indicator register: China, Europe and Italy

For Italy, adoption, output, hours and investment must be read together: a purchased licence does not demonstrate a productivity gain. References: ISTAT and Eurostat.^[20][21]

33. Verified data dashboard

The snapshot combines indicators with different reference periods, shown alongside the values. O = observation or estimate for the period; P = projection by the source. The shares and estimates do not all carry the same precision and should not be aggregated into a single index of power.

Four readings

AI and capital. The availability of funding is concentrated; the economic test lies in final demand and the return on investment. The VC total and the private investment figure from the separate AI Index database are kept apart.

Energy. The IEA trajectory implies a near doubling between 2025 and 2030. The outcome remains conditional on capacity, timing and demand. AI is one component of data-centre consumption.

Italy. Adoption is growing rapidly, while the macro picture continues to leave little room. The potential for diffusion must turn into output and income, not merely technology spending.

Markets. Multiples differ between developed and emerging markets, but the discount is not a sufficient measure of value. Sectors, governance and earnings estimates all weigh on the comparison.

The dashboard retains the values in the report and does not receive real-time data. For decisions taken after the research date, the values should be updated from their respective sources. The accompanying spreadsheet allows observations, assumptions and formulas to be checked without reconstructing them from the charts.

35. Methods, formulas and limitations

Weights and matrix

The initial weights are 35, 25, 20, 10 and 10. If they are changed, each normalised weight is the raw weight divided by the sum of the raw weights. The case in which all weights are zero is invalid and produces no ranking. A theme's score is the sum of the products of the normalised weight and the sensitivity in each scenario. The qualitative scale does not permit the result to be interpreted as utility, probability or expected return.

Debt simulation

The starting point is 138.2% in 2026. The settings illustrate nominal growth of 3%, an average nominal cost of 3.5%, a primary surplus of 1.5% of GDP and zero stock-flow adjustments. The annual recurrence is applied nine times, up to 2035. The average cost is held constant for simplicity; the model does not incorporate refinancing, endogenous inflation, recessions, fiscal reactions or market risk. It is not an IMF forecast.

Valuation and company example

The initial price in the example is 40 times current earnings. After five years, with earnings growing at 15%, the final price is final earnings × final multiple. The annualised price return is (final price / initial price) raised to the power of 1/5, minus 1. For the industrial example, the net benefit is cost of defects × reduction, minus recurring costs; the simple payback is investment divided by benefit, where the latter is positive.

Conventions on sources

FX reserves exclude gold. Private AI investment and venture capital have different perimeters. Robot installations are flows, not stocks. Italian and European AI adoption refers to enterprises with at least ten employees in the sectors covered by the respective surveys. BLS productivity is distinct from TFP; the +1.4% quarterly annualised figure should not be confused with the +2.2% annual figure. MSCI multiples are dated 31 August 2026.

Limitations of the research

The selection of sources is targeted, not an exhaustive systematic review of the literature. Microeconomic evidence does not by itself identify a macro effect; working papers may be revised. No geopolitical probabilities, expected returns, covariances or fair values of individual securities have been estimated. The analyses from 2035 to 2045 are exploratory.

The sources report data available up to the research date, with lags and possible revisions. The analytical content must therefore be updated as evidence, prices or institutions change, keeping a record of the earlier assumptions.

Instrument protocol

ISINs, costs and characteristics come from the issuer documents indicated in sources 28-43 and 48-49. TER and OCF are not a total cost of ownership: for GRID the KID reports 0.63% in recurring costs and further transaction costs estimated at 0.08%.^[43] The preferences by scenario are the report's judgements; they are not derived from estimated expected returns. Availability through intermediaries, executable spreads and tax conditions have not been verified. No dashboard filter automatically endorses a purchase.