Welcome to LB MACRO on Substack

This is our first article on Substack, and it is freely accessible.

Over the past months, we have developed a Web Portal and Mobile App where we publish daily notes, economic and financial analyses, central bank monitoring, and shortly our proprietary forecasts and indicators.

Subscribers on either platform receive full access to our research ecosystem.

All relevant links are available in our profile description.

Overview

The debate over the effects of Artificial Intelligence (AI) on the US and the global economy in general has intensified markedly.

What is beyond dispute is that the AI technological revolution will be transformative.

However, opinions diverge on how it will reshape economic growth, transform job markets and redefine the quality of life across societies.

In our view, it is wise to resist the allure of apocalyptic narratives.

While they make for stimulating intellectual discourse, the long-time history of technological revolutions has consistently failed to validate them.

On the contrary, each wave of innovation has proven to be much more of a blessing for prosperity than a curse.

At their core, technological breakthroughs are positive productivity shocks: they lower costs, expand output and raise real incomes.

History shows that this “income effect”, making the income pie larger, has always outweighed the “substitution effects” of human labour.

AI is going to be no exception, although surely there will be winners and losers, as well as risks, mainly linked to financial fallouts: from credit to bubbles.

In this note, we provide some facts supporting the following conclusions:

i) US “exceptionalism” is grounded on solid foundations, based on human capital and technological progress.

ii) AI is exacerbating the US divergence, boosting productivity more than any other previous technological shock.

iii) The US’s role at the forefront of the AI revolution is driving huge capital investments, far exceeding peers.

iv) AI is affecting the US labour market more in terms of labour reshuffling than labour shedding.

v) Risks stem from high leverage and easy monetary policy.

vi) Although we are not forecasting a crash, should it occur, its economic fall-out would exceed that of the DotCom crash.

Details

1. Economic Growth

Until 2008, the US and the EMU aligned in terms of nominal GDP expressed in dollars: you could say the two economies were equivalent.

Since then, ie from around the Great Financial Crisis, the US has dwarfed the EMU, being more than 70% larger (Fig. 1).

This is the result of the combination of:

i) The price of the US produced goods and services increased way more

The GDP Deflator - expressed in dollars, thus encompassing the FX effect - has risen since 1991 by 110% in the US and by 77% in EMU (Fig. 2).

In other terms, the value of the output produced in the US has increased way more that the output produced in the EMU.

In particular, US technology products and knowledge-intensive services provided a decisive contribution.

ii) An even more striking US outperformance in volume terms

Indeed, since 1991 US real GDP has grown roughly 140%, more than double the EMU’s 65% expansion (Fig. 3).

Europe’s recoveries have been shallower, its expansions shorter, and the output losses from successive crises largely permanent.

In turn, the main contributor to the US outperformance in volume terms was Productivity (Fig. 4).

Indeed, over the same period, Output per hour worked has risen 80% in the US, compared to barely 35% in the EMU.

Moreover, and most importantly, the gap is widening after ChatGPT launch at the end of 2022, as US productivity accelerated above its pre-AI trend.

Fig. 1: The US Dwarfs the EMU in Nominal Terms ...

Fig. 2: ... As Its Output Increased More both in Price ...

Fig. 3: ... And In Volume Terms ...

Fig. 4: ... The Latter Mainly Thanks to Productivity

2. US Labour Productivity Boosted by a Sequence of Technological Revolutions

Unlike Europe, the US has leveraged each technological wave over the last forty years into sustained efficiency gains.

Indeed, each major innovation cycle lifted productivity growth above trend, before gradually fading as the technology and its diffusion matures (Fig. 5).

The Internet revolution, connecting businesses, consumers and markets in real time, ignited a major acceleration in the late 1990s.

The “Control” wave of the mid-2010s, driven by the Cloud, IoT and Industrial Automation, opened a new cycle after the prolonged post-GFC slowdown.

With AI, a new productivity cycle is taking hold, one that is showing the strongest early-stage gains of the three (Fig. 6).

Europe has followed the opposite trajectory, with productivity growth in secular decline for three decades, largely unresponsive to any technological wave.

If history is any guide, the US is once again converting technological disruption into a durable productivity advantage. Europe, once again, looks behind.

Fig. 5: US Productivity Boosted by Technology

Fig. 6: AI Looks the Most Powerful Shock

3. Investments in Fixed Capital - The Compute Arms Race

Productivity gains require capital investments, and in the age of AI, that means computing infrastructure.

Cumulative US investment in AI Hardware - GPUs, CPUs and networking - has reached approximately $57bn, dwarfing the Europe’s $7bn (Fig. 7).

The gap has widened exponentially since 2022, as US Hyperscalers accelerated data centre build outs at a pace no other region can match.

China is absent from the investment data due to non-disclosure, but its presence surfaces clearly in the computing power figures.

According to Epoch AI, the US accounts for 75% of aggregate Supercomputing performance, up sharply since 2022, largely at China’s expense (Fig. 8).

US access to leading-edge chips, combined with tightening export controls that have restricted Chinese procurement, has tilted the balance decisively.

Europe’s tiny fraction reflects a structural absence of the large-scale technology firms that drive AI infrastructure investment.

Fig. 7: US AI Investments dwarf the Rest of the World

Fig. 8: ... As Does Computing Performance

4. Investments in Human Capital

Human capital lays behind the US advantage in terms of productivity, innovation capacity and physical capital.

Indeed, US Research & Development expenditure stands at 3.5% of GDP, up from 2.5% two decades ago, and well above the EU’s 2.2% (Fig. 9).

The output of this investment is visible for example in ICT Patent Applications, where the US generates more than double the EU’s volume (Fig. 10).

Innovation begets innovation, and the US has built a self-reinforcing ecosystem that Europe has been unable to replicate.

At the heart of this ecosystem sits the American university system.

US institutions hold 7 of the top 10 and 59 of the top 200 positions in global university rankings, dominating in science, engineering and computer science.

These are the breeding grounds for the tech companies that drive productivity, from Stanford in Silicon Valley to MIT at the AI frontier.

Fig. 9: The Power of US Human Capital, from R&D ...

Fig. 10: ... To Patent Applications

5. The Current State of the AI’s Impact on US Labour Market

The impact of AI on overall US jobs formation remains difficult to assess with precision at this stage.

AI-Affected Services* have shown a visible deterioration in employment in recent years (Fig. 11).

*(A gauge of AI-exposed BLS classification sectors: Professional & Business Services, Financial Activities and Information).

However, we are very far from the extent of job destructions due to recessions, and there are even signs of stabilizations in the last few months.

Indeed, as of now, job destruction seems confined to some pockets of the labour market (Fig. 12).

For instance, headcount among junior software developers has declined notably since the launch of ChatGPT, while seniors have continued on their trend.

This is consistent with firms substituting entry-level cognitive tasks with AI tools, though the sample remains narrow and the causality uncertain.

The scale of job losses in these sectors, however, remains contained and does not point to a meaningful deterioration in aggregate labour demand.

What looks unquestionably more relevant is the impact of AI on the pace of change in the Occupational Mix.

Dissimilarity indices show the composition of employment shifting faster than during the other major technological waves analysed (Fig. 13).

AI-Affected Services accounts for the bulk of the reallocation (Fig. 14).

Fig. 11: AI Impacted Jobs, but less than a Recession

Fig. 12: AI-Led Job Destruction in Identified Pockets

Fig. 13: AI Impact Strong on Occupational Mix ...

Fig. 14: ... Especially in AI-Affected Services

6. Risks Ahead: Financial Markets

As outlined in the Overview section, the primary risks associated with the AI cycle are financial in nature.

One of the main concerns is that the substantial investments in AI made so far won’t generate a sufficient Return on Investment (ROI).

Hyperscalers have seen Free Cash Flow compress as CapEx accelerates, a structural shift from capital-light to capital-intensive business models (Fig. 15).

In turn, the transition implies lower share buyback capacity going forward, which in recent years supported share prices, amplifying rises and limiting falls.

Still, as shown in the chart, Free Cash Flow is expected to remain largely positive this year.

These dynamics are weighing on equity prices and surfacing in corners of the credit market, with potentially broader implications for financial stability.

Oracle CDS spread, despite an outlier, has recently widened to 160bps, reflecting a weaker standalone credit profile relative to its peers (Fig. 16).

Further down the credit spectrum, High-Yield Tech spreads have diverged from the broader market around Anthropic’s Claude Cowork release (Fig. 17).

Moving from public to private credit, share prices of listed private credit firms and Software sold off in tandem following the announcement (Fig. 18).

Halted redemptions, forced asset sales and deep secondary-market discounts expose the private credit industry’s opaque liquidity and valuation risks.

The growing interconnection between private credit and the banking system makes this a financial stability risk worth monitoring closely.

Fig. 15: Cash Flow Compressed as CapEx Accelerates

Fig. 16: Oracle has a Weaker Credit Profile

Fig. 17: US HY Tech Spreads Widens in 2026 ...

Fig. 18: … Private Credit Heavily Exposed to Software

Although we believe the risks of a financial crash remain limited (see here), the potential economic fallout would far exceed that of the DotCom bubble.

Indeed, in early 2000, US market capitalization peaked at $15trn, or 143% of GDP, as internet mania drove the Nasdaq to a $4trn valuation.

Today:

1. The US market is much larger than in 2000: five times larger in absolute level ($70trn) and more than 50% larger as a ratio to US GDP (220%, Fig. 19)

2. Over the past 15 years, foreign investors have poured capital into US equities, which would spread the effects of a US crash globally

3. Households are more exposed to equity markets, as their direct equity holdings have climbed from 16% to almost 25% of their total wealth (Fig. 20)

Fig. 19: Market Capitalization At All-Time High ...

Fig. 20: ... Similar to Households’ Exposure to Equities

Authors

Luigi Buttiglione
LinkedIn | X

Andrea Pozzoli
LinkedIn

Disclaimer: LB MACRO commentary and forecasts are solely for information purposes; it is not investment research or advice and should not be regarded as an offer or solicitation to buy any instruments mentioned – it should be regarded as unregulated by the Financial Conduct Authority. LB MACRO does not accept liability for the use of, or reliance on, this document and/or its contents by any other party. No personal recommendation is being made to you – any strategies referred to may not be suitable for you and should not be relied upon as a substitute for your independent judgement. Past performance or future forecasts are not a reliable indication of future expectations and investors should be aware that losses and costs can occur on any investment. LB MACRO cannot accept liability for direct or indirect damages, or loss of returns in any way from information provided. LB MACRO is for the use of intended recipients only and should not be forwarded in any form without previous permission being given. Every effort has been made as to the accuracy and information from sources believed to be reliable, but the contents of this document have been made without any substantive study being undertaken into the sources. LB MACRO does not take responsibility for the accuracy or completeness of information provided in this document. By accepting this document you agree that you have read the above disclaimer and to be bound by it.