The Great Productivity Gamble: Trillions Of AI Dollars vs Global Economic Headwinds

AI adoption will create powerful deflationary effects, but companies must demonstrate measurable productivity gains to separate reality from hype.

---

Credit: Alexander Mils, Unsplash

I'm Raphaëlle d'Ornano, founder of D'Ornano + Co., an international financial analysis and strategic advisory firm. Our team developed Advanced Growth Intelligence (AGI)—an analytical framework that helps investors and businesses unlock value by decoding the discontinuities reshaping industries, particularly those driven by generative AI. My work bridges the gap between cutting-edge AI research and real-world business impact, translating complex technological shifts into actionable insights for better investment decisions.

I'm grateful that readers have made this one of Substack's top technology newsletters, and humbled by Michael Spencer's recent recognition as one of his top AI newsletters of 2025.

---

Despite repeatedly promising for months to impose tariffs, the size and scope of President Trump’s attempt to reverse the rules of global trade through high tariffs against dozens of nations caught even many of his staunchest supporters by surprise. The subsequent meltdown in global markets is expected to continue in the next weeks.

Yet, that same week, OpenAI announced a record private funding round. The logic behind OpenAI's $40 billion fundraise at a $300 billion valuation is difficult to reconcile with the current economic landscape—at first glance. As tariffs escalate following last week’s “Liberation Day” and economists downgrade growth forecasts, investors are pouring unprecedented capital into artificial intelligence.

This apparent contradiction reveals a bold wager: AI adoption will accelerate because its productivity gains will create powerful deflationary effects in the coming years. While AI is unlikely to offset the inflationary pressures of trade wars in the short term, over the long term, it should create powerful operating leverage.

Still, this wager seems increasingly precarious. While AI's theoretical potential to transform the $70 trillion annual cost of global wages is substantial, the path to realizing these gains is neither uniform nor guaranteed. For investors navigating this contradictory landscape, distinguishing between companies that can translate AI investment into productivity improvements and those that cannot will prove essential during the coming economic turbulence.

The Infrastructure Investment Paradox

The scale of current AI investment defies conventional economic wisdom.

OpenAI's record-breaking financing comes as the OECD has revised US growth estimates downward to 2.2% in 2025 and just 1.6% in 2026, with core inflation expected to exceed 3%. Wall Street economists have issued even bleaker forecasts, suggesting growth could fall to 0.6% while inflation approaches 4.7%. It should be noted that many of these revisions came in the weeks before the sweeping new U.S. tariff policy was unveiled.

Yet capital continues to flow into AI infrastructure, and the exuberance towards the AI sector is unchanged. Per Pitchbook data, AI startups captured more than half of the newly invested capital in venture capital over Q1. The data center buildout has emerged as one of the most significant industrial investments globally, with companies racing to secure capacity for AI model training. CoreWeave’s recent IPO reflected that trust.

The global hyperscale data center market is forecast to grow from $62.79 billion in 2024 to $608.54 billion by 2030. Of that $40bn raised by OpenAI — $18 bn is earmarked for infrastructure projects such as Stargate. Many of the largest cloud and AI companies — Google, Amazon, Microsoft, and Facebook — announced giant increases in Capex spending for 2025.

The AI infrastructure investment boom represents perhaps the clearest illustration of the Discontinuity that has been unleashed by generative AI. The previous “conventional wisdom” and the playbooks that had guided even the Big Tech leaders are now obsolete.

In this case, these companies are struggling to build the capacity they need to meet projected usage. As Google CFO Anat Ashkenazi said on an earnings call earlier this year: "We exited [2024] with more demand than we had available capacity. So, we are in a tight supply-demand situation, working very hard to bring more capacity online."

OpenAI’s Sam Altman has stated publicly that the company’s “GPUs are melting” and that access to more infrastructure — including data centers, chips, and energy — is the biggest constraint on its growth.

These are not problems that can be solved overnight. Building any data center, even in the most regulatory-friendly circumstances, requires years to secure permits, construction, and get additional power generation online. Recently, David Hogan, Nvidia’s European sales chief, said that access to power is now “the biggest limiting factor” for AI developers as he warned UK leaders that they needed to make massive investments in nuclear power if the country wants to remain competitive.

Amid the current market turmoil and new uncertainties around the costs of procuring all the necessary data center components, these companies have no choice but to look years in the future and make investment decisions today based on the projected adoption curve of AI.

This investment surge represents a massive bet on a single proposition: that AI-driven productivity will serve as the primary counterbalance to inflationary pressures in an otherwise deteriorating economic environment.

Just as infrastructure investors can’t afford to put their plans on hold, the ultimate users of these services must reckon with their own Discontinuity. The traditional playbooks – especially those used during recessionary times – are no longer suited to a moment when these companies face tremendous risk to their competitiveness if they don’t invest in AI.

The Agent Adoption Revolution

While much of the initial AI focus centered on automating discrete tasks, the emergence of AI agents represents a fundamental shift with considerably greater economic implications. Unlike earlier applications that required continuous human guidance, these autonomous systems can execute complex, multi-step workflows independently.

The recent advent of the Model Context Protocol (MCP) marks a critical inflection point in agent capabilities. As I discussed in my earlier analysis of the Manus framework, MCP functions as a standardized interface—effectively a “USB-C port for AI applications" —that enables agents to interact seamlessly with diverse software systems, data repositories, and business processes. This protocol allows agents to navigate autonomously between applications, access specialized tools, and execute workflows spanning multiple systems without continuous human oversight.

Consider the potential transformation in financial services. Traditional AI might assist analysts in generating reports or detecting anomalies in transaction data. By contrast, an AI agent enhanced with MCP capabilities can autonomously monitor markets, identify investment opportunities based on predefined criteria, execute trades, adjust portfolio allocations, generate client reporting, and communicate recommendations—all without human intervention beyond initial parameterization. We are not there yet -but just envision this for a second.

Similar transformations are emerging across industries. ServiceNow has pioneered this approach with its AI Agent platform, enabling organizations to deploy autonomous agents across IT, customer service, and HR functions. ServiceNow’s system allows agents to "collaborate, grounded in human knowledge, to deliver seamless, end-to-end workflow orchestration across systems and domains" while "autonomously solving complex business challenges."These agents access an organization's existing data, workflows, and integrations, executing complex processes while continuously learning and adapting.

This represents an order-of-magnitude increase in potential labor substitution. Goldman Sachs had estimated that AI could eventually substitute up to one-quarter of the current work globally, affecting the equivalent of 300 million full-time positions and trillions in wage expenditure. But that was before agentic deployment and more of a consideration of what % of global work was automatable.

The economic implications are profound. In advanced economies, where labor represents the predominant cost component across most industries, successful agent deployment could dramatically expand margins. AI agents can operate continuously, maintain consistent performance, eliminate human variability, and scale near-instantaneously across an organization.

In addition, agents have the potential to fulfill jobs that are not served by humans today for diverse reasons. The medical sector is seeing massive potential for revenue expansion in the next years as agents take on jobs that were left vacant. Hippocratic.ai, a GenAI startup in the field of AI agents for the healthcare industry, was able to close a large fundraising round in January 2025, reaching unicorn status, with that promise, as their Co-founder & CEO Munjal Shah explained at last month’s Montgomery Summit.

This could profoundly transform P&L statements. There will be lots of talk about potential increased costs for deploying such agents in the coming weeks. But any business leader will have to balance that against the cost of a competitor reaping the benefits and gaining a huge competitive advantage.

Consider the recent memo by Shopify’s CEO in which he mentions the state of reflexive AI as a baseline expectation at Shopify. In a bold move, he states: “Before asking for more Headcount and resources, teams must demonstrate why they cannot get what they want done using AI. What would this area look like if autonomous AI agents were already part of the team? This question can lead to really fun discussions and projects.”

Despite these potential benefits, organizations face significant hurdles in AI implementation. Legacy systems, data quality issues, regulatory concerns, and workforce transitions all represent substantial friction points that could delay or diminish expected productivity gains. The gap between theoretical potential and practical realization remains a critical consideration for investors assessing the timeline for AI-driven margin expansion.

The Measurement Imperative

Despite this potential, evidence suggests that most organizations remain ill-equipped to translate AI investment into tangible and demonstrated productivity improvements. The market has largely failed to address a fundamental question around execution: what separates companies that will successfully implement AI from those that will not?

The answer lies primarily in measurement frameworks. Companies establishing rigorous systems for evaluating AI's impact demonstrate substantially better outcomes than those implementing technology without clear metrics. Yet such measurement remains remarkably rare.

This measurement deficit creates both market inefficiency and investment opportunity. Organizations claiming "exponential productivity gains" from AI typically lack empirical validation for these assertions.

I recently published a 24-page teardown of Klarna’s IPO prospectus (see a preview here). One of the key takeaways from that analysis: “Klarna provides limited quantifiable evidence of AI's impact beyond customer service automation. The prospectus fails to substantiate how AI enhances critical financial metrics such as credit underwriting performance, take rate expansion, or customer acquisition efficiency. The absence of concrete metrics raises questions about the depth of Klarna's AI transformation beyond surface-level applications.”

Companies successfully capturing AI's value implement multi-dimensional measurement systems that quantify impacts at the process, department, and financial levels. They conduct controlled experiments comparing AI-assisted, AI-automated, and traditional approaches to objectively assess performance differences. Most critically, they directly connect technological implementation to margin expansion, providing unambiguous evidence of value creation.

Investment Implications

For investors navigating this contradictory AI adoption landscape, there are several signals to watch closely:

1. Scrutinize AI investment theses. Generalized exposure to "AI themes" is now insufficient. Investors must differentiate between companies with robust measurement frameworks demonstrating tangible productivity improvements and those implementing technology without clear metrics. Earnings reports for last year still show little maturity in this respect. By Q2, earnings statements should better incorporate them, with this being a baseline expectation for 2025 earnings.

2. Sector differentiation will intensify. Knowledge-intensive industries, including financial services, pharmaceuticals, and professional services, will likely capture greater value from AI than manufacturing-based sectors. But some consumer goods companies are leading the charge here and could turn out to be strong beneficiaries of AI-related operating leverage. Even within industries, substantial performance variations will emerge based on implementation capabilities rather than mere technology adoption.

3. Business models (and notably revenue models) will evolve to reflect these margin gains. Just how remains a question. Margin gains must be over the long term to create shareholder value. (See Scott Belsky’s excellent analysis of how AI can make business models antiquated.)

4. Monitor agent integration. Effective integration of “AI assistants” into the workforce provides the clearest indicator of successful AI implementation. Companies demonstrating sustained margin improvement directly attributable to agent deployment represent the most compelling investment opportunities in this space. This direct connection between technological adoption and financial performance cuts through the noise of AI hyperbole.

5. Discipline will win. The "unlimited investment" approach to AI that characterized stronger economic conditions will prove unsustainable as growth slows. Companies with disciplined allocation frameworks will maintain high-value AI initiatives while eliminating unproductive experiments, protecting profitability during economic contraction.

6. Democratization shifts competition. Competitive dynamics will shift as AI capabilities become more democratized. DeepSeek's open-source approach could enable smaller companies and emerging market players to accelerate innovation. Established players relying on proprietary technology advantages may find these eroded more rapidly than anticipated.

The Path Forward

As recessionary pressures mount from trade tensions, the companies best positioned to weather the storm will be those translating AI investments—particularly in autonomous agents—into demonstrable productivity gains and margin expansion. For sophisticated investors, the opportunity lies not in broad AI exposure but in identifying specific organizations with measurement discipline rigorous enough to separate AI reality from hype.

The coming economic turbulence will expose which companies have built foundations for genuine productivity transformation and which have merely adopted fashionable technology. Those that can demonstrate AI's deflationary impact through margin expansion will find themselves uniquely positioned to outperform in an otherwise challenging environment.

In resolving the apparent paradox of massive AI investment amid economic uncertainty, the winners will be those who move beyond the infrastructure investment race to demonstrate measurable, productivity-driven margin expansion that can offset broader inflationary pressures. And this will happen both in the technological and in the material worlds. The question for investors is not whether to invest in AI, but how to identify the organizations capable of transforming technological potential into financial performance.

---

As markets navigate the complexities of generative AI, autonomous agents, and evolving infrastructure, I'm expanding this newsletter with new options to meet diverse information needs:

1️⃣ Free: Weekly insights on macro trends and AI disruption.
2️⃣ Pro: Deep dives on Durable Growth Moat frameworks and sector strategies.
3️⃣ Elite: Executive format analysis of pre-IPO S-1s and company 10-Ks
4️⃣ Alpha: Invitation-only access with comprehensive pre-IPO teardowns, thematic briefings, custom deal sprints, and strategic consultation. To apply for Alpha access, contact me.

You can find more information here on these tiers.

Raphaëlle D’Ornano