In 1987, economist Robert Solow famously observed, “You can see the computer age everywhere but in the productivity statistics.” Nearly forty years on, as we approach the end of 2025, that paradox feels very familiar.
We are witnessing one of the largest infrastructure build-outs in modern economic history. From hyperscale campuses in Nevada to wind‑powered data centres across the Nordics, capital is flowing at unprecedented speed into the physical core of artificial intelligence. Yet the macroeconomic data tells a different story: Total Factor Productivity (TFP)—the measure that truly captures the efficiency of labour and capital—remains flat.
This widening gap between visible investment and invisible returns is reopening Solow’s paradox for the generative AI era. This week, we explore the forces behind this disconnect—and what pragmatic leaders must do to de-risk their AI strategy in 2026.
Focus On: The Trillion-Dollar Bet Meets the Productivity Gap
The scale of capital expenditure fuelling the AI wave is unprecedented. JPMorgan estimates that global data centre and AI infrastructure investment could represent $5–7 trillion by 2030—an industrial-scale reinvention of the digital economy in less than a decade.
But Q4 2025 has delivered a sobering reality check: while the build-up of the AI economy is rising at lightning speed, the corresponding productivity dividend is not yet fully realised.
The OpenAI Situation
Nowhere is this tension more visible than in the economics of leading model providers. HSBC’s latest financial modelling on OpenAI suggests that despite strong revenue growth, the company could face a funding gap of up to $207 billion by 2030.
Why? Because compute costs are scaling as fast—or faster—than revenue. OpenAI’s ambition to secure 36 gigawatts of compute capacity requires a capex profile closer to energy utilities than software companies.
For partners such as Microsoft and suppliers such as Nvidia, this is a lucrative tailwind. But all other enterprises have to confront the reality that even the market leader is struggling with unit economics.
This is not to diminish the revolutionary potential of generative AI. But the caricature of a “money pit with a website on top,” while exaggerated, tells us a deeper truth: as of today, the cost of running intelligence still outpaces the value it generates.
The Macro Illusion
The illusion of momentum becomes even more apparent when zooming out to the macro data. According to economist Jason Furman, US GDP growth in the first half of 2025 was disproportionately driven by investment in data centres and intellectual property.
Take out these AI-preparatory investments and GDP growth falls to around 0.1%—essentially flat.
This reveals a critical instability in the current economic narrative: growth today is being fuelled by preparing for AI (building the factories), not benefiting from AI (unlocking productivity).
If downstream gains fail to materialise quickly enough to justify this investment-led cycle, the correction could be sharp.
Signs of fragility are already emerging:
Credit default swaps for major tech firms such as Oracle are widening.
Corporate bond markets are repricing risk for companies over-leveraging to fund AI capex.
Power grid congestion has become a binding constraint in several US states and European markets.
The smart money is beginning to differentiate between hype-fuelled investment and productivity-backed value creation.
The Enterprise Reality: Adoption ≠ Transformation
Zooming into the enterprise, the picture becomes both clearer and more concerning. According to McKinsey’s State of AI 2025, adoption is almost universal—but meaningful transformation remains rare.
Employees across Fortune 500 companies use GenAI to summarise meetings, generate content, and support coding tasks. Yet very few organisations have achieved double-digit productivity gains at the process level.
Why? Because:
Adoption is easy; transformation is hard.
Most organisations suffer from integration indigestion:
fragmented data
legacy workflows
weak automation foundations
governance models not designed for autonomous agents
If you want to learn more about this, have a look at previous issues such as AI Agents and Automated Operations and The AI-First Company, where we highlighted that transformative outcomes require re-architecture, not just tooling.
De-Risking the Journey
To navigate this Solow-style moment, leaders must pair ambition with discipline. The objective is simple: avoid stranded capex while capturing productivity gains the moment they become real.
1. Pace, Proofs, and Portfolio Discipline
Shift from “AI everywhere” to stage‑gated investment. Fund only what demonstrates measurable progress against operational KPIs such as:
cycle-time reduction
error-rate reduction
first-time-right metrics
cost-to-serve improvements
If a pilot cannot show uplift at the process level, it does not earn the right to scale.
2. Demand Contract Optionality
The OpenAI–HSBC analysis highlighted how dangerous multi-year, fixed‑cost compute contracts can become.
Enterprise leaders should:
negotiate renegotiation levers
avoid hyperscaler lock-in
build multi‑cloud or hybrid optionality
In a volatile market, flexibility is strategic insulation.
3. Focus on “Boring” Productivity
Generative creativity may dominate headlines, but the strongest ROI lives in the back office:
finance reconciliation
legal review
IT service management
procurement automation
These are precisely the domains where TFP impact is measurable, repeatable, and defensible.
Wins here should be used to self-fund higher‑risk initiatives in customer experience or R&D.
4. Measure Micro to Move Macro
Do not wait for TFP to appear in quarterly earnings. Productivity gains first appear locally, not systemically.
Track:
time-to-resolution in service workflows
cash conversion cycle improvements
reduction in manual handoffs
reduction in rework
These micro‑gains will eventually roll up into macro performance—but only if leaders measure and act early.
Follow me
That’s all for this week. To keep up with the latest in generative AI and its relevance to your digital transformation programmes, follow me on LinkedIn or subscribe to this newsletter.
Disclaimer: The views and opinions expressed in Chronicles of Change and on my social media accounts are my own and do not necessarily reflect the official policy or position of S&P Global.