In April 2026, German economists at Makronom reignited a long-simmering debate: the persistent conflation of gross domestic product (GDP) with national wealth creates dangerous illusions of prosperity, particularly as automation, AI-driven productivity gains, and ecological degradation distort traditional metrics. The core issue isn’t merely academic—it shapes fiscal policy, investment decisions, and public perception of economic health in real time. By failing to subtract capital consumption (depreciation) from GDP to arrive at net domestic product (NDP), policymakers overstate sustainable output, mistaking asset depletion for growth. This gap has widened significantly in the AI era, where intangible investments in software, data, and model training are poorly captured in national accounts, yet their rapid obsolescence accelerates hidden wealth erosion.
The Illusion of AI-Driven Growth
Modern GDP calculations treat expenditures on AI infrastructure—such as GPU clusters, TPU pods, and energy-intensive data centers—as immediate contributions to economic output. Yet these assets depreciate at alarming rates: semiconductor hardware faces rapid obsolescence due to Moore’s Law variants evolving into “more-than-Moore” specialization cycles, whereas foundational AI models can lose competitive relevance within 6–18 months as newer architectures emerge. The Bureau of Economic Analysis (BEA) estimates that software and R&D depreciation now accounts for over 30% of fixed asset consumption in the U.S., up from 18% in 2010. In Germany, where industrial AI adoption is surging in manufacturing and logistics, the Federal Statistical Office (Destatis) reports that capital depreciation in ICT assets grew 4.2% year-over-year in Q1 2026, outpacing gross fixed capital formation by 1.8 percentage points.
This distortion is exacerbated by the treatment of AI training costs. When a tech firm spends €50 million training a large language model (LLM) on proprietary data, GDP registers that sum as immediate investment. But if the model becomes obsolete within a year due to architectural breakthroughs—say, a shift from dense transformers to mixture-of-experts (MoE) sparsity patterns or test-time compute optimization—the full economic value may never be realized. As the U.S. BEA notes, current national accounting frameworks lack mechanisms to amortize AI-specific intangibles aligned with their actual economic lifespans, creating a systematic bias toward overstating net wealth.
Why Net Domestic Product Matters More Now
NDP—GDP minus depreciation—offers a clearer picture of whether an economy is truly maintaining or expanding its productive capacity. In ecological economics, this concept extends to “genuine progress indicators” (GPI), which subtract environmental degradation and resource depletion. But even within conventional frameworks, ignoring depreciation risks policies that accelerate capital consumption under the guise of stimulus. For example, tax incentives for rapid AI infrastructure rollout may boost quarterly GDP figures while hastening the need for premature reinvestment—a cycle that benefits equipment vendors but strains long-term fiscal resilience.
Consider the energy dimension: training a single GPT-4-class model consumes approximately 50 GWh of electricity, equivalent to the annual power use of 4,600 U.S. Households. Much of this energy is sourced from grids still reliant on fossil fuels, imposing deferred ecological costs not reflected in GDP. When depreciation of energy-intensive hardware is ignored, these externalities compound the illusion of clean growth. As Dr. Katharina Kohl, Chief Economist at the German Institute for Economic Research (DIW Berlin), stated in a recent interview:
“We are mistaking the churn of AI hardware cycles for economic advancement. Every euro spent replacing yesterday’s GPU cluster with today’s TPU pod shows up as investment—but if we don’t account for the speed of obsolescence, we’re measuring turnover, not wealth.”
The Data Gap in AI-Era National Accounts
One of the most acute information gaps lies in the valuation of data itself. While GDP captures spending on data collection and storage, it does not recognize data as a capital asset that generates ongoing returns—yet it also fails to measure its depreciation through decay, regulatory obsolescence (e.g., GDPR-driven deletion mandates), or competitive devaluation. A 2025 study by the OECD found that data-driven intangibles now contribute up to 15% of value added in digital-intensive sectors, yet remain largely invisible in standard productivity metrics. Without adjusting for data asset depreciation, NDP calculations remain incomplete, especially as synthetic data generation and foundation model distillation alter the lifespan and utility of training corpora.
This has real-world implications for sovereign wealth funds and pension managers assessing national solvency. If a country’s reported GDP growth is fueled by short-lived AI investments whose depreciation is undercounted, its apparent creditworthiness may be inflated. Rating agencies like Moody’s and S&P Global are beginning to scrutinize “quality of growth” indicators, including capital productivity trends. NDP serves as a leading indicator of sustainable fiscal space—more so than GDP alone.
Bridging the Policy-Economics Divide
Closing this gap requires upgrades to national accounting systems that reflect the accelerated turnover of intangible capital in the AI age. The United Nations System of National Accounts (SNA 2025) is currently under revision to better capture digital economy transactions, including cloud computing and AI services. Proposals include: shortening assumed lifespans for AI-related software from 5 years to 2–3 years; creating satellite accounts for AI training expenditures; and integrating physical depreciation metrics (e.g., FLOPS/watt decay in hardware) with economic ones.
Critically, these adjustments must avoid becoming tools for gaming metrics. As Dr. Lars Müller, a senior researcher at the Institute for Fiscal Studies in Bonn, warned:
“Any reform to depreciation schedules must be transparent and rules-based. If we allow politically driven adjustments to extend asset lives artificially, we replace one illusion with another—this time, the illusion of sustainability.”
The path forward lies not in abandoning GDP, but in using NDP as a complementary metric—one that asks not just how much we produced, but how much we truly preserved. In an era where AI can generate synthetic output at near-zero marginal cost, distinguishing between real wealth creation and accounting artifice has never been more urgent.
