Apollo Global Management’s Chief Economist released a stark analysis on October 5, 2026, revealing that the current AI boom has yet to register in Total Factor Productivity (TFP) data. While hundreds of billions of dollars have flooded into data centers, chips, and model training, the economic metric that best proxies genuine technological progress remains stagnant. This disconnect challenges the prevailing narrative that AI is already fundamentally reshaping how businesses operate.

TFP vs. Capital Deepening

The report draws a critical distinction between TFP and labor productivity. TFP measures output gains from better efficiency and innovation, independent of increased inputs. In contrast, labor productivity (output per hour) can rise simply because workers are given more or better equipment, a process known as capital deepening. The data shows output per hour running near 2.5%, which is comfortably above the post-2005 average. However, this strength is accompanied by flat or negative TFP, indicating that the gains are driven by adding more 'machines' rather than making the economy smarter.

The Data Behind the Disconnect

Utilization-adjusted TFP from the San Francisco Fed, using the Fernald (2014) series, is currently sitting slightly below zero. This reading shows no sign of acceleration since the AI capital expenditure cycle began. Historically, TFP has swung between roughly -3% and +4% over the past 40 years with no discernible trend, so the current sub-zero reading is not unprecedented. The report notes that previous transformative technologies, such as electricity and IT, took a decade or more to show up in aggregate productivity numbers, suggesting we may be in a similar lag phase.

Implications for Dev Tools and Infrastructure

For builders and infrastructure teams, this data implies that the current AI investment is largely mechanical. Adding AI tools to a workflow may lift output per hour by providing better 'equipment,' but it hasn't yet proven to improve how inputs are combined at a systemic level. The productivity payoff from AI remains a forecast rather than an observation, which means companies should be cautious about attributing efficiency gains solely to AI adoption without seeing underlying TFP improvements.

Key Takeaways

- TFP is slightly below zero with no acceleration despite massive AI capex. Output per hour is near 2.5%, but this reflects capital deepening, not innovation. Historical parallels suggest it may take a decade for new tech to impact TFP. AI's productivity benefits are currently visible in investment data, not economic efficiency metrics.

The Bottom Line

We are buying more monitors, not better brains. Until TFP turns positive, AI is just expensive infrastructure, not a productivity revolution.