Despite significant investment and high valuations in artificial intelligence technology, meaningful productivity metrics indicate that AI is not fulfilling its efficiency promises. Torsten Sløk, chief economist at Apollo, noted in a blog post that while the AI boom is apparent through massive spending and soaring valuations, it is not reflected in productivity statistics such as the San Francisco Fed's "total factor productivity" (TFP) index. TFP assesses economic output relative to labor and capital, revealing whether companies are producing more without increasing their workforce or equipment.
Read More
Sløk's analysis shows that TFP is currently "slightly below zero" with no indication of improvement since the onset of the AI capital expenditure cycle. This statistic serves as an important indicator of whether the investments in AI translate into actual innovations or enhanced efficiency. Furthermore, as companies incur increasing costs associated with AI implementation, the necessity for demonstrable productivity gains intensifies.
Contrastingly, AI advocates often highlight the per-hour output from labor, which now exceeds the post-2005 average by 2.5%. However, Sløk warns that such figures can reflect various dynamics beyond technological advancement. For instance, capital deepening — providing better tools to employees — can also inflate output metrics. He illustrated this by noting that workers with access to additional monitors can increase productivity without indicating smarter operational strategies. Consequently, while individuals might be completing tasks more efficiently, the underlying operational processes may not have been optimized.