What if there's an overinvestment in AI? (www.mbi-deepdives.com)

The piece argues that overinvestment in AI is all but inevitable because major tech firms prefer to err on the side of spending rather than risking missed opportunity—an incentive structure that breeds boom-bust capex cycles. Citing Byrne Hobart’s logic and the telecom/2000 internet bust (AT&T fell ~65% from peak despite “reasonable” P/E), the author warns that generous AI capex could leave shareholders exposed even when headline valuations don’t look extreme. To illustrate magnitude, the author analyzes Meta’s capex history in three phases: 2010–17 (every $1 capex ≈ $2–$2.5 revenue), 2018–24 (≈ $1/revenue), and 2022–24 where cumulative capex was ~$95B versus ~$47B incremental revenue—implying a normalized 0.6x revenue-to-capex ratio and roughly $15–20B of non-revenue AI capex in 2022–24. Looking ahead, consensus estimates imply the ratio could fall to ~0.3x; using 2026 as an example, $32B projected revenue growth would require ~$53B capex at 0.6x but consensus capex is ~$97B, suggesting ~ $44B of additional non-revenue investment. The upshot: if AI investments underperform or macro ad growth disappoints, big tech (and GPU suppliers like Nvidia) could face meaningful drawdowns—specific downside modeling is deferred to the author’s paid analysis.