Semiconductors & the Bubble Fault Line – Episode 2

Semiconductors are at the centre of the artificial-intelligence investment cycle.

They enable the demand for compute, memory, equipment and advanced infrastructure that has powered the AI build-out. But after years of extraordinary performance, the question is no longer only whether the long-term fundamentals remain attractive.

The question is whether parts of the semiconductor capital-equipment complex have moved into late-stage market behaviour.

In Episode 2 of our Artificial Intelligence mini-series, LFG+ZEST applies a quantitative framework to selected semiconductor names, combining a 40-year trend-residual analysis with a formal log-periodic power-law model.

Both lenses point in the same direction.

The current price extension is above the levels reached during the 2000 dot-com peak, with the selected names standing around 3.0–3.3 standard deviations above their historical mean deviation. The LPPLS model also recovers a log-periodic bubble signature across the 2020–2026 run-up, with the model-implied critical time clustering close to the present.

This is not a crash forecast.

Deviation is a measure of stretch, not a countdown. The signal is one of elevated fragility and reduced margin of safety.

The semiconductor theme remains structurally important, especially in the context of AI infrastructure. But when a market segment reaches a 40-year relative extreme, investors should treat it as late-cycle: size carefully, hedge for tail risks and confirm the signal across the broader complex.

Watch the video summary and follow the “contact us” below to receive the research in PDF.