Cerebras IPO Debut Shows AI Bubble Peak: NVDA, SMCI, ARM Overbought vs Historical Comparisons

Cerebras' stunning IPOInitial Public Offering - a company's first public sale of stock. debut on Thursday sent a stark signal to market watchers: we are in the speculative mania phase of the AI cycle, not the rational growth phase. The Nvidia competitor saw shares skyrocket on opening, drawing comparisons from Wall Street analysts to historical bubble peaks. One measure showed AI chip valuations now surpassing Nasdaq valuations during the dot-com peak, rivaling even the 1700s tulip-mania proportional excesses. This is not hyperbole; it is mathematical observation from serious institutional investors.

The risk is that retail and institutional traders are chasing AI chips because of FOMOFear Of Missing Out - buying because others are profiting. rather than fundamental analysis. Cerebras is pre-revenue relative to its valuation, yet it opened with a moonshot debut. NVIDIA, by contrast, has actual earnings and market dominance, but is now priced for a perpetual bull case with zero room for normalcy. Super Micro Computer (SMCI) has rallied on pure AI capex enthusiasm despite recent accounting-scandal remediation efforts. ARM, the architecture licenser, has benefited from the entire AI buildout narrative but is now at similar valuation multiples to peak dot-com levels.

The broader semiconductor sector is showing cracks beneath the surface. The SOXX (semiconductor ETFExchange-Traded Fund - a basket of securities trading like a single stock.) is at overbought extremes, with RSIRelative Strength Index - momentum oscillator on a 0-100 scale. levels not seen since 2021. Breadth inside the sector is narrowing; smaller, profitable chip makers are underperforming while AI-euphoria plays like NVIDIA and SMCI lead higher. This is a classic late-cycle leadership pattern: speculative hot money chasing the newest, coolest story rather than rotating into value or stability.

The counterargument is that AI infrastructure is not a bubble because the capex is real and the returns on computing are structural, not cyclical. If that thesis holds, current valuations are merely expensive, not irrational. But history shows that even genuinely transformative tech booms (railroads, autos, internet) experienced 50-80% drawdowns from peak to trough during consolidation phases. The timing of such drawdowns is unknowable, but the pattern is predictable.