Meta's expanding cloud ambitions — building out its own AI infrastructure at scale — are reviving a persistent concern in the semiconductor space: that hyperscalers will eventually reduce their dependence on third-party chips, or that the pace of AI capex will outrun monetizable demand, creating an oversupply overhang. The headline rattled NVDA and AMD, both of which have been primary beneficiaries of the AI infrastructure boom.

NVDA reported FY2026 revenue of $215.9B, up 65.5% YoY, with a 71.1% gross margin and $4.90 diluted EPS — a business firing on all cylinders. AMD, more modestly, posted $34.6B in revenue (+34.3% YoY) with a 49.5% gross margin, still a meaningful acceleration. META itself put up $201B in revenue (+22.2% YoY), signaling it has the financial firepower to sustain heavy capex spending on its own silicon and data center buildout.

The bull case for semis rests on the idea that Meta's buildout is additive demand, not substitutive — more AI workloads require more chips, even if Meta develops some custom silicon. NVDA's dominance in training workloads and CUDA's ecosystem lock-in remain structural advantages that don't dissolve quickly.

The bear case is more nuanced: as hyperscalers like Meta, Google, and Microsoft deepen custom silicon programs (Meta's MTIA chip), the addressable market for merchant semis narrows over time. Combined with the possibility that AI monetization lags the infrastructure wave, inventory digestion cycles could hit NVDA and AMD at peak valuation multiples.

The key watchpoints are Meta's capex guidance updates, any incremental commentary on MTIA deployment scale, and whether NVDA's next earnings cycle shows any softening in hyperscaler order velocity. AMD's thinner margins and lower AI revenue concentration make it the higher-risk name if the oversupply narrative gains traction.