Nvidia Gains Despite Custom AI Chip Rivals

What happened

Nvidia shares rose on June 25, 2026, despite headlines that hyperscaler competitors are rolling out custom AI chips—a risk already embedded in the DeepValue report's WAIT rating. The market shrugged off new specific examples of custom silicon (Microsoft's Maia 200, Google's Ironwood TPU, Amazon's Trainium) as the narrative remains fixated on sustained AI capex. However, the report highlights that these custom chips are already scaling: Microsoft has Maia 200 live, Google made Ironwood TPU generally available, and Amazon disclosed Trainium exceeded 50% of its AI chips. The stock's resilience reflects expectations that Nvidia's platform performance lead and roadmap cadence (Blackwell Ultra, Rubin) will maintain share, but the long-term risk of design-out acceleration remains. The news does not alter the report's base case or dangerous downside scenario, as competition was already a key risk factor.

Implication

The stock's uptick on this news suggests the market continues to prioritize hyperscaler capex momentum over competitive risks. However, the DeepValue report assigns only a 20% probability to the bull case where Nvidia maintains dominance, and competition is a key failure mode. Custom chips from Amazon, Google, and Microsoft are already operational, and each new deployment increases the chance of substitution at the margin. The report's WAIT rating with an attractive entry of $180 offers a margin of safety relative to today's $208.65. Until the next quarterly results confirm stable margins and declining inventory provisions, investors should avoid chasing the stock and use any rally as an opportunity to trim toward the target trim above $235.

Thesis delta

The news reinforces our existing thesis that competition from custom silicon is intensifying as expected. No material shift in probability-weighted value; the bear case risk of design-out acceleration remains live. The stock's positive reaction underscores the market's focus on capex, which could be a trap if hardware deployment constraints or transition issues materialize.