Nvidia's stock struggles as Kalshi traders bet chip prices are coming down

Nvidia shares are up about 12% in 2026 but have fallen roughly 3% over the past month, while B200 compute-per-hour pricing dropped from a three-month high of $6.11 on May 30 to $4.22 on June 21. Kalshi traders are turning pessimistic on further upside in Nvidia's chip pricing, even as RBC said GPU rental deals support Nvidia's positioning for 2H 2026 and 2027. The article points to softening near-term pricing power and weaker sentiment, partly offset by supportive long-term demand signals from AI infrastructure spending.

Analysis

The key read-through is not that AI spending is slowing, but that the mix is rotating away from frontier-training GPUs toward storage, networking, and capacity optimization. That matters because the next leg of capex typically rewards the picks-and-shovels beneficiaries before it re-accelerates for the incumbent accelerator vendor; memory pricing and cloud rental economics are a cleaner near-term signal than unit shipment commentary. If B200 rental rates keep softening while memory names stay bid, it suggests buyers are prioritizing utilization and balance-sheet flexibility over raw FLOPS, which is a subtle negative for NVDA margins and a positive for suppliers with tighter inventory cycles.

The second-order effect is that rising neocloud and hyperscaler rental activity can mask a share-loss narrative: even if GPUs remain the default, pricing power is no longer clearly with the OEM. The Google/SpaceX-style capacity contract also implies that large buyers are locking in supply months ahead, which can support visible demand for the ecosystem without immediately lifting spot compute prices; that usually favors companies with commodity-like exposure to the buildout, not the architect. For NVDA, the risk is not a collapse in demand, but a period of multiple compression if investors conclude the scarcity premium is fading.

The contrarian view is that the current weakness may be overextended if compute prices are near a local bottom and supply is still constrained into late summer. A modest rebound in B200 rental rates or another marquee capacity deal would force a sharp reassessment because positioning now appears crowded into memory and infrastructure winners. In that scenario, NVDA can outperform quickly on sentiment alone, even if fundamentals have not materially improved yet.

Key Decisions for Investors

• Reduce NVDA exposure tactically over the next 1-2 weeks; use any intraday strength to trim, since the near-term risk/reward is capped by soft spot compute pricing and rotational flows into memory.
• Go long MU / short NVDA as a 1-3 month pair trade. Rationale: memory remains the cleaner beneficiary of AI capex rotation, while NVDA faces the risk of margin multiple compression if rental pricing keeps drifting lower.
• Add SNDK on weakness for a 4-8 week momentum continuation trade, but keep it smaller than MU: the setup benefits from the same storage bottleneck theme, yet valuation and sentiment are more extended.
• For more convexity, buy NVDA 2-3 month call spreads rather than outright stock if you expect a rebound in compute pricing; this captures upside if rental rates stabilize while limiting carry cost if the downtrend persists.
• Watch B200 spot rental pricing as the trigger. If it reclaims the prior high within 2-4 weeks, cover tactical NVDA shorts immediately; if it fails to stabilize, expect another leg of de-rating across GPU-exposed names.