3 Things Bears Have Wrong About Nvidia Right Now

Nvidia's Q2 FY2027 revenue grew 106% year over year, exceeding the 97% growth analysts had expected, while the company forecast 70% revenue growth for FY2028 versus prior analyst expectations for 45%. Data-center revenue, representing 93% of Nvidia's sales, rose 117% and added $48 billion year over year, far exceeding AMD's $3.5 billion incremental data-center revenue despite AMD's 107% segment growth. Nvidia reported a 75% gross margin and 56% adjusted net margin, while trading below 15x next year's projected earnings; bearish positions total roughly $57 billion, with shares sold short up 40% over the past year.

Analysis

The relevant debate is not whether NVDA can beat consensus once more, but whether hyperscaler AI capex converts into durable inference demand before the next architecture cycle erodes pricing. A forward multiple in the teens is only inexpensive if consensus earnings are resilient; a 10-15 point gross-margin reset from mix, HBM cost pass-through, or accelerated product transitions would materially impair EPS even with continued revenue growth. The article’s valuation framing relies on company guidance and sell-side estimates rather than independently observable cloud-utilization or customer-ROI data.

Competitive risk is asymmetric across the stack. AMD can grow rapidly without taking enough workload share to damage NVDA, but any evidence that ROCm/software portability is reducing switching costs would compress NVDA’s platform premium well before reported market share changes. Conversely, memory suppliers such as MU are a cleaner second-order beneficiary of sustained accelerator volumes, though their earnings remain more exposed to HBM supply additions and conventional DRAM cyclicality than NVDA’s software ecosystem.

Reported short interest is not a reliable squeeze signal at this scale: much of the notional can be index, convertible, options-market-maker, or long-semiconductor-book hedging. Near term, positive sentiment can support upside through earnings revisions, but the more actionable 1-3 month catalysts are hyperscaler capex guidance, lead-time commentary, Blackwell/next-generation ramp yields, and evidence of inference monetization. Over 6-18 months, the key falsifier is a decline in NVDA data-center gross margin without offsetting unit growth, or a major cloud customer disclosing materially lower AI infrastructure returns.

Contrarian view: the bullish consensus may be underestimating the stock’s sensitivity to a modest deceleration because the valuation multiple already embeds unusually high confidence in estimate durability. The bear case is not an immediate demand collapse; it is that supply-chain normalization and customer bargaining power turn a hardware scarcity rent into a lower-margin, still-growing business. That setup favors relative trades over an outright short.

Key Decisions for Investors

• Maintain a core NVDA long only while quarterly data-center gross margin and next-quarter revenue guide remain at or above consensus; use any post-earnings strength to fund downside protection rather than add beta. A margin-guide miss of more than 200 bps or a hyperscaler capex cut should trigger a reduction, as multiple compression would likely dominate an otherwise solid revenue beat.
• Implement a 3-6 month relative-value trade: long NVDA / short AMD in dollar-neutral sizing after confirming AMD’s AI revenue guidance has not accelerated relative to expectations. The thesis is that NVDA’s software, networking, and installed-base economics preserve profit share even if AMD gains units; cover if AMD raises data-center AI guidance materially while NVDA’s gross-margin outlook weakens.
• Use MU as a monitored, smaller satellite long rather than a direct NVDA substitute: add only if HBM contract pricing and supply commitments remain tight through the next two quarters. Risk is that incremental HBM capacity or weaker handset/PC DRAM demand overwhelms AI-related mix benefits; exit on declining HBM pricing or a downward revision to bit-demand guidance.
• Do not position for a short squeeze based on headline short notional. Before purchasing NVDA calls, require options data showing elevated put skew or unusually low upside implied volatility; absent that setup, defined-risk call spreads are preferable to outright calls given event-driven volatility and the risk of a valuation reset despite operational outperformance.