Nvidia CEO Signals Strong Momentum for Shareholders

Nvidia reported fiscal Q2 2027 adjusted EPS of $2.22 and revenue of $96.22B, exceeding Wall Street expectations of $2.10 and $92.17B, respectively; revenue rose 106% year over year. CEO Jensen Huang guided for roughly 70% sales growth in fiscal 2028 versus the 44% analyst consensus and forecast 72%-73% gross margins despite supply constraints. Nvidia also expects its Vera Rubin GPU platform to be its fastest-ever ramp, with estimated AI-compute revenue potential of $40B per gigawatt versus $25B for Blackwell, while CPU revenue is projected to more than double next fiscal year.

Analysis

The important read-through is not another beat, but Nvidia's ability to defend a low-70s gross-margin floor while migrating customers to a higher-content platform. If sustained, this shifts the debate from cyclical accelerator pricing to share-of-rack expansion: CPUs, NVLink/networking and systems can raise Nvidia's wallet share even if GPU unit growth normalizes. The clearest competitive casualty is AMD, whose accelerator opportunity requires meaningful pricing or availability gaps at Nvidia; Intel's server-CPU recovery case is also impaired if hyperscalers increasingly standardize around Nvidia CPU/GPU/networking architectures.

The second-order winners are TSMC, Broadcom and Arista Networks, although each captures a different portion of the buildout. TSMC benefits from leading-edge wafer and advanced-packaging intensity, while AVGO and ANET benefit only if AI clusters continue to scale out rather than consolidate into more integrated rack designs. Conversely, a larger Nvidia networking attach rate creates a medium-term share risk for merchant silicon and optical vendors, including MRVL; investors should not assume all AI infrastructure suppliers participate equally in rising AI capex.

Near term, the guidance upgrade is likely well understood, so upside depends on evidence that supply constraints are deferring—not destroying—demand and that Rubin availability does not cause a Blackwell digestion period. Over the next 1-3 months, hyperscaler capex commentary and TSMC advanced-packaging capacity updates are the key confirmation points. The 6-18 month risk is customer ROI: a pause in Microsoft, Meta, Alphabet or Amazon AI infrastructure spending would expose the operating leverage embedded in expectations, while the cited revenue-per-gigawatt economics remain management targets rather than independently validated demand.

Key Decisions for Investors

• Maintain or add NVDA on post-earnings consolidation rather than chase a headline-driven gap; target a 6-12 month holding period. Thesis is platform-content expansion, not another isolated GPU beat; reduce if gross-margin guidance falls below 70% or a major hyperscaler cuts AI capex.
• Pair long NVDA / short AMD over 3-6 months, sized beta-neutral. Nvidia's broader CPU-networking integration raises the hurdle for AMD to monetize accelerator share; invalidate the trade on evidence of sustained MI-series supply wins at two or more tier-one cloud customers or meaningful AMD AI revenue-guide upgrades.
• Long TSM versus a semiconductor-equipment basket is a cleaner 6-12 month capacity expression if advanced-packaging utilization remains tight; use TSMC monthly revenue and CoWoS capacity commentary as confirmation. Avoid adding if leading-edge utilization weakens despite AI demand, which would indicate broader handset/HPC softness is offsetting the AI mix benefit.
• Watch ANET and MRVL earnings for AI-networking attach-rate evidence before initiating directional exposure. Buy ANET only on confirmed AI-backlog acceleration; consider MRVL underweight if Nvidia's integrated networking revenue outgrows the AI networking market, signaling share capture rather than category expansion.