Looking for a Good Deal on an AI Stock? This Leading Chipmaker Is Cheaper Than Broadcom and AMD.

Nvidia is presented as a comparatively inexpensive AI leader, trading at 27x trailing earnings versus 131x for AMD, 46x for Broadcom, and roughly 33x for the technology sector. Fiscal Q2 2027 revenue more than doubled year over year to $96.2 billion and net income rose 126% to $59.7 billion, while Q3 revenue guidance of $108 billion implies 90% growth. The bullish case rests on Nvidia's estimated 86% GPU market share and sustained AI infrastructure spending, including projected $725 billion hyperscaler capex this year and more than $1.3 trillion in AI-company capex next year.

Analysis

The relevant relative-value question is not whether NVDA is optically cheaper on trailing earnings, but whether its inference economics can remain differentiated as hyperscalers internalize workload-specific silicon. NVDA retains the strongest near-term monetization of aggregate AI capex because its platform captures accelerator, networking and software attach; AVGO is the clearest structural offset because custom ASIC adoption reallocates spend rather than merely expands the market. Over 6-18 months, a rising ASIC mix would pressure NVDA's incremental gross-margin expectations and valuation multiple even if absolute revenue continues growing.

AMD's upside is more execution-sensitive than the headline AI narrative implies. Its CPU franchise benefits if AI deployment shifts from training clusters toward inference and agentic workloads with higher host-processing, memory and networking requirements, but that benefit depends on share gains and supply availability rather than market growth alone. The second-order beneficiary is MRVL, whose optical/interconnect exposure can rise with cluster scale regardless of whether GPU or custom-silicon architectures win; conversely, a capex digestion phase would hit AVGO and NVDA through order timing before it is visible in end-demand metrics.

Near term, bullish sell-side framing is unlikely to be a fresh catalyst after a strong AI complex run. The tradable catalyst over the next 1-3 months is hyperscaler capex commentary and evidence that AI infrastructure is converting into revenue, not just booked capacity. A material reduction in GOOG or META infrastructure-spend plans, or a sequential deceleration in NVDA data-center gross margin, would challenge the thesis; sustained custom-chip deployment without a corresponding reduction in external accelerator purchases would falsify the bearish ASIC-substitution case.

Contrarian view: the market may be underpricing architecture diversification, but overpricing its immediacy. Custom chips are economically compelling at stable, enormous internal workloads; frontier-model iteration and heterogeneous inference requirements still favor programmability. This supports owning NVDA tactically while hedging the longer-duration competitive risk rather than treating AVGO's ASIC pipeline as an immediate zero-sum displacement event.

Key Decisions for Investors

• Initiate a 1-3 month long NVDA / short AMD pair on equal dollar beta: NVDA has lower execution dependence and greater platform capture, while AMD needs sustained product and share validation. Review after the next earnings cycle; exit if NVDA data-center gross margin declines materially sequentially or AMD raises accelerator/CPU share expectations with credible supply evidence.
• Maintain a 6-18 month core long AVGO as a hedge against merchant-GPU concentration, preferably funded by trimming broad semiconductor-beta exposure (SOXX) rather than shorting NVDA outright. The payoff is strongest if hyperscalers disclose expanding internal silicon programs; principal risk is customer concentration, project delays, and ASIC volumes remaining design-win rather than production revenue.
• Add MRVL to the AI-infrastructure watchlist rather than chase immediately: initiate only on independently verified optical/interconnect order acceleration or a post-results dislocation. It offers architecture-agnostic cluster-growth exposure, but the missing data are customer concentration, pricing and inventory normalization.
• Set a capex-revision alert around GOOG and META results over the next two quarters. A combined downward revision to AI infrastructure plans would justify reducing NVDA/AVGO exposure first, as supplier revenue recognition and networking demand can lag customer capex cuts by one to two quarters.