Should You Forget Palantir and Buy These 2 Artificial Intelligence (AI) Stocks Instead?

Broadcom reported AI semiconductor revenue more than tripling year over year in its latest quarter and expects growth to accelerate further, supported by hyperscaler demand for custom AI chips and networking equipment. Applied Digital has secured $36 billion of contracted data-center capacity through three investment-grade hyperscaler agreements, providing substantial revenue visibility as it expands AI campuses. The article positions Broadcom as the lower-risk AI infrastructure investment and Applied Digital as a higher-risk, higher-reward data-center buildout play.

Analysis

AVGO’s differentiated exposure is not simply accelerator share; it monetizes the architectural shift toward customer-specific silicon through both compute-adjacent connectivity and software attach. That creates a higher dollar-content opportunity per AI cluster than a single-chip narrative implies, but also makes quarterly results unusually sensitive to a small set of hyperscaler tape-outs and deployment schedules. Over the next 1-3 months, the key catalyst is evidence that custom-silicon volume ramps are incremental to—not substitutive for—NVDA spend; a weaker networking growth rate would be the earliest sign that cluster build-outs are pausing.

APLD’s equity should be valued as a leveraged project-finance vehicle rather than a conventional AI beneficiary. The contracted-capacity headline does not establish cash-flow quality until investors can verify counterparty concentration, lease commencement dates, tenant-funded versus APLD-funded capex, power delivery, and financing terms. The structural upside over 6-18 months is substantial if completed campuses convert backlog into EBITDA, but dilution, construction overruns, higher-for-longer funding costs, and a single customer deferring energization can overwhelm the apparent revenue visibility.

The non-obvious relative winner from sustained AI campus construction is likely power and electrical-equipment suppliers—ETN, VRT, PWR and GEV—whose orders are tied to grid interconnection, cooling and power-distribution bottlenecks rather than to a specific accelerator architecture. Consensus remains focused on compute; physical-power constraints can extend capex cycles while shifting economics toward suppliers with installed-base service revenue. Conversely, if hyperscalers optimize inference more rapidly than expected, merchant data-center developers face the greatest utilization and pricing risk before semiconductor vendors do.

Key Decisions for Investors

• Maintain/establish long AVGO versus short SOXX over a 3-6 month horizon: AVGO offers custom-silicon, networking and software diversification while SOXX retains broader memory/PC cyclicality. Reassess if AI networking growth decelerates materially for two consecutive reporting periods or a major hyperscaler signals internal-chip deployment delays.
• Do not initiate a core APLD position solely on contracted-capacity claims. Put it on a catalyst watch for disclosed lease start dates, committed project debt, power-delivery milestones and capex per MW; only consider a small long after these are independently reconciled to expected dilution and interest expense.
• Express the infrastructure bottleneck thesis through a 6-12 month basket long ETN/VRT/PWR, funded by an underweight in speculative data-center developers. Target asymmetric upside from order backlog conversion; reduce if utility interconnection queues shorten or hyperscaler capex guidance turns down.
• For NVDA holders, avoid treating AVGO custom silicon as a clean short-NVDA signal. Use any widening AVGO/NVDA relative-strength move as an alert to examine hyperscaler capex allocation, since aggregate cluster spending can support both until utilization, not chip supply, becomes the binding constraint.