Broadcom and Marvell highlighted as underwatched alternatives to Nvidia in the next wave of AI spending

Nvidia’s dominance in AI training hardware is still the benchmark for how the market thinks about the artificial intelligence boom. But a new market-focused piece from contends that the next major phase of AI buildouts may shift more value toward supporting chips and infrastructure suppliers, where Broadcom and Marvell have meaningful opportunities.

The core claim is not that Nvidia’s position is being overturned, but that investors may be overlooking a “quiet shift” in hyperscalers’ spending priorities as they move beyond the most visible training deployments. The article frames the next trillion dollars of investment as a catalyst for broadening the competitive field, with Broadcom and Marvell portrayed as beneficiaries of the procurement mix that comes with scaling data center operations.

The market note specifically groups Broadcom (AVGO) and Marvell as chip designers with a chance to capture demand that is tied to data center infrastructure rather than only to the largest AI accelerators. In that framing, Broadcom and Marvell would be pulled into the same spending cycles as Nvidia, but through different product categories, potentially making their growth more sensitive to cloud capex decisions than investors expect.

Because the post does not provide detailed disclosure of which hyperscalers, contracts, product lines, or financial metrics it is referring to, it is difficult to translate the argument into a precise near-term forecast. The article also does not cite company-specific guidance or named wins. That leaves the thesis largely directional: a call to pay closer attention to infrastructure adjacent players during the next phase of AI infrastructure buildouts.

Broadcom’s relevance to this type of spending typically stems from its large footprint in networking and connectivity technologies used in data centers, while Marvell is positioned in ways that can also tie into data center connectivity and custom silicon designs. The market note’s thrust aligns with that general industry logic, where training is only one part of the AI stack and large-scale deployment depends on high-throughput networking, switching, and data center system integration.

Still, key details remain unspecified in the published post. It does not outline which Broadcom or Marvell offerings are expected to gain, whether the shift is driven by specific product transitions, or how much of the “quiet shift” is already reflected in current estimates. Without those particulars, investors are left with a broader competitive narrative rather than a verifiable, quarter-by-quarter setup.