Beyond Nvidia: Specialized AI Infrastructure Plays Gain Momentum

The artificial intelligence rally is expanding beyond primary chipmakers into the broader infrastructure ecosystem. While large-cap leaders continue to attract the most attention, a subset of specialized firms is demonstrating strong performance and strategic positioning by providing the essential components and power capacity required for AI workloads. Silicon Motion Technology (SIMO) is seeing a surge in demand for its NAND flash controllers, which are critical for the memory chips used in AI hardware. The company reported a 46% year-over-year revenue increase in the fourth quarter of 2025. Management has provided guidance for a strong start to 2026, expecting 80% year-over-year revenue growth in the first quarter. In the infrastructure space, IREN is scaling its GPU-as-a-Service offerings. The company has secured a 4.5 gigawatt grid-connected pipeline and a $9.7 billion, five-year agreement with Microsoft. This specific deal provides Microsoft access to 200 megawatts in a Texas-based data center, generating $1.94 billion in annual recurring revenue (ARR). However, the capital-intensive nature of the neocloud business presents risks. IREN has relied on debt and a $6 billion at-the-market equity program to fund construction and Nvidia chip acquisitions, leading to potential shareholder dilution. Meanwhile, Broadcom (AVGO) continues to leverage its position in custom AI chip design. With a market capitalization approaching $2 trillion, the company differentiates itself from general-purpose chipmakers by focusing on customized silicon tailored for the specific needs of hyperscale tech giants.