Qualcomm (QCOM) stands as a pivotal player in mobile computing. Its chipsets and cellular modems are the backbone of billions of smartphones across the globe, while its licensing arm leverages vital wireless patents to produce highly predictable, high-margin cash flow. This reliable financial foundation gives the company the unique ability to fund massive new growth avenues, including automobiles, virtual reality devices, and AI-enabled PCs, without compromising its financial stability.

Prioritizing these exact characteristics, and identifying fundamentally sound companies with strong margins and sustainable cash generation, is the key philosophy behind our High Quality (HQ) Portfolio.

Nevertheless, the most significant growth potential in computing over recent years has mostly eluded Qualcomm: AI infrastructure. This situation might be beginning to shift.

Recently, much of the growth within the semiconductor sector has stemmed from data centers, propelled by the swift integration of artificial intelligence. Despite GPUs primarily dominating AI training, an increasing proportion of AI expenditure is transitioning toward inference, the process of executing trained models in operational settings. This transition is more substantial than it appears: inference is anticipated to account for two-thirds of AI computing by 2029, representing 80 to 90% of the overall lifetime costs of an AI system. Autonomous AI systems that perform tasks, engage with software, and make decisions independently could further amplify the demand for effective inference hardware.

This transformation aligns well with Qualcomm’s strengths.

Reasons Why Qualcomm May Have An Advantage

The disparity between AI computing demand and power supply is expanding. Constructing a large AI data center can take between 12 to 24 months, while obtaining high-capacity grid connections in major U.S. markets requires 36 to 84 months. Currently, the U.S. interconnection queue surpasses 2,600 GW. Out of the 12 GW of U.S. AI data center capacity projected for 2026, merely 5 GW is in construction, with much of the remainder facing setbacks. Power availability increasingly poses a limiting factor on the expansion of AI infrastructure. See: Is The Power Grid Now Nvidia’s Greatest Growth Limitation?

Energy efficiency has always been fundamental to Qualcomm’s operations. Smartphones operate under stringent battery and thermal limitations, compelling the company to optimize performance per watt — a capability that is now crucial in AI infrastructure, where electricity expenses and power constraints are essential design considerations.

The acquisition of Nuvia has provided Qualcomm with a highly regarded CPU design team and its proprietary Oryon architecture, now implemented in its latest AI PCs and being modified for larger infrastructure applications. This expertise in integrating processing units, memory access, and specialized AI functions onto a single chip helps minimize bottlenecks and enhances performance as workloads grow more intricate.

Current Activities Of Qualcomm

Instead of vying for large AI training processors, Qualcomm concentrates on inference, anticipated to become one of the biggest sectors within the AI infrastructure market. Its recently introduced AI200 and AI250 server systems are engineered for inference, emphasizing capacity over peak performance: the AI200 accommodates 768GB of memory per card, significantly higher than competing accelerators like AMD (AMD) MI350X at 288GB and Nvidia (NVDA) B200-class at approximately 180GB per GPU. This allows clients to operate large models on fewer, lower-power systems, resulting in a reduced overall cost. Qualcomm has also secured a noteworthy customer. Its contract with HUMAIN, Saudi Arabia’s governmental AI initiative, involves a 200 MW deployment valued at around $1 billion.

The Upcoming Stage

Qualcomm’s aspirations reach beyond accelerators. Reports indicate it is developing an independent server CPU featuring as many as 80 custom Oryon cores; successful entrance into the server processor market would grant it a significantly larger portion of data center expenditures, positioning it as a more holistic infrastructure provider. Its recent acquisition of Alphawave Semi for $2.4 billion enhances its ASIC design and high-speed connectivity capabilities, potentially facilitating customized silicon solutions for major cloud providers and enterprise clients.

The Investment Argument

At present, the data center segment is minimal — server sales account for under 2% of total revenue, still classified within the generic IoT category instead of being reported separately. Qualcomm also contends with fierce competition from Intel (INTC) and AMD, along with cloud companies creating their own chips. This reality is precisely why investors are closely monitoring the situation.

AI infrastructure represents one of the largest expenditure opportunities within semiconductors: leading tech firms are poised to collectively spend over $600 billion this year. Even a modest foothold in servers, accelerators, or customized cloud silicon could yield a substantial new revenue source and decrease Qualcomm’s dependency on the established smartphone sector.

The challenge lies in execution: competing against established server manufacturers, AI chip leaders, and hyperscalers developing their own silicon. However, the rising demand for power-efficient AI infrastructure aligns well with Qualcomm’s core competencies. The inquiry is not whether Qualcomm will dominate data centers overnight, but whether it can transform its efficiency advantage into a significant share of one of technology’s rapidly expanding markets — and if successful, the long-term effects could be considerable.

Source link