In the current investment landscape of Artificial Intelligence (AI), Nvidia's name has become synonymous with explosive growth. However, as the market matures and shifts from the training of giant models to their daily execution—the process known as "inference"—another player is preparing to take the lead. Arm Holdings, the British semiconductor design firm owned by SoftBank, doesn't just manufacture chips; it defines the architecture upon which the future of computing is built.

The Great Shift: From Training to Inference

For the past two years, demand for Nvidia's GPUs has been insatiable, as companies like OpenAI and Google needed to train Large Language Models (LLMs). But training is only the beginning. Once a model is trained, it must be "run" billions of times a day to answer user queries, recognize images, or translate text. This is inference.

The critical difference lies in energy consumption. While training requires raw power, inference demands efficiency. This is where Arm's architecture excels. Arm-based chips use a RISC (Reduced Instruction Set Computer) instruction set, which allows complex tasks to be executed with a fraction of the energy required by traditional x86 processors from Intel or AMD. In a world where data centers are straining city power grids, energy efficiency is no longer a luxury but a prerequisite for survival.

The Royalty Model and Ubiquity

Unlike Nvidia or Intel, Arm does not sell physical products to the general public. It sells intellectual property. Every time a company like Apple, Amazon, or even Nvidia itself designs a processor based on Arm's architecture, Arm collects royalties. This creates a business model with exceptionally high margins and, more importantly, a "moat" that is almost impossible to bridge.

According to recent analyses, Arm already holds 99% of the smartphone market, but the real growth is coming from data centers. Amazon's Graviton, Google's Axion, and Microsoft's Cobalt processors are all based on Arm. As these companies seek to reduce their dependence on expensive Nvidia GPUs for inference, they are turning to their own custom, Arm-based processors. This positions the company at the heart of every AI infrastructure, regardless of who wins the "chip wars."

Competition and the Future of PCs

Arm's dominance is now extending to personal computers. The recent entry of Qualcomm's Snapdragon X Elite processors into the Windows PC market marks the end of Intel's monopoly. These chips, naturally based on Arm architecture, offer battery life and AI performance that traditional processors struggle to match.

However, challenges exist. The rise of RISC-V, an open-source architecture, poses a long-term threat as it offers a royalty-free alternative to Arm's licenses. Furthermore, geopolitical instability and export restrictions to China directly affect the company's revenue. Nevertheless, Arm's position as the "architect" of the digital world remains steadfast. If Nvidia was the shovel in the AI gold rush, Arm is the blueprint for the cities that will be built after the rush subsides.

Investor Takeaways

The market often overlooks Arm due to its lower visibility compared to end-product manufacturers. But the economic reality is clear: inference is the future of AI at scale, and inference "speaks" the language of Arm. With the introduction of the v9 architecture, which doubles the royalty rates the company receives per chip, Arm's financials are expected to skyrocket in the coming years, making it perhaps the biggest winner of the revolution we are currently undergoing.