The AMD Era Has Begun.

The speaker celebrates AMD's recent success, highlighting a decade-long investment that has yielded substantial returns. He attributes this success to a long-term perspective and a deep understanding of AMD's fundamentals. The core of AMD's advantage lies in its chiplet platform, which enables the company to produce personalized chips at a marginal cost for a growing variety of specialized workloads. This technology allowed AMD to disrupt Intel initially and is now positioning it as a leader in AI compute.

AMD's tactical decision to equip its chips with more memory than competitors has given it a significant advantage in AI inference. This allows entire AI models to fit on a single chip, leading to faster inferences. Major players like Meta, Oracle, and OpenAI are partnering with AMD, validating this approach. The speaker emphasizes that inference will be massive and distributed, extending from large data centers to edge devices, an area where AMD's chiplet platform is particularly well-suited to capitalize on.

The speaker argues that the excitement around AMD extends beyond recent OpenAI news, fitting into a broader macro thesis: AI is automating intelligence and making companies more efficient. He predicts that over the next five years, new AI compute workloads requiring specific hardware configurations will emerge, and AMD will be well-positioned to meet these demands due to its flexible chiplet architecture. AI at the edge is identified as a massive new market where AMD is expected to be a primary player.

The scale of AI's future impact, especially concerning energy consumption, is described as 'inconceivable.' The speaker emphasizes thinking 10-15 years out, where current impressive numbers are just short-term observations. AMD's platform creates an 'asymmetry factory,' enabling it to simultaneously address multiple trillion-dollar AI opportunities, such as AI PCs, AI-powered gaming, data centers (CPUs and GPUs), and distributed inference. This end-to-end AI capability, as described by Lisa Su, is a significant differentiator.

The speaker argues that the focus should shift from 'AI' labels to the fundamental demand for exponentially more compute over the next decade. AMD's technology is agnostic to these labels, focusing on combining serial and parallelized compute (CPUs with FPGAs) at a marginal cost and with personalization. This first-principles approach, combined with the difficulty for competitors to pivot to chiplet technology (an instance of the innovator's dilemma), gives AMD a significant and hard-to-replicate advantage.

AMD's long-term investment in its chiplet platform makes it challenging for competitors to catch up, similar to Palantir's iterated digital twin platform. The ability to apply chiplets effectively across diverse verticals like AI, data centers (CPUs/GPUs), AI PCs, and edge computing simultaneously is extremely difficult to replicate. The speaker concludes that AMD has a platform that prints personalized chips at a marginal cost, tackling several multi-trillion dollar AI opportunities. He believes in the high odds of success for many of these ventures, making AMD a highly asymmetric investment for the next 5-10 years, independent of short-term AI narratives.