The world of artificial intelligence (AI) is undergoing a significant transformation, and the focus is shifting from training large language models to putting them to work in the real world - a phase known as inference. This shift has profound implications for the tech industry, particularly for chip designers and memory manufacturers.
The AI Inference Supercycle: A New Era
In the AI landscape, the training phase is akin to preparing for an exam, requiring immense computing power and vast datasets. However, the inference phase is where the real action happens - it's the exam itself, where the AI models are put to use, fed with new data, and expected to deliver accurate results.
This transition to inference-focused workloads is a game-changer. It's not just about the big players like Broadcom and Intel, who have been riding the wave of healthy demand for their inference-optimized processors. The real winners, in my opinion, are the memory manufacturers, specifically Micron Technology and Sandisk.
Why Micron and Sandisk Are the Real AI Inference Winners
The inference phase demands more than just powerful processors. It requires efficient data movement between the processor and storage, and that's where memory comes into play. As AI applications become more prevalent, the number of inference requests will skyrocket, and memory will be crucial to ensure quick responses from trained models, especially on edge devices like smartphones and PCs.
Consulting giant Deloitte predicts that inference workloads will account for a significant portion of AI data center computing power this year, and this shift will drive an increased demand for data storage. Western Digital CEO Irving Tan highlighted this inflection point, noting that the larger focus on inference will lead to a greater need for data storage.
Micron and Sandisk are well-positioned to capitalize on this trend. Micron manufactures both dynamic random-access memory (DRAM) and NAND flash, while Sandisk is a pure-play flash storage manufacturer. The severe supply shortage of memory chips, driven by AI-fueled demand, has led to price hikes and a favorable environment for these companies. Gartner predicts a substantial increase in the memory industry's revenue this year, far surpassing the growth anticipated in the non-memory semiconductor market.
The Long-Term Outlook for Micron and Sandisk
The memory shortages are expected to persist until the end of the decade, driven by AI demand from data centers, PCs, and smartphones. This means the incredible growth experienced by Micron and Sandisk is likely to continue. Their stocks have already seen impressive gains, with Micron up by 639% and Sandisk's gains even more astonishing at almost 3,400%.
Despite these gains, both semiconductor stocks are trading at attractive valuations. Micron has a forward earnings multiple of 7.6, while Sandisk's stands at 24. Given their stellar earnings growth rates, which significantly outpace the broader market, these stocks deserve a premium valuation.
In conclusion, the AI inference supercycle is a lucrative opportunity, and investing in Micron and Sandisk, with their eye-popping growth rates and attractive valuations, seems like a smart move. As we navigate this new era of AI, the role of memory manufacturers will only become more critical, and these companies are well-placed to capitalize on this trend.
