Artificial intelligence has been defined by a relentless race for more computing power. Nvidia‘s (NASDAQ:NVDA) GPUs became the stars of that story because they delivered the horsepower needed to train ever-larger AI models. But every technology boom eventually runs into a new constraint.
According to a recent Morgan Stanley report, that next hurdle isn’t computing power — it’s memory. As AI models grow larger and inference workloads become more demanding, data centers need far more memory bandwidth and capacity to keep expensive accelerators fed with data. That shift could reshape where hundreds of billions of dollars in AI infrastructure spending flows over the rest of the decade.
This post originally appeared at 24/7 Wall St.