From Super to Microcomputer: The Quiet Reshaping of How We Compute

In the early 1970s, a college computer lab would have looked more like a furnace room than an office. Men wearing short-sleeved shirts and wide ties attended to the entire setup, which included refrigerator-sized cabinets humming against the walls and vents expelling warm air. Before breakfast, a single machine could fill a hall, require its own cooling system, and still not be able to handle what a modern phone can. It’s difficult to ignore how swiftly that world vanished.

During that time, the supercomputer was the main event. Soft-spoken engineer Seymour Cray, who liked to work alone in a basement in Wisconsin, created devices that stretched the capabilities of silicon. His Cray-1, which had a padded bench around the base and an odd horseshoe shape, came to represent raw computing power in popular culture. They were purchased by governments. They were essential to weather agencies. Eventually, they were used by film studios to render dinosaurs.

The other side of the story then emerged. A more subdued group of engineers, including Bill Pentz and his group, were working on a nearly diametrically opposed project around the same time. They wanted smaller buildings rather than larger ones. That instinct gave rise to the microcomputer, which most of us now simply refer to as a personal computer. There was only one processor. At least in comparison, it was inexpensive. It might be placed on a desk. Additionally, it altered day-to-day existence in ways that the supercomputer never attempted or succeeded in.

The two worlds hardly interacted for decades. Supercomputers were used for tasks requiring thousands of processors operating in parallel, such as data mining, robot design, and astronomy. Spreadsheets, schoolwork, and family photos were all handled by microcomputers. One ran whatever operating system the buyer wanted to install that week, while the other used Linux and custom architecture. Really, there was no overlap until the AI boom made it necessary.

That’s what makes the present moment peculiar. The distinction between the two has begun to become hazy, and the business that stands to gain the most is not the one that most people would anticipate. Before becoming what some on Wall Street now refer to as the unsung partner of the Nvidia era, Super Micro Computer, with its headquarters located in San Jose, was a quiet supplier of server hardware for many years. From $15 billion in fiscal 2024 to $22 billion in fiscal 2025, its revenue increased dramatically. It appears that investors think the run is still ongoing.

In essence, Super Micro creates the link between the two realms of computing. Its servers are neither the lonely desktop nor the room-sized monster, packed with Nvidia’s Hopper and Blackwell chips and increasingly cooled by liquid rather than air. They can be assembled into clusters that function like supercomputers without having a supercomputer-like appearance because they are modular and stackable. Although it sounds like marketing jargon, the company’s name, Server Building Block Solutions, accurately sums up the engineering.

Even though the circle’s shape is strange, there’s a sense that we’ve come full circle. Computing became personal with the advent of the microcomputer. It was powerful because of the supercomputer. Something in between is currently taking place in unmarked warehouses outside of Taipei and throughout Texas and the Netherlands. It’s still unclear if that middle ground will hold or if the next change will push everything back toward one extreme. You get the impression that the engineers themselves aren’t totally certain as you watch this play out.