What a $429 million cube of failed hardware can teach Washington about the courage to shut down what doesn’t work.
Steve Jobs got pushed out of his own company in 1985. Not fired outright, but close enough. Apple’s board stripped him of any operational authority, and he walked out the door with nothing but his stock and his pride. Most executives in that spot take the money and disappear to a beach house. Jobs did something else. He wrote a cheque for seven million dollars of his own money and started over.
The company was called NeXT. The product was a jet black computer cube that cost as much as a used car. It sold roughly 50,000 units over its entire run. By any conventional business measure, NeXT was a failure. Jobs shut down the hardware line in 1993 and laid off more than three hundred employees. Judged on units shipped, the whole venture barely registers as a footnote in computing history.
Judged any other way, NeXT built the future.
The operating system inside that overpriced cube, a piece of software called NeXTSTEP, combined a Unix foundation with an elegant development environment that let programmers build serious software fast. A researcher at CERN named Tim Berners-Lee got his hands on one and used it to write the first web server and the first web browser. He had HTML, HTTP, and URLs defined by Christmas of 1990. John Carmack used NeXT machines while building the engines behind Doom and Quake. Apple, staring down a dying operating system of its own a decade later, bought NeXT for $429 million in December 1996 and brought Jobs back through the front door. NeXTSTEP became the backbone of Mac OS X. Mac OS X became the backbone of the iPhone. The cube that nobody bought ended up running quietly in the pocket of nearly every American.
The story people tell at dinner parties. Here is why it matters to anyone who cares about how government builds things.
I have spent my career working on federal technology projects. I can tell you the one lesson Washington refuses to learn from Silicon Valley: real architecture requires the freedom to fail in public. NeXT failed as a business in full view of the press, the analysts, and Apple’s own board. Jobs did not bury the failure inside a status report marked “on track, yellow.” He shut down the hardware line, took the loss, and kept building the part that actually worked. That kind of honesty is nearly extinct in government IT.
Federal technology programs are not rewarded for developing permanent architecture. They are rewarded for staying green on a dashboard. A contracting officer wants no findings. A program manager wants no headlines. So, agencies keep pouring money into legacy systems that everyone privately admits are dying. (looking at the State of Maryland DOIT) Admitting it in writing poses a risk to someone’s career. We build systems designed to survive an audit, not a decade. NeXTSTEP has survived four decades and counting. Most government platforms cannot survive four years without an emergency rewrite nobody budgeted for.
There is a governance lesson buried inside that cube. It has nothing to do with Steve Jobs being a genius. It has to do with tolerance for honest failure. NeXT was allowed to fail because it was Jobs’s own money on the line. His own name was attached to the outcome. Government programs rarely have either. They run on appropriated funds with no single owner accountable for the architecture ten years out, so nobody has the incentive to make the hard call Jobs made in 1993. Nobody shuts down the failing line. Agencies keep funding it, quarter after quarter, because closing a program is politically harder than sustaining a bad one.
If Washington wants technology that lasts as long as NeXTSTEP did. It needs program owners with something close to the authority Jobs had over his own company: the power to kill what does not work and protect what does, without waiting on a committee that will not exist in five years to approve the decision. Accountability lacking authority is theater. It produces the dashboards, not the results.
The computer failed. The code conquered the world. Good governance ought to distinguish between the two. It should stop propping up the computer just because shutting it down feels as if admitting defeat. Sometimes the only way to build something that lasts forty years is to be willing to bury something after three.