The Internet Was a Rising Tide. AI Is a Black Hole.
The Internet Was a Rising Tide. AI Is a Black Hole.
The internet changed everything, and it meant that in a good way. When the web went mainstream in the mid-1990s, it did not simply transform industries that already existed. It built new ones from nothing. Web designers, e-commerce operations, app developers, cloud hosting providers, digital marketers — none of these existed before the network arrived. Capital flowed in and multiplied. Jobs appeared that had no name ten years earlier. Entire ecosystems of small businesses, agencies, and independent workers found room to grow. A rising tide that lifted all boats.
Artificial intelligence was supposed to follow the same arc. The early narrative was familiar and reassuring. Yes, some jobs would be disrupted, but new categories of work would emerge to replace them. Yes, enormous capital would be required to build the infrastructure, but the returns would justify it. History was the evidence. Every general-purpose technology in the modern era had ultimately been generative. AI, the argument went, would be no different. The transition would be uncomfortable but ultimately additive.
One in Three
Microsoft offered a data point today that is difficult to argue away. The company announced a voluntary exit program targeting approximately 60,000 employees — roughly 27 percent of its global workforce — on the same week it is expected to report strong double-digit earnings growth driven by AI infrastructure investment. The roles excluded from the cuts tell the whole story. AI development, security, and data centers are explicitly protected. The people who built and supported the products that made Microsoft one of the most valuable companies on earth are being offered a check to leave quietly before December.
The internet created the middle layer of the knowledge economy. AI appears to be consuming it.
Nothing Escapes
That is the black hole problem. A black hole consumes. It does not radiate opportunity outward. The capital flowing into AI infrastructure right now is staggering, but it is concentrating at the top — in the data centers, the chip manufacturers, the hyperscaler platforms, and the small number of engineers who build the models themselves. The broad middle layer of knowledge workers — the support engineers, account managers, product specialists, and solution architects who made enterprise software actually function for real customers — is being compressed or eliminated. The internet created that middle. AI appears to be consuming it without replacing it.
The structural difference between the two technology waves matters more than the surface similarities. The internet was additive. It created new surface area in the economy — new businesses, new roles, new niches that generated income for people who showed up and learned the tools. AI, in its current deployment, is operating differently. It is being used primarily to reduce the number of humans required to run existing businesses. It is not generating new categories of work at anything close to the rate it is eliminating existing ones. Microsoft will report strong earnings this week. The stock will likely rise. Wall Street will celebrate the efficiency gains. Somewhere around 60,000 people will be deciding whether to take a severance package or gamble on the next round of cuts.
Even the Builders
The companies building AI are not immune to this dynamic. Frontier labs — the organizations racing to develop the most capable models — are accumulating debt at a scale with no historical parallel in the technology industry. Compute costs do not shrink as the models improve. They grow. Revenue at these organizations is rising, but not fast enough to close the gap. The large platforms can absorb the losses by redirecting income from older, profitable businesses into the bet. The pure-play labs have no such cushion. They are entirely dependent on investor conviction that a profitable version of this technology is coming. That conviction is an article of faith. The debt is not. Even the builders are getting consumed.
Whether AI eventually generates the kind of broad economic opportunity the internet did is a question nobody can answer honestly today. The jobs being eliminated are real and immediate. The debt being accumulated is historic and growing. The returns being promised are distant and unproven. That is not a rising tide. We will see how this story turns out.
Aaron Rose is the founder and publisher of Tech Reader Magazine, covering system architecture, cloud platforms, and AI policy.