Tim O’Reilly: AI Labs Are Repeating Microsoft’s 1990s Monopoly Mistakes

Tech visionary and O'Reilly Media founder Tim O’Reilly is sounding the alarm on the current trajectory of major AI labs, warning that they are attempting to lock users into closed ecosystems. O’Reilly compares the behavior of today's 'hyperscalers' to Microsoft in the 1990s, arguing that these firms are prioritizing shareholder gains and extractive value over genuine utility for the end user.

O’Reilly contends that the industry's current definition of 'open source' is insufficient. While some labs release 'open-weight' models, he argues that true open-source AI must encompass the entire stack, not just the neural-net weights. By controlling the surrounding infrastructure, labs maintain a grip on the ecosystem; conversely, a fully open stack would return control to the designers and users who actually implement the technology.

The Human Cost of Deployment

Beyond technical architecture, O’Reilly challenges the narrative that AI is an inevitable job-killer. He asserts that the technology itself is not taking jobs; rather, the decisions of the people deploying it are the cause. According to O’Reilly, value is being concentrated in the hands of executives who use AI to "gamble on share prices" rather than to enhance productivity. He specifically notes that replacing humans with AI in sectors like customer service will not improve service quality; instead, the two must be integrated to achieve actual improvement.

The Risk of Monopoly Rents

This pursuit of monopoly rents may be a strategic miscalculation for U.S.-based labs. O’Reilly warns that the attempt to maintain high costs through closed systems could be undermined by Chinese businesses, which he suggests may build comparable models at a significantly lower cost. This competitive pressure could collapse the financial walls that current frontier labs are attempting to build around their proprietary models.

Fighting 'Enshittification'

To combat the degradation of digital platforms—a process often termed 'enshittification'—O’Reilly now co-leads the AI Disclosures Project. This initiative explores the possibility of requiring companies to disclose platform metrics, providing a layer of transparency that could prevent platforms from pivoting from user-value to pure extraction once they have achieved market dominance.

The Path Forward

As AI becomes foundational infrastructure, the industry faces a choice between 'appliance' models controlled by a few labs and 'infrastructure' models based on open standards. O’Reilly views AI as 'normal technology' subject to social and physical constraints, rather than a direct path to Artificial General Intelligence (AGI). The critical question is whether the market will succumb to the current 'winner-takes-all' narrative or return to the historical success of open standards to drive the next wave of innovation.