Surviving the New Economics of a Post-Agentic World

IBM's stock collapsed 25 percent in a single trading session, erasing approximately 70 billion dollars in market value, the worst single-day drop for the company in over 50 years and worse than its losses during the 1987 Black Monday crash. The New York Times described IBM's situation as a potential canary in the coal mine for wider economic disruption. On the same day, Workday fell approximately 10 percent, Salesforce dropped around 9 percent, ServiceNow fell 8 percent, and Adobe declined 6 percent, while Nvidia and Intel rose as investors rotated toward hardware and infrastructure.

The sector-wide pressure was attributed to companies diverting IT and technology budgets away from enterprise software and services toward hardware and infrastructure. Part of this shift reflects panic buying of hardware driven by fear of being priced out of AI model access, as agentic AI requires far more context and long-running processes, increasing infrastructure cost pressure significantly. This represents a structural reallocation rather than a temporary dip, with executive-level capital being pulled away from historically core operations toward AI risk mitigation.

Companies are simultaneously facing pressure from expensive US proprietary AI models and the potential restriction of Chinese open-weight models. China is moving to limit its open-source AI exports in response to US export controls, closing off a fallback strategy many organizations had been relying on. This dual squeeze leaves enterprises with fewer affordable options and forces harder choices about which AI infrastructure bets to make under uncertainty.

A Chinese token black market has emerged as a gray market where brokers resell discounted API access to Western AI platforms including OpenAI and Anthropic to mainland Chinese users who cannot directly pay for or access those services. The speakers cited this as a predictable outcome when regulation is used to artificially shape access for political ends, suggesting that enforcement gaps will persist regardless of the formal restrictions in place.

The shift to massive agentic infrastructure is already operational, not theoretical. One company recently contacted already has 70,000 agents running in production and another has approximately 6,000. The economics change substantially when moving from a personal interactive agent to a long-lived cloud-hosted agent running for hours, days, or weeks, and agentic loops incorporating hundreds of thousands or millions of interacting agents represent the emerging operational paradigm. Future enterprise software interactions are expected to occur agent-to-agent without direct human involvement in most processes, displacing traditional GUI and web interfaces.

Companies like NetSuite are pursuing an MCP connector strategy on the assumption that their value lies in their data platform surfaced agentically, while some vertical software vendors are building proprietary agent sets rather than connecting to the broader ecosystem. Managing agentics at massive scale was identified as a wide-open entrepreneurial opportunity, and the concept of loop engineering was predicted to be antiquated before the end of the current year. Agentic technology is simultaneously replacing existing technologies and existing human roles, with mass layoffs in tech reflecting companies experimenting with hyperproductive alternatives to human labor. This does not remove all humans from the equation but fundamentally changes their roles, and the rate of change in business fundamentals is accelerating exponentially.

Anthropic released a paper titled Verbalizable Representations Form a Global Workspace in Language Models, in which the company observed that models create internal structures called workspaces that fulfill the same function as working memory in a human brain. The routing and reporting capability these workspace structures enable was described as a key mechanism underlying the agentic transformation currently underway. Whether the mechanisms involved constitute a form of consciousness remains unresolved, as there is no widely accepted unified definition of consciousness and a philosophical debate persists between those requiring a strict neuroscience-based definition and those who accept functional equivalence as sufficient.