Gergely Orosz Compares Current AI Infrastructure and Adoption to Cloud Adoption in the 2010s
Renowned engineer blogger Gergely Orosz points out that many characteristics of today's AI infrastructure development and enterprise adoption are highly similar to the cloud adoption process in the 2010s.
Cloud was initially thought to significantly reduce costs, but later it may actually increase expenditures; adoption and integration cycles can last for years; the biggest winners are often companies close to the infrastructure layer; cloud becomes a major budget item that companies need to plan for long-term; customers do not care whether the backend uses cloud or AI.
He also emphasizes that both cloud and AI are significant innovations that can give rise to entirely new business models, and ignoring them will leave tech companies significantly behind.
Source: Public Information
ABAB AI Insight
Gergely Orosz has long observed the evolution of engineering culture and infrastructure in tech companies, previously comparing the cloud era with the AI era. This article continues his analysis of the "technology hype to reality implementation curve," having detailed cases of uncontrolled cloud spending by multiple companies in 2025.
In terms of capital pathways, companies are forced to shift budgets from traditional IT to long-term planning for AI infrastructure (training, inference, data pipelines), accelerating the concentration of resources towards companies close to the model and computing layers. The motivation is to avoid the "save first, pay later" trap in the adoption curve while fostering new avenues for AI spending optimization similar to DuckBill Group.
Just as AWS and Azure became the biggest winners during the corporate cloud migration from 2010 to 2018, and the rise of cloud cost optimization tools followed, the AI industry is currently transitioning from early hype to long-term infrastructure spending control.
Essentially, this is a restructuring of the industry chain: AI infrastructure is shifting pricing power from end application companies to underlying computing, models, and optimization platforms. The mechanism involves long-cycle integration and continuous high expenditure characteristics, leading capital to shift from short-term tool procurement to long-term infrastructure subscriptions, forming an oligopoly structure similar to the cloud era.
ABAB News · Law of Cognition
All major technologies are initially promoted as cost-saving, but ultimately become the largest fixed expenses for companies. The closer you are to the infrastructure, the easier it is to reap the maximum benefits from the entire adoption curve. Customers never care what technology you use; they only care about the results you ultimately deliver.