Solow's lesson and the dotcom comparison
In an editorial published by the Guardian, the economist John Quiggin, a professor at the School of Economics at the University of Queensland, takes up an observation by Nobel laureate Robert Solow from 1987: the computer age was visible everywhere except in the productivity statistics. At the time, computers had existed for forty years. Productivity gains arrived in the 1990s with the internet, but were short-lived. According to Quiggin, four years after the launch of ChatGPT the pattern is repeating: much emphasis and very limited productivity growth.
As in the years of the dotcom boom, the main economic effects are concentrated in the stock market and in infrastructure investment: fiber optics then, data centers today. The bet is on future demand that, the economist writes, has not yet materialized.
The announced layoffs and the reasons given
Quiggin examines the cuts announced by NAB, Qantas, Canva and Atlassian, some of which were explicitly attributed to AI, and asks whether the feared collapse in employment has arrived. His answer is no. NAB denied that AI played a role in the decision and, according to the author, there is no reason not to believe it: since 2010 the bank has announced mass layoffs almost every year, while its total number of employees has remained unchanged, also as a result of resignations and retirements. The model, he argues, is the typical one of large companies: hire when needed and lay off when convenient.
In Qantas's case, according to Quiggin, AI is invoked to muddy the waters: the outsourcing of 850 jobs to India is part of the company's long-term strategy. If those jobs are going to be automated anyway, he observes, there is no explanation for why it would lay workers off first and open an operation abroad.
Canva, Atlassian and the risk for the big model providers
Canva appears more exposed: its subscription software model for simple graphics is being challenged by the possibility of producing a poster or an advertisement in seconds with AI, perhaps of questionable quality but sufficient for many users. Atlassian, by contrast, operates in project management and organization-level data, functions that fragmented use of AI by individual employees is unlikely to replace; Quiggin does not rule out that AI could enable reverse engineering of software, but the question remains open.
Finally, the author takes aim at the big model providers themselves, such as OpenAI and Anthropic, as well as Meta and Alphabet. Low-cost open-weight models such as DeepSeek and Mistral, he argues, are about a year behind the leaders while investing a fraction of the resources. If language models become like email—indispensable but not profitable—the risk of a collapse will concern the stock market more than employment.
