Most of my colleagues today have never lived through a tech downturn. I don't count 2020, for reasons I'll come to soon. That means the last downturn was 2008, which was 17 years ago, and I have heard leadership at my company citing that. But I think it's a bad comparison; we should instead look at the 2001 dot-com bust before that.
In 2008, speculation in housing (and especially sub-prime lending) led to a global recession that swept the tech industry along with everyone else. As an industry, however, tech was relatively well-balanced. When the S&L crisis exploded, the government shored up the finance sector, consumer and employer confidence improved, and in tech our recovery was fairly quick for both employment (by 2010, according to FRED) and market valuations (October 2008 to Feb. 2011).
In 2001, by contrast, the tech industry was the problem: too many companies hired too many people, and a sock puppet without an sustainable business plan commanded $400M valuation. That was a bubble in tech, and the resulting burst hurt: stock valuations took over a decade to recover (November 2007, just in time for 2008's crash), as did job recovery (2017, as a percentage of total employment).
In 2020, the pandemic and our responses caused a general downturn, like 2008, not a tech sector one. But it turns out that software is pretty easy for work-from-home, so the truly disastrous impacts weren’t so bad on our sector. Hiring was already booming going into the pandemic, and the free money policies that kept the economy overall from tanking only made tech hiring boom more. So I don’t count 2020 as a downturn at all; it was a boom on top of a boom, in jobs and stock valuations both. Until 2022 for stock… and the layoff headlines started in 2023. What changed? Mostly the end of free money, making the costs of hiring harder to justify.
Which means what? First, it means that even senior leaders might be working from the wrong parallel, and expecting a fast rebound. More important, I think, is the push for LLMs aka Generative AI during a downturn. Even if you believe the AI claims, then in a boom time, employers would still be hiring, and a productivity-multiplying tool would magnify that investment in people. But in a lean time, the bias is on cost control, so a productivity multiplier is gets the required work done with fewer people: no hiring. Alternatively, if you don't believe the AI claims... then we're riding a bubble that's going to pop while we're already down. Either way the AI boom goes, I fear we might not see hiring recover for a long while!
What does that mean? The obvious short-term problem is on the employee side; we have thousands of experts who were recently laid off. There are jobs---my mailbox is still cluttered with unsolicited recruiter spam---but mostly in shaky-sounding startups and second-tier companies; the big leaders are sitting this round out. That leads to more competition for jobs that are riskier, have lower compensation, and/or stretch engineers less. The subtle problem is on the employer side: in the constrained environment, "growth" means hiring senior people instead of, or to replace, more junior ones. With both together, today’s new computer science grads face the highest unemployment rate of any degree; software as a field looks far less appealing.
The really big problem, however, is the future effect of this: senior job definitions entail mentoring of junior engineers, leading designs, defining and revising architectures, and we aren't necessarily creating more of that level of work. Instead, we're hiring senior people into, effectively, more junior roles, but at salaries for their seniority. Which may keep their bills paid, but might not keep them or stretch them; instead, it leads either to finding more stimulating work, into under-contributing complacency. So whenever we catch up to AI's impact, good or bad, and come out of the cold for this downturn, it may be to a deep labor shortage, if we aren't drawing new talent into the market and are driving existing talent either out of the market or into comfortable mediocrity.
We would work our way out of that, of course, but it would take time. Maybe lots of time, if AI gets better and better, or if a post-AI crash extends the current downturn!
