The Jobspocalypse Is Cancelled
Despite the valuations of their companies largely resting on the promise of displacing expensive employees with AI, both Sam Altman and Dario Amodei have handbrake-turned away from kissing goodbye to work. Instead, they now suggest we’ll probably be manacled to our jobs for a while yet.
The cynics will say it’s a marketing push to help balance increasing public unrest. AI is slurping up natural resources and trillions in capital, all while promising to drive up unemployment. A tough sell. Others will argue tech leaders are genuinely taking stock, recognizing that human labor is far more resilient and their AI tools far less capable than originally forecasted. The truth is likely somewhere between the two.
AI is now painfully unpopular. Luminaries from Big Tech have been booed out of graduation ceremonies for vaunting its capabilities in commencement speeches. Local governments in the pocket of data center developers go viral on social media as they’re berated by their constituents. And backlash against future employment loss has seen workers rally around ideas that range from mocking AI’s weaknesses to open sabotage.
At the same time, the genuine technical promise of AI has become more measured. Several years ago, the idea that AI could compose poems at the drop of a hat and conjure up complex research made an immediate impact on the job market seem inevitable. Over the years, as videos of Will Smith eating spaghetti became more believable, the threat seemed more plausible still.
But as far as economic data is concerned, there’s nothing to write home about. While AI-related companies have tacked billions onto their market valuations since late 2022, feeding concerns of a massive AI bubble, baseline estimates of actual potential corporate profit gains remain significantly lower. Even entry-level jobs, the traditional proxy for mass unemployment since ChatGPT arrived, have challenged the view that AI is causing any systemic issues at large. Indeed, recent research shows an increase in entry level work. Productivity gains are also disappointing. Predictions for net productivity growth in highly exposed G7 economies hovers at a modest 0.4 to 1.3 percentage points over a 10-year horizon.
It’s true, there have been very public layoffs across Big Tech. Yet in most cases, the mention of AI isn’t a displacement story, but rather one about shuffling resources around. Tech giants would rather spend cash on Nvidia chips and training models than on established areas of their business. Combined CapEx spending from US Big Tech hyperscalers is projected to surpass $600 billion according to financial reports tracking Hyperscaler CapEx infrastructure debt, swallowing an unprecedented portion of their revenues, and more than double pre-ChatGPT levels. Layoffs and austerity in other areas of their business help balance the books. In many cases, the business logic is flawed. But because investors would punish them harshly if they didn’t present a robust AI strategy, cut they must.
While the issue is complex, there’s a simple thread running through the AI story: it’s simply not as good, or at the very least, universally applicable, as everyone thought. Pundits can churn out poor content on LinkedIn with abandon, and contextually richer chatbots are now pervasive. But established use cases that move the needle at a macroeconomic level remain few and far between.
More pressing for the AI optimists, is the fact public bodies, the media, and citizens are actively rallying around examples of corporate failure. Countless stories are emerging where red-faced executives have declared AI engagements a bust, forcing them to hire back boring old humans to actually get the work done.
Crucially, these costly rollbacks today aren’t nearly as expensive as they will be tomorrow. AI is currently subsidized twice over: directly by private equity, and indirectly by consumers covering higher utility bills and tax enticements. The capital to sustain the former, and the political will for the latter, are rapidly running out.
That means one thing: price increases. And if businesses are struggling to make the math work now, they’ll find it even harder when they have to cover a fairer share of the cost burden. And AI companies will find it equally difficult to justify the bill if, as Altman and Amodei now suggest, AI won’t free up as much money from the salaries as we were first led to believe.