AI & Tech Brief: Beyond the hyperscalers

David Stout, CEO of webAI, has decided that the future of enterprise AI is 'sovereign' — as in, your very own personal model, not some rented mind from a hyperscaler. The pitch is that businesses will prefer to run their own models rather than 'renting frontier models' from the cloud giants. This is, of course, the same crowd that couldn't be bothered to install their own email servers a decade ago. But sure, let's pretend that every mid-sized manufacturer will suddenly become an AI model operator, because nothing screams 'efficiency' like running your own inference cluster while your competitors are blithely API-calling their way through the quarter.

The 'sovereign AI' argument is catnip for executives who distrust Big Tech but trust themselves to manage something exponentially more complex than their current SAP deployment. Stout is betting that the post-hype hangover from hyperscaler lock-in will drive businesses to demand 'independence' — conveniently ignoring that most companies can barely secure their Office 365 tenant. And what happens when their sovereign model starts hallucinating supply chain schedules? Will they call it 'artisanal error'? The real magic is that sovereign AI solves the wrong problem: it gives you control over the model but not over the compute costs or the data-centre maintenance nightmares.

The punchline, as ever, is that the hyperscalers are probably laughing their way to the bank. They'll happily sell you the GPUs, the cloud credits, and eventually the consulting services to 'sovereignly' migrate back when your homegrown model proves insufficient. Meanwhile, Stout gets to ride the sovereignty wave, and we get another acronym to file alongside 'edge AI', 'federated learning', and 'private cloud' — all of which were supposed to make us free and instead just made us more confused. Ah, progress.