I’ve been watching this year’s developer conference season with a mix of fascination and bewilderment. Nvidia’s Jensen Huang just walked on stage and basically told us that everything we know about laptops is wrong. Not just wrong, but obsolete. The guy described a completely new way of using computers, and a completely new kind of machine to support it.
And my first thought was: really?
Look, I’m all for artificial intelligence pushing boundaries. As someone who writes about tech for a living, I’ve seen plenty of legitimate innovations. But there’s something about this particular push that feels less like solving a problem and more like creating one.
The Vergecast crew broke down all the announcements from Microsoft Build and Google I/O this week. We saw Gemini Spark, Nvidia’s RTX Spark, Microsoft’s Scout and Solara projects. AI agents popping up everywhere, doing everything, from scheduling your meetings to apparently “rethinking” what a laptop even is. The pitch is that we need specialized hardware to run these models locally, that our current machines simply aren’t cut out for the AI future.
Here’s where I get stuck as a developer. Most of the AI workflows I actually need don’t require me to run a model locally. Cloud APIs have gotten good enough that latency isn’t a real problem for my day-to-day work. I can spin up a notebook on any cloud provider and have access to considerably more compute than I’d ever need on a laptop. The question nobody seems to be answering is: what specific problem does an AI-native laptop solve that doesn’t already have a cheaper, more flexible solution?
The industry keeps telling us that AI is going to change everything. And maybe it will. But there’s a difference between a technology that enables new possibilities and one that requires us to buy new hardware to access them. The smartphone worked because it solved real problems people had. The cloud worked because it actually reduced costs and complexity for most use cases.
What I’m seeing feels more like hardware companies looking for a product cycle to chase. We’re deep into the “every company is an AI company” phase of this cycle, and the hardware push feels like the logical next step to keep the revenue flowing.
The thing that bugs me most is that we’re not even asking the fundamental question properly. It’s not “should laptops be more powerful to run AI models?” It’s “what are the actual workflows where local AI processing provides meaningful value over cloud alternatives?” Privacy might be one answer. Offline capability another. But I’m not seeing enough honest discussion about whether those tradeoffs are worth the premium hardware prices.
Maybe I’m just old school. I remember when “AI PC” meant something different, and the industry cycled through plenty of overhyped categories before. But I’d rather see compelling use cases that make me want these machines rather than being told I need them because the future is coming.
The Vergecast folks are taking calls about this new wave of AI products. They want to know how people really feel about the new daily format and all the tech announcements flooding in. I’m curious too. Because right now, I’m sitting on the fence, waiting for someone to show me what I’m supposedly missing. The hardware announcements are loud, but the actual value proposition feels remarkably quiet.