What Dell and Salesforce Tell Us About Where the AI Money Actually Goes

Haven't we seen this movie before?

I called my old colleague at Siemens last week, and we were laughing about the PLM software wars of the early 2000s. Everyone was buying licenses, hiring consultants, installing servers. But actually getting value out of those systems? That took another decade. Some companies never figured it out.

The AI buildout feels similar. Companies are buying Nvidia chips and Dell servers like their lives depend on it. They're terrified of being left behind. But the actual applications, the software that makes all that compute useful, that's still being figured out. Salesforce is trying to bolt AI onto their existing CRM platform, and honestly, it remains unclear whether that's going to work or whether some startup will just build something native from scratch.

What does this mean for robotics?

Now, you might be wondering why I'm writing about enterprise software companies. Fair point. But this dynamic matters for our industry too.

The robotics sector is going through the same split. Companies selling robot hardware, sensors, actuators, those are the Dell equivalents. They're doing fine, mostly. But companies trying to build AI-native robot software? They're in Salesforce's position, trying to prove they can adapt before someone else eats their lunch.

I'll be honest, I don't know how this plays out. The integration challenges are enormous. When I was working on the KR C4 controller (this would've been around 2012), we spent years just getting basic sensor fusion working reliably. The idea that you can just drop a large language model into an industrial robot and have it, sort of, figure things out? I'm skeptical. Not because the technology isn't impressive, but because manufacturing environments are unforgiving in ways that chatbots aren't.

So who should we be watching?

Dell's numbers suggest the infrastructure buildout has legs. $60 billion in AI server sales means someone's buying, and that demand doesn't appear to be slowing down. Whether all that compute actually gets used productively is a different question.

Salesforce's wobble is more concerning if you're in the software business. The market is basically saying: we're not sure you'll survive this transition. That's rough for a company that's been dominant for two decades.

For robotics, I think the lesson is this: if you're selling hardware that enables AI, you're probably in good shape. If you're selling software that AI might replace or fundamentally change, you'd better have a clear story about how you're adapting. The middle ground, companies that are sort of doing both, that's where things get murky.

I don't have a neat conclusion here. We're in the messy middle of a technology transition, and the winners and losers won't be clear for years. But these two earnings reports, one triumphant, one nervous, they're a snapshot of where we are right now. Infrastructure's hot. Applications are uncertain. And everyone's spending money they hope will pay off eventually.

James Chen · 7 hours ago · 4 min