HashiCorp's Mitchell Hashimoto Says Companies Are Losing Their Minds Over AI
Everyone keeps telling us we're in the golden age of AI adoption. That companies are finally "getting smart" about artificial intelligence.
Bullshit.
Mitchell Hashimoto, the HashiCorp co-founder who gave us Terraform and actually knows how to build infrastructure that doesn't fall over, just dropped a truth bomb that's got the entire tech industry squirming. His viral post about companies suffering from "AI psychosis" racked up 1,857 points and 1,048 comments on Hacker News—the kind of engagement that only happens when someone hits a nerve.
<> "I believe there are entire companies right now under AI psychosis—adopting AI everywhere, reorganizing teams around it, and spending heavily without clear business cases or proven ROI."/>
This isn't some random blogger throwing shade. Hashimoto built HashiCorp into an infrastructure giant before IBM acquired it in 2024. He's watched companies scale from garage startups to enterprise behemoths. And he's telling us that executives are making strategic decisions under extreme AI hype, treating ChatGPT's existence as justification to pivot entire organizations.
I've seen this movie before. Remember when everything had to be "mobile-first"? When "cloud-native" was the magic phrase that opened VC wallets? When "blockchain" could transform your grandmother's knitting circle into a unicorn?
The pattern is always the same:
- Leadership gets FOMO about the hot new thing
- Budgets get reshuffled to chase the trend
- Teams get reorganized around solutions looking for problems
- ROI becomes a "nice-to-have" instead of a requirement
But AI psychosis feels different. More intense. The stakes are higher because the technology is genuinely powerful—just not in the way most companies are using it.
Hashimoto's diagnosis cuts deeper than surface-level criticism. He's talking about organizational delusion—entire C-suites detached from practical constraints, approving AI initiatives because "AI" sounds compelling in board meetings.
The Elephant in the Room
Here's what nobody wants to admit: most enterprise AI deployments are expensive demos.
Sure, there are real wins. Code completion actually helps developers. Customer support triage can deflect tickets. Document summarization saves time. But these targeted use cases get lost in the noise of companies trying to "AI transform" everything from their hiring process to their coffee machines.
The technical reality is brutal:
1. LLM inference costs scale unpredictably when you actually have users
2. Hallucinations aren't a bug you can patch—they're fundamental to how these systems work
3. Integration complexity turns simple AI features into maintenance nightmares
4. Security and compliance requirements multiply when AI touches real data
Developers know this. We're the ones getting vague requirements like "make it more AI-powered" and "add some machine learning magic." We're building retrieval systems, access controls, audit logging, and fallback behaviors while executives think they ordered a simple chatbot.
The worst part? Companies are reorganizing entire engineering teams around AI-first strategies without understanding what they're actually trying to solve. It's innovation theater disguised as digital transformation.
Hashimoto's post resonated because it named something we've all been feeling. The emperor has no clothes, and his robes are made of transformer tokens.
Will this change anything? Probably not. The hype cycle has its own momentum, and there's too much money chasing the AI dream. But at least now we have a name for the madness.
AI psychosis. It's real, it's spreading, and your company might already be infected.
