AI Startups Worth Building

In partnership with

Hey there,

While researching what's going on in the AI space lately, I came across these events. 

They made me rethink which AI businesses are actually worth building right now.

Let me show you what I mean.

The $800 Billion Problem

Deutsche Bank just revealed that the AI bubble is literally the only thing preventing US from sliding into recession.

They've identified an $800 billion gap between projected AI revenues and the actual investment needed to keep this whole thing running.

George Saravelos from Deutsche Bank put it bluntly: "Nvidia is currently carrying the weight of US economic growth."

One company. Carrying an entire economy.

If you're building on expensive cloud AI infrastructure or counting on costs coming down, the economics don't add up. When this corrects, access and pricing will change dramatically.

Entry-Level Jobs Are Disappearing

Meanwhile, Goodwill's CEO is preparing for an "influx of unemployed young people" because AI is destroying entry-level jobs right now.

"What I'm seeing is of the overall unemployment, people without college degrees have no jobs," Steve Preston said.

The same AI boom propping up the economy is eliminating the jobs that help young people build careers.

Here’s an opportunity: There's massive demand for reskilling, but not the traditional kind. Young people need practical skills that work alongside AI. Training programs teaching AI-augmented workflows, industry-specific prompt engineering, or AI tool management are going to explode.

The Psychological Cost Nobody's Talking About

While everyone debates economics and regulation, people are developing genuine psychological dependencies on AI systems.

Researchers are documenting cases of "AI psychosis" - people convinced they have special relationships with ChatGPT, creating "seed prompts" to "awaken" it, building entire belief systems around these interactions.

These aren't isolated cases. There are entire communities sharing techniques to make AI "remember" them between sessions, creating what they call "spores" and "voices."

OpenAI got so concerned they deployed GPT-5 as a moderator to intervene when conversations get too intense.

Tools that create healthier AI relationships - usage limits, mandatory breaks, features encouraging human interaction - will become competitive advantages as dependency concerns grow. This will be huge for enterprise AI tools.

Connecting the Dots

Let’s look at what's happening:

• Our economy depends on AI infrastructure spending that doesn't match revenue reality

• We're eliminating traditional career pathways for an entire generation

• Tech leaders are fighting any meaningful oversight

• Users are forming unhealthy dependencies on these systems

When that $800 billion gap hits reality, we'll have:

• A generation without job prospects

• An economy that forgot how to grow without AI spending

• People psychologically dependent on systems that might become restricted

• Sudden, poorly-designed regulations changing everything overnight

Most AI startups are building for a world where this bubble continues forever.

That's a mistake.

Where the Real Opportunity Lies

Here's what I'm seeing work, regardless of what happens with the AI bubble:

1. Local AI Infrastructure

The LocalLLaMA community hit 500,000 members and launched a Discord for technical discussions. They're building tools for local AI inference on consumer hardware.

Companies are realizing they don't want sensitive data going through cloud APIs.

Build:

• Local deployment consulting for specific industries

• Hardware optimization services that make local AI practical

• Custom model fine-tuning for niche use cases

• Privacy-focused AI solutions for regulated industries

Why it works: Survives pricing changes, regulation, and cloud service restrictions. Becomes MORE valuable if cloud AI gets expensive or limited.

2. AI Skills Training (The Practical Kind)

Not coding bootcamps. Practical training on:

• Using AI tools for specific professions (legal research, financial analysis, medical coding)

• Building AI workflows that enhance human work

• Quality control and AI output verification

• Managing AI teams in traditional businesses

Why it works: Addresses the job displacement problem directly. Demand increases whether AI succeeds or fails - either people need to learn to work with it, or they need to differentiate from it.

3. AI Wellness and Management Tools

As dependency becomes recognized:

• Usage tracking and healthy limit tools

• AI interaction coaching and best practices

• Tools helping people use AI productively without becoming dependent

• Corporate AI wellness programs (yes, this will be mandatory)

Why it works: Regulatory pressure will make this required. Companies will need to show they're managing AI usage responsibly. First movers will set the standards everyone else has to follow.

4. Hyper-Niche AI Applications

Instead of competing with ChatGPT:

• Industry-specific tools solving one problem extremely well

• AI for industries Big Tech ignores (construction, agriculture, local government)

• Accessibility-focused applications

• Tools helping small businesses compete with AI-powered enterprises

Why it works: These solve real problems with measurable ROI. They don't depend on hype or cheap API access to justify their existence.

5. "AI Resilient" Business Models

Services that become MORE valuable if AI access becomes restricted or expensive:

• Human verification for AI-generated content

• AI output quality assurance for regulated industries

• Hybrid human-AI workflows for high-stakes decisions

• Emergency backup systems for AI-dependent businesses

Why it works: You're building the insurance policy everyone will need. When (not if) something goes wrong with AI access or quality, these businesses will be essential.

How to Think About It

Sam Altman has talked about OpenAI's "small family of devices" that will transform computing. But the real innovation is happening with people building AI tools that don't require massive data centers.

The startups that will still exist in five years are asking:

• What happens when AI costs 10x more than today?

• What if cloud AI services become restricted or regulated?

• How do I build value that doesn't depend on infinite cheap AI access?

• What human problem am I actually solving, not just automating?

The AI boom democratized access to powerful technology. But it also created a gold rush mentality where everyone's building the same infrastructure-dependent businesses.

The opportunity isn't in doing what everyone else is doing, but in building what will matter when the fundamentals change.

Because when Deutsche Bank's $800 billion gap becomes reality, the survivors won't be the ones burning cash on infrastructure.

They'll be the ones who built real solutions for real problems.

What are you building? And more importantly - could it survive without cheap cloud APIs and hype-driven funding?

- Aashish

P.S. If you're working on any of these opportunities or have thoughts on what's actually worth building right now, reply and let's discuss. I'm always looking for entrepreneurs thinking beyond the current bubble.

Startups who switch to Intercom can save up to $12,000/year

Startups who read beehiiv can receive a 90% discount on Intercom's AI-first customer service platform, plus Fin—the #1 AI agent for customer service—free for a full year.

That's like having a full-time human support agent at no cost.

What’s included?

• 6 Advanced Seats

• Fin Copilot for free

• 300 Fin Resolutions per month

Who’s eligible?

Intercom’s program is for high-growth, high-potential companies that are:

• Up to series A (including A)

• Currently not an Intercom customer

• Up to 15 employees

Apply now