The day we stop calling them AI startups

Imagine a founder stepping onto a stage and announcing they’re building an internet startup. 

You’d raise an eyebrow – the internet is infrastructure, not a startup differentiator. 

But in the early 2000s, startups did proudly describe themselves as internet companies. That label disappeared over time. The internet didn’t vanish, obviously; it became assumed. 

And AI might be heading in the same direction. The team at DeepFest wrote about this recently. Today, thousands of companies describe themselves as AI startups. Investors track the category closely, and product launches revolve around the latest models. But AI is spreading across industries – and it’s fast becoming part of the assumed infrastructure of new tech-native businesses. 

From product to infrastructure

The numbers already suggest that shift is underway.

According to McKinsey’s global survey on AI, 78% of organisations reported using AI in 2024, up from 55% the previous year. As the technology becomes more accessible, adoption is spreading across industries rather than remaining concentrated among specialist firms.

At the same time, the economic momentum continues to build. The Stanford AI Index notes that generative AI attracted USD $33.9 billion in private investment in 2024, increasing nearly 19% year-on-year.

But the most important trend is diffusion. 

AI is being embedded into the software and systems that already exist: 

• logistics platforms optimising supply chains
• financial tools detecting fraud
• biotech companies accelerating drug discovery
• creative software generating design and media

So it’s behaving less like a standalone product category and more like a foundational technology layer – similar to the internet, mobile computing, or cloud infrastructure.

Economists describe innovations like this as general-purpose technologies: breakthroughs with real impact that emerges as they spread across the entire economy. 

And once that diffusion happens, the label itself becomes redundant. 

Where differentiation moves

If AI becomes infrastructure, founders will have to compete on different terrain.

Four areas are already emerging as fresh differentiators: 

1. Proprietary data

Large models are becoming widely available. But high-quality domain data remains scarce.

Startups with access to specialised datasets (think clinical records, industrial sensor networks, financial transaction histories) can build products competitors cannot easily replicate.

The advantage shifts from raw intelligence to contextual intelligence.

2. Distribution and trust

When many companies can access similar models, technology alone is rarely enough. The real differentiators become distribution channels, brand credibility and trusted relationships with users. In sectors like healthcare, finance and government, trust can be as valuable as the technology itself.

Infrastructure still needs human interfaces.

3. Governance and compliance

As AI becomes embedded in critical systems, regulatory scrutiny intensifies. The EU AI Act and similar frameworks around the world signal a shift from high-level strategy toward enforceable compliance requirements.

Startups that succeed will treat governance as a design principle rather than an afterthought – building systems that are auditable, explainable, and compliant from day one.

4. Compute efficiency

AI is also becoming physical infrastructure. Data centres consumed around 415 terawatt-hours of electricity in 2024 – roughly 1.5% of global electricity demand. By 2030, that demand could more than double to about 945 TWh, driven largely by AI workloads.

In that environment, efficiency becomes a competitive advantage. The companies that thrive may not be those with the largest models, but those that deliver powerful capabilities with the lowest compute footprint.

From AI-native to AI-ambient

There’s also a cultural transition underway. As AI becomes embedded across products and workflows, it begins to fade into the background.

Instead of consciously interacting with AI tools, people simply use software that happens to be intelligent.

AI becomes ambient.

And that raises an important question for founders and builders today: If every competitor had access to the same frontier models tomorrow, would your business still work?

If the answer is no, the opportunity may not lie in building an AI company. Instead, it lies in building a great company in a world where AI is simply part of the infrastructure.

Join us at LEAP from 31 August – 3 September 2026 to explore what comes next when AI stops being the headline and starts being the baseline.