Is AI forcing everyone to pick a side?

Across the tech industry, a new kind of structural pressure is building. Companies are being pushed into clearer positions as AI becomes embedded in how products are built, decisions are made, and value is created.

As time moves on, three paths are emerging: build, buy, or gradually lose relevance. It’s more of a sorting mechanism than a trend. 

AI has become a baseline capability 

The DeepFest team wrote about the end of the AI startup a few weeks back – because using AI is no longer a differentiator of its own. And this is the case beyond just the startup realm. According to McKinsey’s 2025 global survey, around 72% of organisations now use AI in at least one function.

At this level of adoption, AI starts to resemble electricity – expected, ambient, and woven into everything.

At the same time, the cost of building frontier models continues to climb. Stanford’s AI Index estimates GPT-4 training at roughly USD $78 million, with some projections exceeding $100 million.

And this creates an unusual dynamic: access to AI is broad, but ownership of the underlying capability is increasingly concentrated. 

A narrowing centre of ownership 

The foundation layer is coalescing around a small group of companies:

• OpenAI
• Google
• Anthropic
• Meta

Their advantage rests on a combination of capital, compute infrastructure, and access to specialised talent. Training costs have increased at an average rate of around 2.4× per year since 2016. 

As a result, the model layer resembles a high-barrier industry. Participation is limited, and the distance between leaders and followers continues to widen.

Against this backdrop, for most companies the strategic question shifts away from whether to build models, and towards how to position on top of them.

Three strategic positions 

1. Build

A small cohort of organisations are developing AI systems in-house.

They bring together proprietary datasets, research capability, and the capital required to sustain long development cycles. Meta’s investment in open-weight models offers one expression of this approach, combining internal capability with ecosystem leverage.

Building provides control over performance, cost structure, and direction. It also requires sustained investment and a tolerance for complexity.

2. Buy

A much larger group integrates AI through external platforms.

APIs from OpenAI, Microsoft, and Google allow teams to deploy AI quickly, experiment widely, and ship new features at pace.

This route lowers the barrier to entry and accelerates product cycles. It also places many companies on similar technical foundations, which compresses differentiation over time.

3. Drift

Some organisations adopt AI incrementally, without a clear point of view on where it compounds or how it influences their core product.

As AI capabilities spread, this middle ground becomes harder to sustain. Inference costs have dropped sharply – by as much as 280× in certain benchmarks between 2022 and 2024.

Lower costs enable faster adoption across the market. They also accelerate feature parity, making it easier for competitors to replicate surface-level improvements.

In that environment, unclear positioning gradually translates into reduced visibility.

Where advantage is accumulating 

As models become widely accessible, advantage shifts to other layers, including:

• Proprietary data
  Systems that learn from user behaviour (recommendations, feedback loops, domain-specific data) compound in value over time. Public models provide capability; private data shapes performance.
• Distribution
  Access to users becomes a powerful lever. Embedding AI into products that people already rely on creates continuity and depth. This is why incumbents with large installed bases can extend their position as they integrate AI into existing workflows.
• Workflow ownership
  Products that sit inside daily tasks (writing, coding, analysis, operations) gain resilience. They become part of how work gets done, rather than an additional tool alongside it.

These layers form the new competitive ground. 

A change in shape (not just speed) 

People often describe AI in terms of acceleration: faster development, quicker iteration, shorter cycles. But the deeper change here lies in structure. 

The widespread availability of powerful models flattens certain parts of the landscape, while raising the importance of others. Capability spreads, and differentiation inevitably relocates.

From this, we see an emerging market where many companies can build with AI, but far fewer can build enduring positions with it. 

What are the implications for builders and investors? 

For founders, the focus turns toward designing systems that improve with use. Data loops, user engagement, and integration depth begin to define long-term value.

For investors, attention shifts to control points – where companies own inputs, capture feedback, or sit closest to decision-making workflows.

For operators, the opportunity lies in applying AI where it compounds over time, embedding it into processes that benefit from continuous improvement.

The industry is already reorganising 

Very few organisations will go out of their way to articulate their position explicitly. But it can be seen through product decisions and hiring priorities – and of course, capital allocations. 

Over time, the differences between the three paths will become clearly visible – in growth, resilience, and relevance. 

At LEAP, these strategic choices take shape on the ground – through conversations in the Investor Lounge, targeted meetings in the Matchmaking Zone, and knowledge-sharing throughout keynotes and panels. 

Because as AI changes the entire tech landscape, positioning is becoming the defining advantage. 

Join us at LEAP from 31 August – 3 September 2026 to immerse yourself in the future of tech.