Murat Abdrakhmanov on which AI startups will survive and where to find a competitive advantage

You can also watch the recording of the presentation on our YouTube channel — this article summarizes the key points.

Murat Abdrakhmanov, Founder of MA7 Ventures

About Me

I have been working in venture capital for 11 years: my portfolio includes over 70 deals and 15 exits. I am also a serial tech entrepreneur and investor, including in Higgsfield.

Over the years, we have accumulated significant experience, and now we are actively investing — primarily in AI.

Startup Classification

We divide startups into three phases.

Phase One — copy-paste models.
These are startups that replicate existing and proven business models from other markets. This approach worked for a long time: you could see what was successful abroad and replicate it locally. It is a simple and clear path, but today it is practically exhausted — at least in Kazakhstan.

Phase Two — differentiation.
Startups begin with copy-paste but then find their niche, delve deeper into the problem, and create unique value.

An example is Prosper Pay, which is transforming from a "salary advance" service into a neobank for the gig economy.
Or Hero’s Journey, which reimagined the Invictus format and is scaling to international markets.

Phase Three — AI-first / AI-native startups.
These are the projects we are currently working with.

Investment Strategy

In the last year, we have revised our investment approach. Previously, we entered at the Seed stage, but now we enter at the Pre-Seed stage.

The reason is simple: AI startups develop extremely quickly.
For example, Higgsfield became a unicorn in just 12 months: from $5 million in revenue to a $330 million Series A round.

In 2025, we invested in eight AI-native startups with global potential.
We have reduced check sizes and are entering at the pre-revenue stage — even before revenue appears.

AI is becoming a commodity — like the internet or electricity. It is basic infrastructure without which no industry will soon be able to function.

Why It Is Difficult to Build AI Startups Now

On one hand, major players — LLM developers and infrastructure providers — are moving into verticals and beginning to compete directly with startups.

Any new feature within large models can "kill" hundreds of smaller products.

On the other hand, traditional SaaS companies are actively integrating AI and defending their positions.

As a result, the market is becoming extremely competitive, and finding a sustainable advantage is becoming more difficult.

Main AI Trends

— AI agents are becoming mainstream — SaaS platforms are transforming into agentic systems
— SaaS is losing valuation — valuations are falling for both new and mature companies
— wrappers over LLMs no longer work
— vertical AI products with deep niche expertise are winning
— Open source is developing rapidly, especially Chinese models
— The gap between open source and commercial solutions is shrinking to just a few months.

There is also growing interest in implementing AI in the corporate sector — especially in closed infrastructures with high security requirements.

A separate trend is the development of Physical World Models: AI for industry, transportation, and robotics. About $1.5 billion is already being invested in this direction.

Where to Find Competitive Advantage

1. Proprietary data
Unique data that the company collects itself. This is the most sustainable moat.

2. Distribution
Sometimes more important than the product. The winner is the one who has access to the user.

3. Domain expertise
Deep expertise in a niche is critically important.

4. Speed — execution velocity
Companies that quickly develop and release products win.

For example, teams with weekly releases and high work intensity.