Key takeaways

• The AI infrastructure market is not currently in a bubble, suggesting continued growth potential.
• Nebius faces significant competition from hyperscalers with much larger capital expenditures.
• The threat of excessive industry consolidation could impact Nebius’s business strategy.
• AI adoption is still in its early stages, with many companies just beginning to integrate AI into their operations.
• Specialized AI models can outperform universal models in specific applications.
• There are numerous unsolved tasks in AI, driving the need for advanced solutions.
• The decreasing cost of AI models leads to increased consumption and the ability to solve more complex tasks.
• Both open-source and specialized AI models have room to coexist in the market.
• Physical expansion is crucial for infrastructure companies but is often slowed by real-world complications.
• The AI infrastructure race is ongoing, with significant implications for businesses and developers.
• Nebius’s current capital expenditure is significantly lower than that of its competitors.
• The market for AI solutions is robust, with ample opportunities for growth and innovation.
• The complexity of AI tasks is increasing, necessitating more sophisticated models and solutions.
• AI adoption is expected to grow significantly, with many more use cases emerging in the near future.
• The balance between open-source and specialized models is key to addressing diverse market needs.

Guest intro

Roman Chernin is Co-Founder and Chief Business Officer of Nebius, the AI infrastructure company building and operating large-scale compute clusters for AI labs, enterprises, and developers. He previously spent 12 years at Yandex, where he led the Search platform and later served as CEO of the Geoservices business unit.

The threat of industry consolidation

• The main threat for Nebius is excessive consolidation in the industry.

  — Roman Chernin

• Consolidation could limit opportunities for smaller companies like Nebius.

• We have to move; the AI infrastructure race is on.

  — Roman Chernin

• The competitive landscape is rapidly changing, necessitating quick adaptation.
• Nebius must navigate the challenges posed by larger, more established competitors.
• Industry consolidation may lead to fewer players dominating the market.
• Smaller companies may struggle to compete with the financial power of hyperscalers.
• Strategic partnerships could be a way to mitigate the risks of consolidation.
• The pace of consolidation could dictate the future landscape of AI infrastructure.

• Our capex program this year is $2,025,000,000,000 dollars; our competitors hyperscalers have eight times bigger.

  — Roman Chernin

Competing against hyperscalers

• Nebius operates in a capital-intensive industry dominated by hyperscalers.

• Nebius is competing against hyperscalers with significantly larger capital expenditures.

  — Roman Chernin

• The financial disparity presents a significant challenge for Nebius.
• Nebius’s strategy must account for the financial muscle of its competitors.
• Competing with hyperscalers requires innovation and strategic resource allocation.
• The scale of investment by hyperscalers dwarfs that of smaller companies.
• Nebius must leverage its strengths to carve out a niche in the market.
• The competition is fierce, with hyperscalers having a substantial advantage.

The state of AI infrastructure

• I don’t believe we’re in an AI infrastructure bubble right now.

  — Roman Chernin

• The market is perceived as having strong growth potential.
• Investment in AI infrastructure continues to be robust.
• Current market conditions do not suggest an impending bubble.
• The demand for AI infrastructure is expected to grow steadily.
• Companies are increasingly investing in AI capabilities.
• The perception of a bubble can influence investment decisions.
• The AI infrastructure market is still in its formative stages.

Early stages of AI adoption

• We are just at the beginning of real AI adoption, with many more use cases to emerge.

  — Roman Chernin

• Most companies are only starting to integrate AI into their operations.

• It’s just the first steps.

  — Roman Chernin

• The potential for AI adoption is vast, with numerous untapped opportunities.
• The pace of adoption varies across industries and regions.
• Companies are exploring AI to enhance efficiency and innovation.
• Early adopters may gain a competitive advantage in their respective fields.
• The evolution of AI technology will drive further adoption.

Specialized vs. universal models

• Specialized models can outperform universal models in specific use cases.

  — Roman Chernin

• Tailored solutions can optimize performance and cost.
• The choice between specialized and universal models depends on the application.
• Specialized models offer advantages in niche areas.
• Universal models provide broader applicability but may lack specificity.
• The development of specialized models is a strategic focus for many companies.
• The balance between the two approaches is crucial for effective AI deployment.

• There are many unsolved tasks in AI that continue to drive growth for frontier models.

  — Roman Chernin

Economics of AI models

• The cost of AI models is decreasing, leading to increased consumption.

• We can solve more complex tasks with the same budget.

  — Roman Chernin

• Lower costs make AI solutions more accessible to a broader range of companies.
• The economics of AI models influence market dynamics and adoption rates.
• Cheaper AI models enable the tackling of previously unsolvable tasks.
• The trend towards cost reduction is expected to continue.
• Companies can achieve more with the same level of investment.
• The economic viability of AI solutions is a key consideration for businesses.

Market space for AI models

• There is enough space in the market for both open source and specialized models to coexist.

  — Roman Chernin

• The diversity of AI solutions reflects the varied needs of the market.
• Open-source models offer flexibility and community-driven development.
• Specialized models cater to specific industry requirements.
• The coexistence of different models promotes innovation and choice.
• Companies can leverage both types of models to address diverse challenges.
• The market is large enough to accommodate multiple approaches.
• The interplay between open-source and specialized models is a dynamic aspect of the AI landscape.

Challenges of physical expansion

• Infrastructure companies must expand physically to meet demand.

• There are a lot of complications of the real world that prevent you to move fast enough.

  — Roman Chernin

• Physical expansion is crucial for maintaining competitiveness.
• Real-world challenges can slow the growth of infrastructure companies.
• Companies must navigate logistical and regulatory hurdles.
• The scale of expansion impacts the company’s ability to serve its customers.
• Physical capacity is a key determinant of success in the infrastructure sector.
• Balancing growth with operational challenges is a strategic priority.