Public transcript

Davos 2026 LIVE: Microsoft CEO Satya Nadella Attends World Economic Forum 2026 In Davos | WION

By: WION
Length: 1 h 12 min
Language: English
Text from: the video's captions
Transcribed: 14 October 2026

Read the full transcript Open on YouTube

Notes in English.

Summary

At the World Economic Forum in Davos 2026, Microsoft CEO Satya Nadella sits down with BlackRock's Larry Fink for a wide-ranging conversation on AI diffusion, productivity, energy, and sovereignty. Nadella argues that AI is spreading more evenly across the world than any prior technology, that "tokens per dollar per watt" will drive GDP growth, and that firms must redesign workflows, build skills, and control their own models and data context. He also shares advice for Europe on competitiveness and rejects the idea of one dominant AI model, describing a multi-model, orchestration-based future. The transcript repeats the same conversation twice, and the final minutes contain only non-verbal audio.

Key points

Nadella says AI models and their outputs are available almost everywhere, spreading more evenly than the PC or early mobile eras, citing a rural Indian farmer who used an early GPT bot to reason over farm subsidies in a local language.
He argues a telltale sign of an AI bubble would be if only tech firms and the supply side are discussed; diffusion of benefits across sectors and countries is what prevents one.
Nadella describes using Copilot to prepare for his roughly 50 bilateral meetings at Davos, saying it flattens and inverts information flow inside organizations and forces structural redesign.
His formula for firms is mindset (change the workflow with the technology), skill set (use it and learn guardrails to trust it), and data set (context engineering with the firm's tacit knowledge).
He claims GDP growth will correlate with "tokens per dollar per watt," and that token pricing drops by roughly half every 3 months.
On Europe, he says European competitiveness depends on globally competitive output, praises the German Mittelstand, and warns that firms without control over their own models and tacit knowledge are leaking enterprise value.
He predicts a multi-model world where firms orchestrate closed-source, open-source, and their own models with their data to create IP they control.
Larry Fink adds that at BlackRock, computations that took 12 hours now take minutes when processing 14 trillion dollars of other people's money.

Questions it answers

00:00AI diffusion worldwide

Is AI spreading evenly across the world or leaving people behind?

Nadella says models and their outputs are available almost everywhere, more so than in the PC or early mobile eras. He cites a rural Indian farmer using an early GPT bot to reason over farm subsidies in a local language, showing AI can bring agency to those who lacked it.

Mobile and connectivity rails let tokens be delivered evenly around the world.
Capital investment and a supportive environment, including government action on the grid, remain necessary conditions.

03:40The AI bubble question

Could the current AI boom be a bubble?

Nadella says a telltale sign of a bubble would be if only tech firms and the supply side are discussed. He is confident because diffusion is already producing real outcomes, like AI-accelerated clinical trials, and countries that diffuse AI fastest will win.

For it not to be a bubble, benefits must be evenly spread beyond tech firms.
Diffusion, not technology creation, determines the winners.

06:02AI and organizational change

How is AI changing the structure of companies?

Nadella describes using Copilot to prepare briefs for his ~50 Davos bilaterals, getting a 360 view he shares across all functions, flattening and inverting traditional information flow. Firms must redesign structures so information flows freely.

AI flattens information flow, requiring structural redesign.
His formula is mindset, skill set, and data set (context engineering).

10:48Large vs. small firms

Who is adopting AI faster, big companies or small ones?

Nadella calls it a barbell: small starting companies adopt easily, while large firms have data and relationships but face bigger change-management challenges. Neither incumbents nor new entrants can coast.

Large firms risk being schooled by small players achieving scale with these tools.
Productivity results depend on leadership will within each organization.

12:08Country differences and energy

Does AI adoption differ by country, and what role does power play?

Nadella says talent quality in Jakarta, Istanbul, or Mexico City is not that different from Seattle, though the US shows more energy and risk capital. He claims GDP growth will correlate with tokens per dollar per watt, and token pricing drops by about half every 3 months.

Financial sector AI adoption is far faster than cloud adoption was.
Cheap energy, construction costs, and silicon cost curves determine total TCO.

16:12A message for Europe

What should Europe do about AI and competitiveness?

Nadella says European competitiveness is about globally competitive output, not protection, praising the German Mittelstand's engineering prowess. Europe should build locally and think globally rather than only focusing on data protection.

Europe's human capital is world-class; it must invest in energy and tokens locally.
Privacy and safety leadership are features, but must be complemented with global competitiveness.

19:05Sovereignty and the multi-model future

Is data sovereignty misunderstood, and will one AI model dominate?

Nadella says the key sovereignty issue is firms embedding their tacit knowledge in models they control, not where data centers sit. He predicts a multi-model world where firms orchestrate closed-source, open-source, and their own models with their data to create IP they control.

Firms without control over their models leak enterprise value to model companies.
The IP of any firm is how it orchestrates models with its own context.

24:15Conversation repeats and close

What happens in the rest of the recording?

The same conversation replays from roughly 24:15 onward, and the final minutes from about 1:09:28 contain only non-verbal audio.

The transcript duplicates the full interview.
No new content appears after the repeat.

Notes

Overview

Satya Nadella, in conversation with Larry Fink at Davos 2026, discusses how AI diffuses across economies, firms, and countries, and what conditions determine who wins.

Global diffusion of AI

Nadella argues AI outputs ("tokens") are delivered more evenly worldwide than in the PC or early mobile eras, thanks to mobile and connectivity rails.
He cites a 2023 demo of a rural Indian farmer using an early GPT bot to reason over farm subsidies in a local language, even showing agentic behavior like completing a form.
Necessary conditions include capital investment, an environment that attracts capital, and government action on the grid, since private capital alone cannot modernize power infrastructure.

The AI bubble question

A telltale sign of a bubble would be if conversation stays focused only on tech firms and the supply side.
He is confident because diffusion is happening, e.g., AI-accelerated clinical trials bringing drugs to market.
Countries and companies that diffuse the technology fastest will be the winners, not the technology creators.

Changing organizations

Preparing for his ~50 bilateral meetings at Davos, Nadella now asks Copilot for a brief; it gives a 360 view and he shares it across functions, flattening information flow.
His formula: mindset (change the workflow with the technology), skill set (use it, learn guardrails to trust it), and data set (context engineering with the firm's tacit knowledge).
Adoption is a "barbell": small starting companies adopt easily; large firms have scale advantages but bigger change-management challenges.

Energy and economics

He claims GDP growth will correlate with "tokens per dollar per watt"; token pricing drops by roughly half every 3 months.
Larry Fink notes BlackRock computations that took 12 hours now take minutes, processing 14 trillion dollars of other people's money.

Europe and sovereignty

European competitiveness is about globally competitive output, not protection; he praises the German Mittelstand.
The big sovereignty issue is firms embedding their tacit knowledge in models they control, rather than leaking enterprise value to model companies.

The multi-model future

Nadella predicts no single dominant model: firms will orchestrate closed-source, open-source, and their own models with their data, creating IP they control.

Text from the video's own captions. Summary and notes written by AI from the transcript, so check anything important against the recording.