FULL: Microsoft CEO Nadella in Conversation With BlackRock’s Fink at World Economic Forum | AI1G
- By
- DRM News
- Length
- 35 min
- Language
- English
- Text from
- the video's captions
- Transcribed
- 13 October 2026
Summary
At the World Economic Forum in Davos, BlackRock's Larry Fink interviews Microsoft CEO Satya Nadella about AI's evolution, diffusion across economies, and its impact on organizations and countries. Nadella argues AI is a continuation of the long arc of computation and that its value depends on turning tokens into real productivity gains in health, education, and industry. He discusses organizational change, the barbell effect between small and large firms, Europe's competitiveness, data sovereignty, and why the future will be a multimodel world.
Key points
- Nadella frames AI as part of a continuous 70-year arc of digitizing people, places, and things to build analytical and predictive power.
- He traces AI's progress in software engineering from GitHub Copilot code completions to chat, agent mode, and fully autonomous agents working 24/7.
- Fink says BlackRock can now compute in minutes what took 12 hours, processing $14 trillion of other people's money.
- Nadella says society will lose social permission to use scarce energy for tokens unless they improve health, education, and public sector outcomes.
- Diffusion depends on tokens per dollar per watt getting more efficient, plus broad skilling of the workforce.
- AI adoption is a barbell: small greenfield companies adopt easily while large incumbents face a bigger change management challenge.
- Nadella argues Europe should focus on global competitiveness of its products rather than just data protection, and that firm-level sovereignty over tacit knowledge in models matters most.
- He predicts a multimodel world where firms orchestrate multiple models with their own data and context engineering as their IP.
Questions it answers
00:00AI and the arc of computation
How does Nadella see AI in the history of computing?
He frames it as the latest step in a 70-year arc of digitizing people, places, and things to build predictive power, comparable to or greater than the PC, web, mobile, and cloud eras.
- AI is a continuation of a continuous arc of computation.
- Software's low marginal cost enables ever-greater insight and capability.
01:41AI in software engineering
How has AI changed software development?
Nadella's belief built up as GitHub Copilot moved from code completions to chat, agent mode, and autonomous agents working 24/7, though the developer retains agency and levels of abstraction shift.
- AI progressed from completions to fully autonomous agents.
- Human agency remains central; abstraction levels will change.
- AI lets documents transform into websites or apps via code.
04:47Productivity at BlackRock
What results is BlackRock seeing from AI?
Fink says computations that took 12 hours now take minutes, processing $14 trillion of other people's money across hundreds of thousands of mandates, which he says would be impossible without the technology.
- 12-hour computations now take minutes at BlackRock.
- $14 trillion in assets is processed with AI support.
06:11Diffusion and social permission
Why does diffusion of AI matter so much?
Nadella says society will quickly lose social permission to spend scarce energy generating tokens unless they improve health, education, public sector efficiency, and private sector competitiveness everywhere.
- Diffusion is everything; tokens must produce useful outcomes.
- Supply side requires tokens per dollar per watt to improve monotonically.
- Token factories will spread like electricity via a ubiquitous grid.
12:11Global south and skilling
Will AI reach less-developed economies evenly?
Nadella cites a rural Indian farmer using an early GPT bot in a local language to reason over farm subsidies and complete a form, showing AI can restore agency where technology was previously inaccessible, but capital investment, grid modernization, and skilling are necessary conditions.
- Models are available everywhere more evenly than PCs or smartphones were.
- Skilling is the factor most correlated with diffusion.
- Public capital is needed for grid modernization.
14:53Bubble fears
Is there an AI 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 AI is already accelerating things like drug clinical trials, and demand worldwide depends on local surplus everywhere.
- If only tech firms benefit, that signals a bubble.
- 50% of Microsoft's capital spending is outside the United States.
- Growth cannot be driven by capital expenditure alone.
17:21Organizations redesigned
How must companies change as AI diffuses?
Nadella describes how Copilot flattens information flow — his Davos meeting prep now comes as a 360 brief he shares across functions — requiring structural redesign, with a formula of mindset, skills, and context/data engineering.
- Information flow inverts and flattens organizations.
- Formula: mindset, skill set, and data/context set.
- Productivity results require hard change-management work.
22:48Barbell adoption
Who is adopting AI — big or small companies?
It is a barbell: small greenfield companies adopt easily, while large incumbents have data and relationships but face a bigger change management challenge and risk being schooled by smaller, faster movers.
- Small firms start fresh with the tools.
- Large firms must keep their rate of change up.
- Neither incumbents nor entrants can coast.
Notes
The arc of computation and AI's place in it
- Nadella describes 70 years of computing as one continuous arc: digitizing artifacts about people, places, and things, then building analytical and predictive power on top.
- AI is of the same class as the PC, web, mobile, and cloud — possibly greater — because software is a malleable resource without traditional marginal cost economics.
- His confidence in this generation of AI was built watching GitHub Copilot do code completions, then chat, then agent mode, and now fully autonomous agents that can work 24/7 on an entire project.
- He compares the moment to the early 1980s: nobody predicted 4 billion people would type daily, and a whole class of
Text from the video's own captions. Summary and notes written by AI from the transcript, so check anything important against the recording.