Nadella criticizes AI giants and bets on cheaper models

What happened?

In an interview with The Next Web, Satya Nadella, CEO of Microsoft, openly criticized the major players in artificial intelligence, including OpenAI (in which Microsoft has invested billions). Nadella pointed out that these companies cannot keep promising massive job destruction while demanding the power to build whatever they want without accountability. In response, Microsoft is betting on cheaper AI models that give customers more control and seek to restore public trust. The interview, published on April 15, 2025, reflects a significant shift in tone from previous years, when Nadella was one of the main advocates of collaboration with OpenAI.

Why is it important?

Nadella's statements mark a turning point in the dominant narrative about AI. Until now, major labs (OpenAI, Google DeepMind, Anthropic) had pushed a narrative that AI would replace millions of jobs and that unfettered development was necessary to achieve 'superintelligence.' Nadella, one of the architects of the AI boom thanks to Microsoft's investment in OpenAI — which exceeds $13 billion since 2019 — now warns that this narrative is unsustainable and dangerous. This reflects growing concern in the industry about the social and political consequences of AI, as well as a possible strategic shift toward more democratized and less concentrated AI. According to Goldman Sachs data (2024), AI could automate up to 25% of global work tasks, but Nadella questions whether that is a desirable goal, instead advocating for AI that augments human capabilities.

Historically, Microsoft has been a pillar of generative AI development. In 2023, it integrated GPT-4 into Bing and its Office 365 suite, driving mass adoption. However, the costs associated with frontier models, such as GPT-4 training which according to SemiAnalysis estimates (2024) exceeded $100 million, and energy consumption — a single training run can emit as many tons of CO2 as 100 transatlantic flights — have led Microsoft to rethink its strategy. Nadella's criticism also comes amid increasing regulation: the European Union passed the AI Act in 2024, and the United States issued an executive order on AI in 2023. By positioning itself as a responsible player, Microsoft seeks to influence these regulatory frameworks.

Consequences and context

Nadella's stance could have several consequences:

• Pressure on OpenAI and others: Microsoft is OpenAI's main investor, with an estimated 49% stake according to sources (Reuters, 2024). His criticism could translate into changes in governance or development priorities. In fact, Microsoft has already reduced its dependence on OpenAI: in March 2025, it launched its own language model Phi-3, which directly competes with GPT-3.5 in performance but at a computational cost 10 times lower.
• Cheaper and more accessible models: Microsoft has already launched models like Phi-3, which are smaller and more efficient, directly competing with OpenAI's massive models. Phi-3, with only 3.8 billion parameters, achieves results comparable to 7-billion-parameter models on reasoning tasks and can run on mobile devices. This contrasts with the trend of scaling models, such as GPT-4 (estimated 1.8 trillion parameters) or Google's Gemini Ultra.
• Regulation: Nadella's statements could influence regulatory debates, giving more arguments to those calling for controls on AI developments. For example, Senator Elizabeth Warren has cited Nadella's criticism in her proposed Algorithmic Accountability Act (April 2025).
• Public trust: By acknowledging that alarmist rhetoric can be counterproductive, Microsoft seeks to position itself as a responsible, user-focused player. Pew Research polls (2024) show that 52% of Americans feel more concern than excitement about AI, and Nadella's statements could help restore trust.

Historically, Nadella has been an advocate of AI as a tool to empower people, not replace them. This criticism aligns with his vision of 'AI as a copilot,' in contrast to the 'AI as a substitute' vision promoted by figures like OpenAI CEO Sam Altman, who in 2023 stated that AI could replace most jobs within a decade. The tension between these two visions has intensified since OpenAI attempted to restructure as a for-profit company in 2024, causing friction with Microsoft over control of the technology.

What should readers know?

Readers should understand that the AI market is in a maturation phase. The promise of ever-larger and more expensive models is being questioned by the reality of energy costs — training GPT-4 consumed approximately 50 GWh, equivalent to the annual consumption of 4,600 U.S. homes — and the need for practical applications. Microsoft's bet on cheaper, more controllable models could accelerate enterprise adoption of AI, but it also raises questions about the quality and safety of these smaller models. For instance, Phi-3 performs worse on creative writing and complex reasoning tasks compared to GPT-4, limiting its use in critical applications.

Moreover, Nadella's criticism reveals internal tensions in the AI ecosystem, where commercial and ethical interests often clash. Microsoft, which has generated significant revenue from Azure AI services — over $10 billion in 2024 — now seeks to diversify its offerings to avoid relying solely on OpenAI. This strategy recalls Microsoft's transition in the 1990s from proprietary software to the cloud, when Satya Nadella led the transformation toward Azure. Back then, Microsoft managed to surpass competitors like IBM by offering more open and flexible solutions.

For users and businesses, the signal is clear: AI is not a monolith. There will be options for different needs and budgets, but one must also be attentive to the quality and biases of each model. Nadella's statement is a reminder that technology should serve people, not the other way around. As he said in the interview: 'We cannot build a future where AI eats the economy; we have to build a future where AI serves the economy.'