Nadella is not the first technology leader to articulate a view of what appears to be the next phase of the artificial intelligence era, one where the conversation is shifting from model capabilities to economic outcomes.
Microsoft’s peers in the AI industry, OpenAI and Anthropic, have recently published a blog post and an essay, respectively. While the writings focus on different themes, they share a common underlying question: What happens after AI becomes capable?
The answer increasingly revolves around productivity, economic growth and how societies adapt to a world where intelligence itself becomes abundant.
Why the conversation is changing
The first phase of AI was about proving that large language models could perform tasks once considered uniquely human.
The industry needed to demonstrate that these systems could write, code, reason, summarise information, analyse data and assist with complex workflows.
That phase is largely complete.
Today’s frontier models are capable enough to perform a broad range of knowledge work. As those capabilities become increasingly common, the focus naturally shifts elsewhere.
The bigger challenge now is turning AI capability into measurable economic outcomes.
That shift is visible across the industry’s biggest players.
Nadella’s vision: The era of abundant intelligence
Nadella’s article can be read as Microsoft’s blueprint for what comes next.
At the centre of his argument is the idea of abundant intelligence. Just as computing power and internet connectivity became widely available over time, he believes AI intelligence will eventually become abundant as well.
“What is at stake is not some digital tool or system and its use, but how organisations continue to learn, build IP, differentiate, and thrive in a world where AI models can continuously absorb the expertise of humans and organisations and commoditise it,” Nadella wrote.
Put simply, if every company has access to powerful AI models, the differentiator will no longer be the model itself.
It will be the unique knowledge, workflows, expertise and business processes companies build on top of those models.
To explain this, Nadella introduces two concepts: human capital and token capital.
Human capital consists of expertise, judgement, relationships, creativity, institutional memory and domain knowledge.
Token capital refers to the AI systems, agents, workflows and digital capabilities organisations build and own.
Importantly, Nadella does not see these as competing forces.
Instead, he argues they form what he describes as a learning loop. Human expertise improves AI systems. Better AI systems make people more effective. Those improved outcomes generate new knowledge that further strengthens both human and AI capabilities.
Over time, that loop becomes a company’s intellectual property.
AI should augment, not replace
Perhaps the most notable aspect of Nadella’s vision is what it does not predict.
There are no warnings of mass unemployment. No forecasts of fully autonomous companies. No suggestion that human workers become irrelevant.
Instead, Nadella repeatedly argues that human capital becomes more valuable as AI improves.
“Human capital does not become less valuable as token capital grows. It only becomes more valuable! I believe human agency will be the driver of token capital growth,” he wrote.
In this view, humans continue to set goals, identify opportunities, exercise judgement, build relationships and make decisions. AI simply amplifies those capabilities and scales them across an organisation.
This is Microsoft’s preferred vision of the future: AI agents working alongside employees rather than replacing them.
OpenAI sees a similar path
A remarkably similar theme appears in OpenAI’s recent writing.
The company explicitly argues against a future where everything is automated. Instead, OpenAI says AI should help people pursue their goals rather than replace human judgement.
As AI systems become more capable, OpenAI believes humans will continue to play a critical role in setting direction, making trade-offs, applying values and deciding what is worth doing.
In other words, both Microsoft and OpenAI are presenting AI primarily as an amplifier of human capability rather than a substitute for it.
For both companies, the future is not one of autonomous organisations operating without people. It is one where AI becomes deeply embedded into existing workflows and helps workers become more productive.
Anthropic sees a different possibility
Anthropic Chief Executive Officer Dario Amodei takes a more cautious view.
In a recent essay, he asks a different question: What if augmentation eventually becomes substitution?
What if AI stops complementing human work and starts replacing it?
Historically, technological revolutions displaced some jobs while creating others. Amodei argues that AI could be different because it directly targets cognitive work traditionally performed by humans.
If the transition happens quickly enough, labour markets may struggle to adapt.
Anthropic therefore argues that governments should prepare for the possibility that labour-market disruption could be larger and faster than previous technological transitions.
Preparing for disruption
Anthropic’s proposals focus on managing potential labour-market shocks before they occur.
Among the ideas discussed are:
-
Wage insurance for workers moving into lower-paying jobs -
Incentives for businesses to retain employees -
Expanded labour-market monitoring -
Workforce retraining programmes
If job displacement becomes widespread and long-lasting, Anthropic argues governments may eventually need to consider broader interventions such as universal basic income or other forms of direct income support funded by AI-driven economic growth.
The underlying argument is straightforward.
If AI dramatically expands economic output, societies will need mechanisms to ensure that prosperity reaches people whose jobs are affected by the transition.
Economic growth becomes the new scoreboard
Despite their differences on labour disruption, Microsoft, OpenAI and Anthropic ultimately converge on the same central theme: economic growth.
This is where the conversation has evolved most significantly.
For years, the technology industry measured progress through technical breakthroughs. The AI industry initially followed a similar path, focusing on benchmark scores, reasoning ability and model capabilities.
Now the focus is shifting towards productivity.
Nadella repeatedly argues that AI’s success should be judged by whether it raises productivity and creates value across industries.
OpenAI compares AI to electricity, a technology that transformed economies because it enabled entirely new forms of productivity and innovation.
Anthropic similarly believes AI could dramatically accelerate scientific discovery, research and economic growth.
The common message is that capability alone is no longer enough.
The next challenge is turning capability into broad economic value.
Who captures the value?
This is where another shared concern emerges.
Microsoft, OpenAI and Anthropic all raise questions about concentration versus distribution.
Nadella’s warning echoes concerns that emerged during the first wave of globalisation. While globalisation generated enormous economic growth, many communities felt left behind because the benefits were not distributed evenly.
He worries AI could produce a similar outcome.
If a small number of companies capture most of the value generated by AI while workers and businesses see their expertise commoditised, public support for the technology could weaken.
This is where Nadella’s concept of societal permission becomes important.
Society will continue supporting AI only if people see tangible benefits in their everyday lives through better healthcare, improved education, stronger public services, scientific breakthroughs and higher living standards.
OpenAI arrives at a similar conclusion through a different route.
The company argues that power should be distributed broadly rather than concentrated among a handful of firms. Its vision positions AI as infrastructure that creates value across society rather than concentrating it in a few hands.
Anthropic’s concerns about labour displacement ultimately point in the same direction. If AI generates enormous wealth but that wealth remains concentrated, governments may need new mechanisms to distribute some of those gains more broadly.
The real question, therefore, may not be whether AI creates wealth.
It may be who benefits from it.
From a technology debate to an economic debate
Viewed together, the writings from Microsoft, OpenAI and Anthropic reveal an important shift within the industry.
Microsoft is focused on embedding AI into organisations and creating learning loops between humans and machines.
OpenAI is positioning AI as a foundational layer of the economy capable of generating widespread prosperity.
Anthropic agrees with the economic potential but warns that growth alone may not guarantee a smooth transition for workers.
The differences matter.
But so does the common thread.
All three companies are increasingly discussing productivity, economic growth, labour markets, value distribution and public trust rather than benchmark scores and model rankings.
The model race is not over. But the conversation has clearly evolved. For nearly two years, the defining question in AI was how powerful the technology could become. The emerging question is arguably more important: How will the economy benefit when it does?
