AI economy and monopolies

The AI economy and monopolies are colliding in real time: from Judge Mehta’s decision on Google and data sharing to the consumer vs. enterprise split and the energy bottleneck of datacenters. The question isn’t who has the data, but who uses it better.

The King’s Wild West-or His Return?

Five years of legal battle later, the U.S. Department of Justice defeats Google over monopolistic practices. Judge Amit Mehta does not order a breakup. He does, however, order something more painful: the sharing of secrets.

What Google won

It keeps Android and Chrome.
It continues to pay $20B per year to Apple for default search.
Alphabet’s stock rose 7.8%.

What it loses

Exclusivity is over.
It must give competitors a “snapshot” of its search data-decades of indexing as an instant download.
For five years, search results will be open under usage licenses.

Judge Amit Mehta wrote it plainly: “It is not the job of judges to see into the future.” And he acknowledged the obvious: generative AI changed everything.

Why AI is the real winner

In 2020, ChatGPT didn’t even exist. Today, 1 in 4 users search through AI chatbots. OpenAI, Perplexity, even DuckDuckGo suddenly see access to the treasure: search data that until yesterday was locked away.

In the trial, OpenAI admitted that building its own full index would take years and hundreds of millions. Google’s refusal back then was predictable: “you’re competitors.” Now, data sharing opens an era where AI agents feed from the same “diet” as the king. The question shifts: not who has the data, but who activates it better.

Two markets inside the AI economy

That same week, OpenAI and Anthropic publish major usage studies. The comparison is skewed from the start-OpenAI does not measure Team/Enterprise, while Anthropic mostly measures Team/Enterprise-but the patterns are revealing.

The Consumer AI Economy (ChatGPT-led)

73% non-work usage.
49% of interactions are “asking” for guidance and creativity.
Users see it as a personal advisor.

The Enterprise AI Economy (Claude-led)

77% API usage signals automation.
36–39% focuses on coding.
The “directive” style is rising rapidly: full delegation of tasks.

The big “coding divide”: ChatGPT ~4.2% coding among a casual audience vs. Claude ~36–39% among a professional audience. Asymmetric data? Yes. But the trend is clear: Claude is winning the worksite; ChatGPT is winning everyday life.

The geography of inequality

Small, hyper-digital countries (Israel, Singapore) show usage multiples above what you’d expect. Large emerging markets (India, Nigeria) lag per capita, but display automation-dominant patterns. India has coding usage >50% versus ~36% global, suggesting specialization in AI-powered development services.

Put simply: where AI augments, it increases value; where it only automates, it compresses price. This is the new digital geoeconomic divide.

Energy as a lever of monopoly: the bottleneck of the AI economy

NVIDIA is committing up to $100B for 10 GW of AI datacenters for OpenAI. Put simply: power for ~7.5 million homes, dedicated to training/running models.
In parallel:

OpenAI–Oracle deal of ~$300B for ~4.5 GW.
Microsoft: a new mega-datacenter in Wisconsin.
Meta: 5 GW “superclusters” and investments in gas plants.
Energy puzzle: the bottleneck isn’t chips-it’s electricity.

The irony? AI demands colossal energy, yet helps accelerate energy (e.g., fusion optimization). Until then, the mix relies on natural gas and grid-scale batteries.