AI News – July 26, 2026: Who Controls AI? The EU Challenges Google, Oracle’s Cuts, and the Infrastructure Race

Artificial Intelligence is leaving the experimental phase and showing its harshest face: that of infrastructural control. In the weekly edition of AI News (July

If the past months were dominated by enthusiasm for the new "wonders" of Generative Artificial Intelligence, the week of July 20-26, 2026 marks a phase of harsh and pragmatic maturation. The debate has radically shifted from "who has the best model" to "who controls access to the technology."

Between antitrust issues, human costs, and titanic infrastructure investments, the news of the last seven days confirms that platforms, regulation, and data centers have become the true battlefield. Here are the 5 fundamental stories of the week.

1. The EU vs. AI Lock-in on Android

The most disruptive news of the week comes from Europe and touches the nerve center of technological distribution: mobile operating systems.

🔍 What happened: The European Union has initiated formal action to force Google to open the Android ecosystem to rival AI models and assistants. The goal is to prevent the so-called lock-in, i.e., the practice of binding users exclusively to the proprietary Artificial Intelligence (Gemini) pre-installed on devices.

💡 Why it matters: This regulatory move could reshape the entire mobile economy. The EU is drawing a clear line: whoever owns the operating system cannot monopolize access to AI. It is a crucial battle for free competition, which will allow startups and third-party companies to compete for the home screen of billions of smartphones.

2. The Cost of Infrastructure: Oracle Cuts 30,000 Jobs

The race for computing power comes with a huge industrial and human price, and companies are willing to do anything to finance it.

🔍 What happened: Oracle announced a drastic restructuring plan involving the elimination of 30,000 jobs globally. This massive workforce reduction is not driven by a crisis, but by the need to free up enormous capital to finance its push into data centers and, in particular, the colossal Stargate project (the supercomputer network dedicated to training future AI models).

💡 Why it matters: This event lays bare the ruthless economics of Artificial Intelligence. Software is literally devouring capital to build hardware. Legacy companies are dismantling entire traditional divisions to concentrate every resource on energy procurement and server farms.

3. DeepSeek's Squeeze and the Race for Portability

In the language model market, the dynamics of infrastructure dependency are intensifying, pushing users to seek ways out.

🔍 What happened: The week saw growing pressure around DeepSeek, with impending deadlines on migration terms for developers. This squeeze forces companies relying on its APIs to quickly revise their technology stacks or accept new, stricter usage policies.

💡 Why it matters: The DeepSeek case raises the critical issue of model portability. Developing a product based on a single provider exposes a company to mortal risks if that provider changes the rules or costs. The market is desperately seeking agnostic architectures that allow jumping from one LLM to another without rewriting all the code.

4. The EU Accelerates: Institutional Hub and 400 Million in Funds

Public institutions are no longer just legislating; they are entering the field as financiers and orchestrators of innovation.

🔍 What happened: In parallel with the antitrust action on Android, the European Parliament announced the creation of a new centralized hub for Artificial Intelligence governance. To support this initiative, a European public fund of 400 million euros was allocated to support sovereign AI startups and projects (i.e., developed entirely within EU borders).

💡 Why it matters: Europe is trying to close the gap with the United States and China. Establishing massive public funds means recognizing that AI is not just a commercial market, but a strategic national security infrastructure over which governments must maintain independent control and development.

5. The Normalization of the "AI Infrastructure First" Model

The industry narrative has changed forever: the magic of the prompt gives way to the hard law of silicon.

🔍 What happened: End-of-July market analyses converge on one irrefutable reality: the consolidation of the AI infrastructure first paradigm. Analysts point out that the true winners of the current market are not the companies building the final applications, but the "pick and shovel" suppliers: those who make chips, those who manage cooling systems, those who provide electricity, and those who own the data centers.

💡 Why it matters: We are witnessing an industrial maturation. Just as the internet era was won by those who laid submarine cables and built the cloud, the AI era will be dominated by those who control the compute. Access to hardware has officially become the bottleneck and the most valuable currency of the decade.

Conclusions: Infrastructure is Destiny

The week of July 20-26, 2026 definitively closes the curtain on romantic and experimental Artificial Intelligence. The news tells us of a sector that has become a heavy industry. Whether it is the EU's legal battles to dismantle monopolies on mobile devices or Oracle's painful cuts to finance servers, the message is one: the geopolitics and economy of the future will be written by those who hold material control over Artificial Intelligence. Software eats the world, but infrastructure eats software.