Europe Opens Android as Google’s Gemini Timeline Slips

Europe forced open two Google advantages

The European Commission issued two binding measures under the Digital Markets Act aimed at reducing Google’s structural advantages in mobile AI and search.

The first requires Google to give competing AI assistants access to Android features on terms comparable to Gemini. That includes capabilities such as voice activation, acting inside apps, suggesting replies and using device context. The Commission says the decision covers 11 Android features relevant to AI services.

The second measure addresses Google Search data. Eligible third-party search engines—and AI chatbots with search functionality—must be able to obtain anonymized query, ranking, click and view data under fair terms. The Commission also specified privacy safeguards, pricing principles and a process for evaluating cybersecurity risks before data is shared.

The near-term test is implementation. Access on paper will matter only if rival assistants can activate reliably, complete actions across apps and obtain search data at a price and cadence that supports competitive products.

Google’s flagship model reportedly missed its window

The regulatory decision arrived on the same day that Google faced a separate execution problem. Reuters reported, citing Bloomberg, that Gemini 3.5 Pro was months behind schedule while Google worked to improve the model, particularly its coding performance.

Sundar Pichai had said at Google I/O that the model was expected in June. Google told Reuters that it was testing 3.5 Pro, an upgraded Flash model and other systems with partners. Alphabet shares fell nearly 3% after the report.

A missed model date is not proof that Google has fallen out of the frontier race. The company still controls a deep research organization, custom silicon, cloud distribution, Android and a large consumer product surface. But those advantages also raise expectations: a model delay becomes more visible when competitors are shipping quickly and customers are making platform decisions now.

TSMC paired record earnings with another capacity bet

The chip cycle delivered a much more concrete demand signal. TSMC reported second-quarter revenue of $40.2 billion, a 67.7% gross margin and a 60.3% operating margin. Separate coverage said net profit rose 77% year over year to a record as demand for advanced AI processors remained strong.

TSMC also pledged a further $100 billion of U.S. investment, bringing its planned American commitment to $265 billion. The company raised its 2026 capital-spending range as it prepared more advanced manufacturing and packaging capacity.

Those numbers matter because the foundry sits beneath nearly every major AI platform. Model companies can change architectures, cloud providers can design custom accelerators and chip vendors can compete on systems, but leading-edge manufacturing and advanced packaging remain difficult bottlenecks to replace.

Japan turned Rubin capacity into industrial policy

Nvidia and Noetra announced a 140-megawatt Vera Rubin AI factory in Japan, supported by Japan’s Ministry of Economy, Trade and Industry.

The planned system includes 13,750 Vera CPUs and 27,500 Rubin GPUs. It will support Japan’s FRONTia project and the development of open multimodal foundation models for robotics, digital twins, manufacturing, logistics, healthcare and other physical-AI applications.

The important detail is not simply the GPU count. Japan is tying national compute capacity to domestic industrial data, robotics expertise and broadly available model weights. That is a different strategy from treating AI infrastructure as generic cloud capacity purchased one workload at a time.

Netflix put generative AI inside ordinary production budgets

Netflix disclosed in its second-quarter shareholder letter that generative-AI workflows had been used in roughly 300 titles during 2026, with most of the work concentrated in post-production.

The company cited _Glory_, _Brasil 70: A Saga do Tri_ and _The American Experiment_ as examples. The tools were used for enhanced crowds, historical battle sequences and establishing shots. Netflix framed the benefit in operational terms: higher-quality output delivered faster and at lower cost, including shots that some productions otherwise could not afford.

This is a more useful adoption signal than a vague announcement about an “AI strategy.” It identifies where the technology is entering the workflow: specific, expensive visual tasks with measurable time and budget constraints.

Linux rejected a blanket ban on AI-assisted work

The day’s developer-culture story came from Linus Torvalds. During a kernel mailing-list dispute, he said Linux was not an anti-AI project and told opponents of AI-assisted work that they could fork the project or walk away.

That does not mean the Linux kernel is lowering its standards or accepting generated patches without scrutiny. The kernel’s review process still depends on technical correctness, maintainability, provenance and human accountability. Torvalds’ intervention instead rejects a categorical rule that code should be excluded merely because AI tools contributed to its creation.

Thursday’s clearest contrast sat inside Google: Europe moved to weaken the company’s distribution advantages just as its next flagship model reportedly needed more time. The immediate watchpoints are narrow: whether rival assistants gain meaningful Android capabilities, whether shared search data is commercially usable, and whether Gemini 3.5 Pro ships with the coding performance Google is seeking.