29 October 2026 · The morning thread

Security is becoming central to the AI race, while cloud growth shows the strength of demand for infrastructure. Google introduces Gemini 4 Carbon and shares new model-usage figures; Microsoft reports quarterly results; Anthropic expands support for research. Four official announcements from 28 October, selected for this morning’s briefing.

Google · Carbon hunts for flaws, but access remains selective

Google has introduced Gemini 4 Carbon, a model focused in part on cyberdefence and long-running engineering tasks. The company reports improvements in vulnerability detection and patch generation. Rollout begins with selected cyber defenders through the Fairwind programme, ahead of broader access. This is not a general public release. Benchmark results do not guarantee that every flaw will be found or fixed in a real system.

Google · announcement of 28 October ↗

Alphabet · Gemini reaches one billion users

In its third-quarter earnings remarks, Sundar Pichai says the Gemini app has reached one billion users. First-party model APIs process nearly 32 billion tokens per minute, more than four times the level a year earlier. Google Cloud reports year-on-year revenue growth of 103%. These are company-reported figures on usage and business activity, not an independent assessment of answer reliability.

Alphabet · third-quarter results, 28 October ↗

Microsoft · Azure grows 48%, cloud revenue reaches $63.9 billion

Microsoft ended the first quarter of fiscal 2027, which closed on 30 September 2026, with total revenue of $92.3 billion, up 19% year on year. Microsoft Cloud revenue reached $63.9 billion, up 30%, while Azure grew 48%. For AI watchers, these figures indicate the scale of supporting infrastructure. These business lines include many activities, however, and should not be described as revenue generated solely by artificial intelligence.

Microsoft · quarterly results published on 28 October ↗

Anthropic · Claude credits and partnerships for science

Anthropic is expanding AI for Science through partnerships with 20 foundations, institutions and research groups. The programme offers free Claude credits for high-impact scientific projects. Examples include data analysis and tools connecting AI agents to laboratory equipment. The Model Hardware Standard, developed with HHMI Janelia Research Campus, remains in research preview. Enabling experiments and widening access to tools matters, but does not by itself demonstrate new discoveries or clinical benefits.

Anthropic · AI for Science, announcement of 28 October ↗

The desk’s view · Growth needs trust

Our reading is that competition increasingly operates on two fronts: capability and control. Cloud figures show growing demand; security and research raise questions about the quality of outcomes. For businesses and professionals, the practical test remains the same: distinguish available tools from announced plans, verify outputs and retain human responsibility for decisions.

While you were sleeping · all editions ↗

Sources and documents

Document date or reference period: Original announcements of 28 October 2026. Sources confirm publication dates, but not all provide a time falling within the Italian overnight window..

Desk selection and summaries with AI assistance, based on the linked original sources. Figures, capabilities and programmes are attributed to the companies. The concluding perspective is desk commentary.

Report an update or correction →