Billions for OpenAI, patience with Google & automated hacks
Monday, December 22, 2025
Today is marked by massive financial movements and strategic delays. While billions are being mobilized behind the scenes for the next phase of AI development, users will need to exercise a little more patience than expected in their daily lives. At the same time, security researchers are warning of a new era of cybercrime, which could become fully automated by 2026.
Here are the most important AI news stories of the day:
SoftBank plans a $22.5 billion bet on OpenAI
Reports indicate that Japanese tech investor SoftBank is on the verge of a historic deal: a $22.5 billion funding round for OpenAI is expected to be finalized by the end of the year. SoftBank CEO Masayoshi Son apparently plans to finance this sum, in part, through loans secured by shares in chip designer Arm. This move underscores SoftBank’s aggressive strategy to position itself at the heart of “superintelligence” development.
Google: Gemini transition is taking longer
Google has confirmed that the complete replacement of the classic “Google Assistant” with the new Gemini AI model on smart home devices will take longer than originally planned. While smartphones are already being migrated, hardware integrations in smart speakers and displays are proving more complex. The original timeline of March 2026 is now uncertain, suggesting that Google is prioritizing quality and stability over speed.
Additional info: At the same time, the global rollout of Gemini 3 Flash in Google Search has been completed, meaning faster and more logical answers for users worldwide.
Anthropic is building massive infrastructure with Hut 8
To keep pace with the computing demands of its Claude models, Anthropic has announced an infrastructure partnership with mining and data center operator Hut 8. The goal is to build massive computing capacity (up to 2 gigawatts of power), which should be available from 2027 onwards.
This is a clear indication that major AI labs are already cementing their hardware foundation for the next generation of models.
Warning about “Industrialized Cybercrime”
Two new reports from Trend Micro and the Boston Consulting Group (BCG) paint a bleak picture for 2026. Trend Micro predicts that cybercrime will be fully “industrialized” and automated by AI agents in the coming year, allowing attacks to scale without human intervention.
The BCG study adds that cyber risks in companies are currently growing faster than the implemented protective measures.
OpenAI publishes GPT-5.2 codex
There was also news for developers: OpenAI has released the specialized GPT-5.2 Codex model. It specifically targets complex programming tasks, refactoring legacy code, and improved cybersecurity. The model is being rolled out gradually to identify potential abuses in software development at an early stage.
Beliebte Beiträge
Why is my Excel File so big?
Solve the problem with Excel spreadsheets that take up several megabytes of space without any apparent reason.
The most important tips for Excel
These 10 tips for Microsoft Excel 2016/2019 with questions that many users ask every day are guaranteed to help you and save a lot of time.
More fonts in Word through Google Fonts
You can find out how to expand the font selection in Word with just a few steps and above all free in our article.
Move pages and insert new ones in word
You probably also know the situation that you are sitting on a very long document, and in the middle of it falls on the actually between page 2 and page 3 still a page in between must.
The Sreference and the Wreference in Excel
Microsoft Excel: How to get S-reference and W-reference purposefully data out of a source.
Calculate with date, time and working days in Excel
Calculating correctly with date, time and working days in Excel always causes uncertainty. We will explain step by step how it works.
Offers 2024: Word & Excel Templates
