What do large language models need more than anything else?

Training data.

The Wikimedia Foundation announced landmark licensing agreements with top AI companies this past week. They signed deals with Amazon, Meta, Microsoft, Mistral, and Perplexity. These tech giants can now access Wikipedia's data through a new paid Wikimedia Enterprise platform.

No more sketchy web scraping.

These agreements replace the old method of scraping the open web. Now, companies get structured, machine-readable data delivered via specialized APIs. This covers over 65 million articles in more than 300 languages.

The enterprise platform offers real-time updates and snapshots. This helps models stay current with live edits while reducing the need to clean up raw and messy data.

Wikimedia aims to offset rising infrastructure costs caused by aggressive AI scrapers. Plus, this move avoids litigation by treating factual content as paid data rather than copyright battles.

What it means: Wikipedia was basically the original large language model.

Now they are formalizing that role.

It is lawsuits, partnerships, or bankruptcy for media companies. There is no third option. This deal secures Wikipedia's future while feeding the AI beast.

It is officially official.

In a joint statement from both companies, Apple announced a significant multi-year partnership with Google to power some of Apple's most important AI services. This includes a much more personalized Siri.

They will be using Google Gemini's models.

This move marks a massive shift in Apple's strategy. They tried to do it themselves for four years. They promoted Apple Intelligence. It didn't really work.

Now they are calling in the big guns.

Apple will rely on Gemini to enhance services, marking the first time Apple has used Google's foundational technology in its ecosystem. Financial terms were not disclosed, but past agreements have been valued in the billions.

Regulatory scrutiny is expected in the US, UK, and EU. But let's be honest. Apple isn't even a contender in the AI space right now. So this might not be viewed as a monopoly move.

What it means: Apple admitted defeat.

They couldn't build a competitive model in time.

If you ask Apple Intelligence something now, it often just kicks you to ChatGPT. Google Gemini is about to become the brain behind the iPhone.

OpenAI is trying to create a little compute diversity, and their recent deal signals taht. 

According to reports, OpenAI has struck a computing partnership with Cerebras Systems valued at over $10 billion. The multi-year deal runs through 2028.

It gives OpenAI access to Cerebras' specialized Wafer Scale Engine chips.

The goal? Reduce the company's reliance on NVIDIA's GPUs for large-scale AI training. Cerebras hardware will be delivered in phases to support OpenAI's push for a more resilient infrastructure.

Industry analysts see this as a direct challenge to NVIDIA's dominance in the AI chip market. This is the first time a major AI lab is betting heavily on non-NVIDIA alternatives at this scale.

What it means: OpenAI needs more compute than NVIDIA can provide.

They have models ready but can't serve them due to bottlenecks.

Diversifying their chip supply is the only way to scale quickly enough to meet their massive revenue goals.

Google is finally doing something we have wanted for a long time.

Google says that Gemini's new personal intelligence feature is rolling out as a beta version to AI Pro and Ultra subscribers in the US. It promises a more personalized AI experience by securely connecting with your Google apps.

Gemini can now link up with your Gmail, Google Photos, YouTube, and search history.

It creates tailored AI responses. You can get personal answers about travel plans, purchases, or recommendations based on your actual life.

Privacy is emphasized here. Connecting apps is optional. Google claims Gemini does not train directly on your personal data like your inbox or photos.

The feature works across web, Android, and iOS. Users can control which apps are connected or disconnected at any time.

What it means: This is the true promise of an AI assistant.

It isn't just a chatbot anymore. It is a tool that knows your schedule, your emails, and your photos.

If you can get past the creepy factor, the utility is unmatched.

Google is trying to completely change how we shop.

Google is expanding its Gemini AI chatbot with new shopping features, partnering with major retailers like Walmart, Shopify, and Wayfair. They are turning the app into a virtual merchant.

Here is the wild part.

You will be able to check out in AI mode without even visiting the websites. The new instant checkout function lets customers buy products directly inside the Gemini chat.

Walmart customers who link their accounts will receive personalized recommendations based on previous purchases. You can even combine new purchases with existing online carts.

Payments are processed using cards linked to Google accounts. They plan to add PayPal soon.

This is part of a broader race among Google, OpenAI, and Amazon. They all want to create seamless experiences that take users from browsing to buying in a single conversation.

What it means: The traditional internet is fading.

Why spend hours clicking through Wayfair when AI can do it for you?

Retailers are becoming backend warehouses while the AI interface becomes the new storefront.

This might change how medicine is made forever.

NVIDIA and Eli Lilly have announced a major $1 billion partnership to open a cutting-edge AI drug discovery lab in the Bay Area. They are aiming to speed up the development of new medicines with a five-year commitment.

The new lab opens by March.

It will house teams from both companies working side by side. They will use AI and high-performance computing to accelerate research.

Lilly's new AI factory is powered by NVIDIA's DGX SuperPOD and future Vera Rubin systems. It is designed to scale up drug research in a hybrid cloud environment.

The initiative centers on generating high-quality ground truth data through large-scale lab experiments. The goal is a closed-loop feedback system. Better experiments produce better data. Better data improves models.

What it means: Drug discovery is moving from test tubes to code.

This partnership industrializes the process.

Instead of random trial and error, they are building an engine that learns how to create medicine faster and more efficiently.

This might be the biggest story of the week.

Anthropic is making headlines with the launch of Claude Cowork. It is a new AI agent designed to handle everyday computer tasks for non-technical users. This move brings AI that works the way we do.

It is available right now.

Claude Cowork can autonomously manipulate, read, analyze, and create files on a user's computer. It is a hands-off assistant for organizing downloads, converting receipts, or drafting documents.

Unlike the developer-focused Claude Code, this is built for regular people. It simplifies complex tasks without requiring coding knowledge.

Anthropic's team said they built Cowork in just a week and a half. And they used Claude Code to build it. An AI built this tool in days.

It connects your local computer, your browser, and your cloud skills into one workflow.

What it means: This is the first real step toward the future of work.

It isn't just a chat box. It controls your actual computer.

When AI can control your mouse and keyboard, the definition of "productivity" changes completely.

Oh, YOU didn't think this was coming?

Awkward.

OpenAI confirmed that they will begin showing ads to adult users on the free version of ChatGPT. This starts in the US in the coming weeks. It is their first official move into digital advertising.

Paid subscribers on Plus, Pro, and Enterprise will not see ads.

But there is a catch.

The newly launched low-cost "ChatGPT Go" subscription will also include ads. Ads will appear at the bottom of answers and will be clearly labeled.

OpenAI promises never to sell user data to advertisers. Sensitive topics like politics and health are excluded.

Why are they doing this? OpenAI has nearly 900 million weekly active users. But only 5% of them are paid. That leaves hundreds of millions of people using the service for free.

Ads could help meet their ambitious financial goals as they target $20 billion in revenue.

What it means: If you aren't paying, you are the product.

OpenAI loses money on free users. Ads were inevitable.

This shifts the model from pure growth to sustainable monetization. Expect people to complain, but they will keep using it.