🦾How You Can Leverage:

Oh, YOU didn't restructure your entire company to lay the groundwork for a $1 trillion IPO this week?

Thought we all did?

OpenAI did. Microsoft somehow won by losing equity. Amazon axed 14,000 due to AI (and there’s more on the chopping block). Google hit 650 million users.

And that’s just the start.

This week wasn't about flashy model releases. Nope. This was about partnerships, billion dollar deals, and corporate chess moves that'll shape the AI industry for years.

What'd you miss?

Here's what matters for AI updates this week.

OpenAI announced this past week that it completed its major recapitalization, reshaping governance and technically deepening its Microsoft partnership.

The new structure? The OpenAI Foundation now holds a $130 billion equity stake with 26% control of the for-profit arm called OpenAI Group PBC. Employees and investors hold 47%.

Microsoft's stake dropped from 49% to 27%.

But here's where it gets interesting. That 27% stake is now worth $135 billion. Microsoft's original investment? Around $13 billion.

Yuuuuup. They 10x'd their money.

Plus, OpenAI just committed to buying $250 billion in Azure cloud services. Microsoft lost its exclusive compute provider rights, but gained a quarter trillion dollars in guaranteed future revenue.

Not bad, right?

With the restructure, OpenAI also secured clearance for its recent ongoing funding round and cleared the path for what's coming next. (More on that later)

What it means: 

Microsoft won this deal despite any headlines saying otherwise. (Well, it’s not like OpenAI lost TBH)

Microsoft multiplied their investment by 10x and locked in $250 billion in cloud revenue.

Sometimes losing exclusivity is worth it when the checks are this big.

For OpenAI? The whole nonprofit to for profit story may be over on paper, but expect the (not so serious?) lawsuits from Elon Musk to continue.

How many jobs can one company cut before calling it a restructure instead of a bloodbath?

Amazon has already confirmed 14,000 corporate job cuts tied directly to AI, marking its largest AI driven reduction to date. That number is widely expected to hit 30,000 by next year.

These aren't pandemic overcorrection layoffs. Amazon already handled those. The company explicitly said this is about AI replacing mid level and admin roles across HR, devices, advertising, and Prime Video.

The HR department alone is getting slashed by 15% globally.

Let's be honest. This isn't a surprise. Last year, Amazon announced its internal AI tool Amazon Q saved more than 4,500 developer years annually in just one use case.

According to The New York Times, Amazon also plans to hire 600,000 fewer warehouse workers over the next eight years as it leans into AI robotics.

The company wants to operate like the world's largest startup.

If you missed our episode on this, make sure to go check it.

What it means: 

Amazon is shifting from paying people to paying for infrastructure.

The OPEX to CAPEX pivot is real, and 30,000 corporate employees are paying the price.

When a company saves thousands of developer years with one AI tool, the writing was on the wall.

According to reports from CNBC and Bloomberg, Apple is actively considering even more AI app providers for system level integration into its operating systems.

Why?

Because Apple Intelligence doesn't work.

CEO Tim Cook told CNBC that Apple's intention is to integrate with more people over time. The company has already embedded ChatGPT into Siri and Google Gemini support is reportedly in progress.

Now? Potential integrations with Anthropic's Claude and Perplexity are on the table.

Cook said Apple is making good progress on an AI upgraded Siri targeted for next year. That's the same thing Apple said last year about this year.

The only real Apple AI news lately? Lawsuits, delays, or announcements about needing more third party help.

What it means: 

Apple's AI strategy is outsourcing everything because it can't build competitive models internally.

ChatGPT does the heavy lifting. Gemini handles search. Claude might join soon.

Apple Intelligence might just turn into a router to everyone else's ACTUAL intelligence, wrapped up in Apple Marketing and ‘private cloud’ technicalities.

At its Canva World Tour conference this past week, Canva announced a new foundational AI design model that generates editable layered designs instead of flat images.

This works across posts, presentations, whiteboards, and websites.

The model understands design layers and formats. Users start with a prompt and then directly tweak objects without the usual prompt to perfection friction that drives designers insane.

Canva AI, the platform's assistant, is now available throughout the entire interface. You can use the at mention command and talk to Canva AI whether you're working in standard design, video, or anything else.

The assistant also gained new creation features including generating 3D objects and copying the art style of any design.

That last part could speed up brand consistent content production for teams who need to maintain visual identity across dozens of assets.

What it means: 

Canva didn't license someone else's model. They built their own.

The focus on editable layers instead of flat images solves the biggest pain point with AI design tools.

You can actually use what it generates instead of starting over from scratch.

Want Claude directly inside Microsoft Excel with real time market data?

Anthropic is rolling out Claude AI tools that integrate directly with Excel, letting financial analysts interact with the AI assistant inside their spreadsheets. This is one of Anthropic's biggest moves into financial services.

But here's the catch though.

The rollout is absurdly limited. Only 1,000 users in the first round. You've gotta be on a Max Team or Enterprise plan just to join the wait list.

The new Claude sidebar inside Excel can read, analyze, and modify Excel workbooks. It explains every change at the cell level and helps build financial models.

The company is also introducing six new agent skills with prebuilt workflows for tasks like discounted cash flow models, comparable company analysis, and processing data room documents.

Beyond Excel, select Claude models will be available in Microsoft Copilot Studio and Researcher Agent as well.

What it means: 

Microsoft's Copilot struggled for 18 months but has been crushing it the last three months.

Part of that success? Partnering outside OpenAI for the first time.

When you're paying other companies instead of just paying yourself, that's a signal something is working.

Google announced some new updates and a complete redesign of its Vibe Coding products inside Google AI Studio this past week.

The new Vibe Coding workflow turns a single sentence into a working AI powered app in minutes.

Users can describe multimodal apps in plain language like a voice powered video generator or an Imagen 3 powered image editor. The system uses Gemini models to handle core code wiring, UI scaffolding, and core logic.

There's also a revamped app gallery that provides a visual library of Gemini powered examples. You can browse, preview, or remix them just with natural language.

The new annotation mode inside AI Studio lets users click on interface elements and request changes in natural language. You can click and mark up designs with multiple notes and just say fix this, move this left, or let users upload a photo here.

Welp, most of this functionality already existed. It was just buried and hard to find.

What it means: 

Google isn't adding brand new capabilities. They're making existing features easier to discover.

The real unlock is annotation mode where you markup designs and give instructions instead of rewriting prompts.

Vibe coding just got way more accessible for people who think in clicks instead of code.

Microsoft just announced a new app builder and workflow agent for Microsoft 365 Copilot customers in the Frontier program.

Employees can now build apps, automate workflows, and create agents using natural language inside Microsoft Copilot.

The app builder creates working apps in minutes without database setup. It grounds content in Microsoft 365 files and stores new data in Microsoft Lists with share by link distribution.

The workflows feature turns plain language instructions into automated steps across Teams, SharePoint, Outlook, and Planner. It shows each step in real time and allows edits in the same conversation.

Copilot Studio inside Copilot now builds personalized work grounded agents that tap SharePoint, meeting transcripts, team chats, emails, and external systems like ServiceNow and Jira.

In a launch scenario, Copilot could spin up an app to track campaign milestones, send Monday Teams updates with Planner deadlines, and deploy an agent to answer what's next using only company resources.

Microsoft has been shipping major updates almost every week lately. Not just twice a year anymore.

What it means: 

Microsoft is turning Copilot into a low code platform where employees build their own tools.

No IT tickets. No waiting on developers. Just describe what you need and it builds it.

The work grounded part matters because it only uses your company's actual data instead of hallucinating answers.

During its quarter three earnings this past week, Google announced that the standalone Gemini app now has over 650 million monthly active users.

That's absolute nutty growth. A year ago? Only 90 million users. In March? 350 million users.

They almost doubled their user base between March and September.

OpenAI still leads with 800 million weekly active users. Different metric, but Google is closing the gap fast.

CEO Sundar Pichai credited Imagen 3, the image generation and editing model, for driving 23 million new Gemini app users in September alone. The company also announced Gemini 3 is coming later this year.

Google already exceeded the CEO's goal of 500 million users by 2025.

Google's AI Mode reportedly has over 100 million monthly users and is now seeing 75 million daily users. That feature is so good it'll eventually become the default search mode.

What it means: 

Google went from AI search disaster with Bard to legitimate ChatGPT competitor in under two years.

Imagen 3 is a user acquisition machine. Gemini 2.5 Pro is the best model by many benchmarks.

They could realistically catch OpenAI in total users by end of year.

Ready for some Dr. Evil numbers?

OpenAI is reportedly preparing for an initial public offering that could target a $1 trillion market valuation, making it potentially the largest IPO in US history.

According to reports from Reuters and Bloomberg, the IPO could happen as early as 2026. The company recently achieved a $500 billion valuation in October through a secondary share sale.

If successful, this IPO could raise $60 billion in outside capital. That would triple the previous US IPO record set by Alibaba at $21.8 billion.

The restructuring we covered earlier? That laid the groundwork for OpenAI to go public.

In an interview late last week, Sam Altman acknowledged that an IPO was a very real option because of the capital intensive work the company is doing.

What it means: 

We think OpenAI will become a Top 3 company in the world within three years. We called the whole ‘NVIDIA will be the world’s biggest company’ thing in 2023 before most had heard of them.

(Bookmark this one.)

OpenAI’s recapitalization cleared legal obstacles. The valuation keeps climbing. The infrastructure demands require massive capital.

An IPO isn't just likely anymore. It's necessary.