Meta Just Launched Its Own AI Image Tool — and Microsoft Is Quietly Firing Its AI Vendors
The AI industry is doing something surprising this week: pulling back. Big companies are building more of their own tools, questioning who they actually need to buy from, and rethinking whether throwing money at outside AI providers still makes sense. Meanwhile, a new image generator just entered a very crowded race. Here’s what’s happening and why it matters to you.
Microsoft Is Replacing Outside AI With Its Own
Every time you ask Copilot to help you write a formula in Excel or draft a reply in Outlook, something is happening behind the scenes: Microsoft pays a fee to whoever’s AI model is doing the work. For years, that meant sending money to OpenAI and Anthropic. Now, according to TechCrunch, Microsoft is quietly swapping those outside models for its own internal ones, called MAI models, on a growing number of everyday tasks.
Think of it like a restaurant that used to order its bread from a bakery down the street. That works fine at first. But once the restaurant grows large enough, it becomes cheaper and smarter to bake the bread in-house. Microsoft has reached that scale. Running billions of AI-assisted requests through Excel and Outlook adds up fast, and using outside providers for each one is expensive.
For everyday users, the practical experience may not change much. The AI features in Microsoft 365 should still work the same way. What does change is who controls the technology — and that matters more over time. A company that owns its own models can update them faster, customize them more freely, and stop writing large checks to its AI partners.
Why this matters: Microsoft’s relationship with OpenAI has been one of the most closely watched partnerships in tech. Quietly replacing OpenAI’s models in its own products suggests that relationship is becoming less central than it once was.
“Microsoft shifts Excel and Outlook prompts to its own MAI models, cutting reliance on OpenAI and Anthropic.”
Meta Built Its Own Image Generator — and It’s Available Now
Meta has launched Muse Image, an AI image generation tool that lets you type a description and receive a picture. It comes from Meta’s Superintelligence Labs — a research division the company set up earlier this year to work on more ambitious AI projects — and it is now live inside Meta AI, the assistant accessible through Facebook, Instagram, and WhatsApp.
Image generation tools work by learning patterns from enormous collections of photographs and artwork. When you type “a golden retriever on a snowy mountain at sunset,” the model uses those learned patterns to assemble a new image that fits your description. Muse Image is Meta’s first major release from this research division, as AI at Meta’s blog explains, and it signals that Meta wants to be taken seriously in creative AI tools, not just conversational ones.
For the average person scrolling through Instagram or chatting in WhatsApp, this means AI-generated images are now one tap away inside apps you already use. You won’t need a separate subscription to a tool like Midjourney or Adobe Firefly to create a quick visual. Meta’s reach is enormous, which means this technology will land in front of people who never went looking for it.
Why this matters: Meta has over three billion users across its platforms. Embedding image generation directly into those apps means generative AI just got dramatically more accessible to people who aren’t already enthusiasts.
“Muse Image is the first image generation model from Meta Superintelligence Labs.”
A Former OpenAI Executive Joins a Rocket Startup
Kevin Weil, who served as Chief Product Officer at OpenAI before leaving earlier this year, has joined the board of Stoke Space, a startup building fully reusable rockets. TechCrunch reports that Weil brings experience from Twitter, Meta, and Planet Labs to the role. Stoke Space is trying to make getting to orbit dramatically cheaper by reusing every part of the rocket — including the upper stage, which most rockets discard after a single flight.
Board members at startups like this typically offer strategic guidance, help with fundraising, and open doors to valuable networks. Weil’s background in scaling fast-moving technology companies makes him a logical fit for a hardware startup with similarly ambitious goals. The crossover is notable: AI executives accumulating influence in aerospace is becoming a pattern, not an anomaly.
For most people, this is a story about where top tech talent flows when they leave AI companies. It suggests that ambitions around space infrastructure are attracting serious people, and that the overlap between AI and other deep-tech industries is growing.
Why this matters: When experienced executives from AI move into adjacent industries, they often bring both capital networks and a mindset shaped by rapid iteration — which can accelerate how quickly those industries move.
Also Happening in AI
The UK’s Financial Conduct Authority is warning that financial firms are racing to adopt AI faster than regulators can track, with officials flagging an “arms race” dynamic in the sector, per Ars Technica. On a more cautionary note, Wired profiled Verity Harding, a former DeepMind executive who argues that treating AI development as a geopolitical competition increases the risk of serious mistakes. On the enterprise side, Prime Intellect closed a $130 million Series A to help large companies build their own AI agents — software that can complete multi-step tasks autonomously — according to TechCrunch. And French startup ZML released free software called ZML/LLMD designed to make AI models run faster across a wider range of chips, potentially lowering costs for developers who don’t want to depend on Nvidia hardware exclusively.
What to Watch
The Microsoft story and the Prime Intellect funding round point in the same direction: companies are done being fully dependent on a handful of AI providers. Watch for more big tech firms announcing that they’re bringing AI capabilities in-house over the next quarter. The more interesting question is what this means for OpenAI and Anthropic, whose business models depend on enterprises paying to use their models. If their biggest customers become their biggest competitors, the economics of the AI industry shift significantly.