Two longtime technology rivals are working together on something unexpected: making advanced artificial intelligence run outside massive cloud data centers.
What happened?
On October 15, Meta introduced Muse Spark Open, a powerful AI model designed for coding, reasoning and complex multi-step tasks. Meta
The model contains 816 billion parameters, although only 42 billion are active at a time to process information. Meta · VentureBeat
But the most interesting part wasn’t its size.
Meta worked with Apple to make the model run locally on two of Apple’s M5 Ultra Mac Studio computers, using Apple’s MLX software. Instead of relying entirely on remote servers, developers can run this advanced AI system on hardware they control. Meta · Apple Machine Learning Research · MacRumors
That’s an important milestone for local AI computing.
And there’s an unexpected twist: Meta and Apple have spent years competing over privacy, advertising and technology platforms. Now they’re collaborating to make powerful AI more accessible.
Why it matters
Today, many advanced AI systems depend on enormous data centers filled with expensive processors. That creates significant demand for electricity, cooling, networking and specialized infrastructure.
Running AI locally offers a different approach. Companies could potentially process sensitive information without continuously sending it to outside cloud services, and developers could gain more control over how their AI systems operate and are customized.
However, there’s an important limitation.
Meta’s setup still requires expensive professional hardware. Each Mac Studio needs 512GB of memory, a configuration Apple hasn’t released yet; it’s expected later in October, and its price hasn’t been announced. The 256GB version already costs more than $10,000. This isn’t something an ordinary laptop can run today. MacRumors · VentureBeat
And while Meta calls Muse Spark Open the most capable U.S. open-weight model to date, models from Chinese developers such as Kimi K3 and GLM-5.3 outperform it on several benchmarks. VentureBeat
The significance isn’t that Meta has won the AI race. It’s that another powerful AI model is becoming available outside a closed, cloud-only environment.
The bigger picture: who controls AI?
Here’s where things become particularly interesting.
For years, the AI industry has been building increasingly powerful systems inside centralized data centers. But what happens if more of that intelligence can run directly on company-owned hardware?
Businesses could have more choices about where their data goes, which providers they depend on, and how much control they keep over their AI tools.
There’s another important detail. Although Meta is releasing the model’s weights, its license includes commercial limits: companies with more than $20 million in annual revenue that offer the model as a service must sign a separate agreement with Meta. Internal company use is exempt. VentureBeat
So even as AI becomes more accessible, the battle over its distribution and commercial use continues.
The next AI competition may be about who controls the infrastructure, not simply who builds the smartest model.
The investment connection
This development touches several industries.
Apple could benefit if powerful local AI creates more demand for high-memory computers and specialized processors. Meta could strengthen its position with developers by encouraging companies to build around its models.
Meanwhile, cloud providers such as Amazon, Microsoft and Google may face a more complicated future. Some workloads could eventually move onto local hardware, while demand for cloud-based AI training and large-scale computing keeps growing.
The opportunity isn’t necessarily winner-takes-all. It may be a future where local and cloud AI coexist, serving different needs.
What we’re watching next
Three questions stand out:
- Can advanced AI models become efficient enough to run on more affordable computers?
- Will businesses increasingly choose local AI for privacy, cost and control?
- Can Meta compete with leading global open-weight models while keeping its licensing strategy?
The bottom line
Meta and Apple’s collaboration is about more than one AI model running on two computers. It’s an early sign that powerful AI may not always need to live inside enormous data centers.
The future of AI may depend not only on how intelligent these systems become, but also on who gets to own, operate and control them.
And that could reshape the economics of the entire industry.
Stay sharp, stay organized.
Educational commentary, not personalized investment advice. Companies named are examples, not recommendations. Source: Meta’s October 15 announcement.