Apple and Meta put Meta’s open-weight Muse model on high-end Macs

Apple and Meta are joining forces to bring one of America’s most powerful open-weight AI models to Apple’s high-end Macs, Axios reported¹. The Financial Times described the arrangement as a model tailored for Mac computers², and CNBC reported that the two companies are working to make Meta’s Muse model run on Macs³.

Axios called the collaboration a striking détente between two longtime Silicon Valley rivals whose leaders have spent years trading barbs over privacy and the future of computing¹. The companies still compete over devices and the next generation of computing, Axios reported, but both see an opportunity in bringing powerful AI capabilities to personal computers¹. “We’re excited for mommy and daddy to be playing nice again,” Meta’s AI chief Alexandr Wang told Axios¹. CNBC noted that Tim Cook and Mark Zuckerberg have clashed for years³.

Why it matters to investors

The deal sits on the open-model side of the industry’s central argument. Zuckerberg and other AI industry leaders have championed open models, the Financial Times reported², and Muse is the model Meta has pushed furthest in that direction; AI Magazine noted Zuckerberg’s own account of the personal AI product Muse Spark, which he set out at Meta Connect⁵.

There is a capital-spending angle as well. Muse gave a new shot of life to the AI trade when it was released last month, and its success could draw investor scrutiny away from hyperscaler spending, according to an account of the earnings season ahead⁴. Running a frontier-class model on a customer’s own hardware is a different cost structure from serving it from a data centre, which is the comparison investors are being invited to make.

The AI race is reshaping relationships across the technology industry, Axios argued in explaining why two rivals would co-operate at all¹. For investors, the practical question is narrower than the symbolism: whether putting a capable open model on a laptop changes where the margin on AI sits, between the company that trains the model and the company that sells the machine it runs on^2,3.

What to watch

The commercial terms have not been disclosed in any of this coverage, and neither Apple nor Meta has set out who pays whom^1,2,3. Whether the arrangement extends beyond high-end Macs is also open¹.

For Meta, the question is whether an open model distributed through a rival’s hardware supports the case its executives have made for openness^2,5. For Apple, it is whether a partner’s model does the work its own AI effort has not. The next set of results from the large AI spenders is where the capex argument⁴ gets tested.