Assessment
The report saying Meta plans to double compute next year to 14 GW alongside recent deal announcements increases my confidence that Meta’s capex will rise again next year.
I=8 Meta launches Muse Spark 1.1, claims it outperforms Gemini on several agentic AI, coding, and multimodal benchmarks
Zuckerberg described the model as having “state-of-the-art or very close to it” agentic reasoning and tool use.
He said that Muse Spark 1.1 performed better than Google’s Gemini model in several areas related to agents, coding and multimodal capabilities.
“That is a pretty interesting milestone because I think this may be the first time, at least that I can remember, that Meta’s models are better than all of the Google models,” Zuckerberg said ahead of the release.
Zuckerberg said they are still trailing some of the larger AI labs, including Anthropic and OpenAI but believes the upcoming Watermelon model will “push this maximum frontier of intelligence.”
He declined to say when Watermelon will arrive, saying that the focus is on quality.
Muse Spark 1.1 is priced at $1.25 per million input tokens and $4.25 per million output tokens, making it more expensive than OpenAI’s GPT-5 mini and Anthropic’s Claude Haiku 4.5, but cheaper than Anthropic’s premium Claude Sonnet 4.6.
Assessment
Some third-party tests already confirm Zuckerberg’s argument that it outperform’s Gemini and is leading on a number of areas (Notion).
I=7 Zuckerberg said Meta doesn’t have excess compute yet but exploring cloud business makes sense
Zuckerberg told Bloomberg that the offers they get may make sense to consider leasing out compute instead of putting them to own internal use.
“The offers that you get for using the compute are so high that it may make sense, in some cases, to rent out or consider those kind of deals instead of your own internal uses,” Zuckerberg said.
Zuckerberg said Meta has not already overbuilt or has excess compute to lease out.
He said he’s interested on recent SpaceX’s approach to lease compute to Anthropic and Google.
“The SpaceX model I think is quite interesting in terms of just making these short-term deals that are at a big premium,” Zuckerberg said. “So we get offers for all kinds of stuff like this and we’ll evaluate them and see what makes sense.”
In an interview, Meta’s CTO Andrew Bosworth AI is entering the biggest period of uncertainty in the technology industry (min 0:00–0:30).
“This is the least clear vision we’ve had of even the relatively near future in a long time.”
“Two years ago we started to get a sense like, whoa, this is bigger than we thought. I think last year it got to a point of seriousness that it became an almost existential conversation.”
He said data will become the biggest competitive advantage in AI (min 15:31–17:52).
He highlighted that Meta’s AI advantage comes from consumer distribution and commerce data (min 40:31–43:40).
“Meta AI is absolutely the best at local recommendations.”
“It’s really, really good at commerce, at shopping, helping people make decisions.”
“Our ability now to give each of those people a super intelligent AI that’s going to help their business grow is great.”
He pointed out that AI infrastructure will ultimately depend on energy availability (min 47:46–48:37)