Meta Earnings Reveal: Llama 4 Is Being Trained On More That 100,000 H100 Chips! | Techmeme Ride Home
It is the meat of tech earnings season, so let’s get through this. Meta Q3 revenue up 19% YoY, net income up 35% YoY, and family daily active people up 5% YoY to 3.29B on average for September 2024. Mark Zuckerberg says Threads has almost 275M MAUs and is adding more than a million sign ups per day, while Meta AI has passed 500M MAUs in less than a year. Quoting Engadget:
Last month at Meta Connect, Mark Zuckerberg said that Meta AI was “on track” to become the most-used generative AI assistant in the world. The company has now passed a significant milestone toward that goal, with Meta AI passing the 500 million user mark, Zuckerberg revealed during the company’s latest earnings call.
The half billion user mark comes just barely a year after the social network first launched its AI assistant last fall. Zuckerberg said the company still expects to become the “most-used” assistant by the end of 2024, though he's never specified how the company is measuring that metric.
Meta’s assistant isn’t the only AI tool that’s boosting the company’s business. Zuckerberg said that AI improvements in its feed and video recommendations have led to an 8 percent increase in time spent on Facebook and a 5 percent increase for Instagram this year. Advertisers are also taking advantage of the company’s AI tools, he said, with more than 15 million ads created with generative AI in the last month alone. “We believe that there's a lot more upside here,” Zuckerberg said. ENDQUOTE
Mark Zuckerberg says Meta is “executing quite well” in building out AI infrastructure, as Meta raises the low end of its 2024 capex guidance from $37B to $38B. Yeah, you know, we just need to nudge that a few tiny percentage points. Which accounts for… and extra BILLION dollars. No biggie.
Zuckerberg also said Meta is training Llama 4 models on a cluster of 100K+ H100 chips, “or bigger than anything that I've seen reported for what others are doing” Quoting Wired:
Increasing the scale of AI training with more computing power and data is widely believed to be key to developing significantly more capable AI models. While Meta appears to have the lead now, most of the big players in the field are likely working toward using compute clusters with more than 100,000 advanced chips. In March, Meta and Nvidia shared details about clusters of about 25,000 H100s that were used to develop Llama 3. In July, Elon Musk touted his xAI venture having worked with X and Nvidia to set up 100,000 H100s. “It’s the most powerful AI training cluster in the world!” Musk wrote on X at the time.
On Wednesday, Zuckerberg declined to offer details on Llama 4’s potential advanced capabilities but vaguely referred to “new modalities,” “stronger reasoning,” and “much faster.”
Meta’s approach to AI is proving a wild card in the corporate race for dominance. Llama models can be downloaded in their entirety for free in contrast to the models developed by OpenAI, Google, and most other major companies, which can be accessed only through an API. Llama has proven hugely popular with startups and researchers looking to have complete control over their models, data, and compute costs.
Although touted as “open source” by Meta, the Llama license does impose some restrictions on the model’s commercial use. Meta also does not disclose details of the models’ training, which limits outsiders’ ability to probe how it works. The company released the first version of Llama in July of 2023 and made the latest version, Llama 3.2, available this September.
Managing such a gargantuan array of chips to develop Llama 4 is likely to present unique engineering challenges and require vast amounts of energy. Meta executives on Wednesday sidestepped an analyst question about energy access constraints in parts of the US that have hampered companies’ efforts to develop more powerful AI.
According to one estimate, a cluster of 100,000 H100 chips would require 150 megawatts of power. The largest national lab supercomputer in the United States, El Capitan, by contrast requires 30 megawatts of power. Meta expects to spend as much as $40 billion in capital this year to furnish data centers and other infrastructure, an increase of more than 42 percent from 2023. The company expects even more torrid growth in that spending next year.
Meta’s total operating costs have grown about 9 percent this year. But overall sales—largely from ads—have surged more than 22 percent, leaving the company with fatter margins and larger profits even as it pours billions of dollars into the Llama efforts. ENDQUOTE
#llama4 #llama3 #ainews #ai #meta #technews