4.8 • 56 Ratings
🗓️ 16 May 2025
⏱️ 8 minutes
🧾️ Download transcript
Click on a timestamp to play from that location
0:00.0 | This is a key story. So let me continue to break it down. Meta essentially here is spending like it's 2023 when scale was everything. |
0:06.7 | But here's the rub. The pace of AI breakthroughs is slowing and the market has already been shifting towards smaller, more efficient models. |
0:14.2 | So even the name here, behemoth, it feels out of step. And increasingly, so does the pace of meta's capax spend. |
0:20.7 | Let me put it in perspective, |
0:21.7 | $72 billion this year, and as a share of revenue, is far more aggressive than even the hyperscalers |
0:27.7 | who at least have massive enterprise cloud businesses to backstop the spend. So the question for |
0:33.5 | investors isn't just why Behemoth is late, but it's whether Meta's AI strategy is out of |
0:38.7 | sync with where the real momentum is heading. Meta had no comment on the delay. Now, some of the |
0:43.2 | most compelling models right now, they are getting smaller, cheaper, faster to run. Open |
0:48.1 | AI's GPT-40, it's optimized for costs and real-time interaction. Google's lighterweight, |
0:53.2 | Gemini 2.0 Flash. |
0:54.9 | You've got Chinese open source models from Alibaba and Deepseek. They all follow that trend, |
0:58.6 | too, underscoring practicality over power. Now, meta has released a few smaller Lama 3 models, |
1:03.9 | but they came with controversy. The version that briefly topped AI leaderboards, it wasn't the same |
1:08.8 | as the one that was shipped to the public. |
1:11.1 | So Meta here isn't hitting either end of the market. It doesn't have a benchmark beating |
1:15.6 | flagship, essentially what amounts to prestige in the AI race, nor does Lama have breakout adoption. |
1:21.3 | And as the journal piece points out, 11 of the 14 researchers behind the original Lama paper, they are gone, raising questions about |
1:29.0 | Meta's ability to maintain momentum in the key and only gaining traction open source AI |
1:34.4 | race. |
1:35.4 | There's also this question of monetization, guys. |
1:37.4 | Open source gives meta reach, but not necessarily revenue. |
... |
Please login to see the full transcript.
Disclaimer: The podcast and artwork embedded on this page are from CNBC, and are the property of its owner and not affiliated with or endorsed by Tapesearch.
Generated transcripts are the property of CNBC and are distributed freely under the Fair Use doctrine. Transcripts generated by Tapesearch are not guaranteed to be accurate.
Copyright © Tapesearch 2025.