Latest / Neural Newscast / Nvidia Names OpenAI and Anthropic as Vera Chip Users [Model Behavior]
Transcript
- [00:00] Nina Park: From Neural Newscast, this is Model Behavior, [00:02] Nina Park: AI-focused news and analysis on the models shaping our world. [00:12] Nina Park: I am Nina Park. Welcome to Model Behavior. This program examines the intricate ways AI systems [00:19] Nina Park: are built, [00:20] Nina Park: deployed, and operated in professional environments. Today we are analyzing a significant shift in [00:27] Announcer: the semiconductor landscape as Nvidia moves to expand its presence [00:31] Announcer: within the data center infrastructure by introducing its own central processing units. Nina, [00:39] Announcer: our reporting today comes directly from the Computex conference in Taiwan. [00:44] Announcer: Bloomberg is currently reporting that Nvidia CEO Jenson Huang has officially named the first wave [00:49] Announcer: of customers for their upcoming microprocessor known as the Vira CPU. This is a strategic move [00:56] Announcer: for a company that has historically dominated the market through graphics processing units. [01:00] Announcer: That is correct, Thatcher. During his keynote presentation, Huang specifically identified OpenAI, [01:07] Announcer: Anthropic, and SpaceX as the primary early adopters of the Vira CPU. These organizations [01:13] Announcer: are expected to integrate the new chips into their data centers, to bolster [01:17] Announcer: their existing infrastructure. It is a clear signal that Nvidia intends to secure [01:22] Announcer: the entire compute stack rather than just providing accelerators. It's a compelling roster [01:28] Announcer: of partners. While we certainly expect AI labs like OpenAI and Anthropic to lead the way, [01:33] Announcer: the inclusion of SpaceX is a noteworthy datapoint for the industry. Nina, [01:38] Announcer: did the presentation provide any specific technical details on how the Vela architecture [01:42] Announcer: differs from the Hopper or Blackwell systems we've discussed in previous episodes? [01:47] Announcer: The technical details remain focused on deployment scale for now, but the timeline is [01:52] Announcer: the critical factor for investors. Huang noted that the Vira product will enter full production in [01:58] Announcer: the third quarter of this year. This rapid schedule suggests the silicon is already [02:03] Announcer: through its primary testing phases and that Nvidia is confident in its ability to scale [02:08] Announcer: this architecture immediately. That confidence is necessary for a firm with a $5 trillion valuation. [02:14] Announcer: But there is some friction in the broader narrative. While these chips are coming, [02:19] Announcer: Nvidia recently issued a somewhat conservative forecast as competition from Intel and AMD increases. [02:25] Announcer: Nina, how does the Viralaunch address the fact that rivals are catching up in the data center space? [02:30] Thatcher Collins: Diversification appears to be [02:32] Thatcher Collins: the primary strategy that's here. Beyond these high-end data center chips, [02:37] Announcer: Nvidia is [02:37] Thatcher Collins: making a concerted [02:39] Announcer: push into the Windows laptop market. By developing silicon for personal computers, [02:44] Announcer: they are taking the competition directly to Intel and AMD on their traditional home turf. [02:51] Announcer: They are no longer content being solely a provider of backend infrastructure. [02:55] SPEAKER_03: The internal corporate dynamics are also shifting in interesting ways we saw reports [03:01] SPEAKER_03: earlier this month that Jensen Huang's own compensation package shrunk by 27% [03:07] SPEAKER_03: due to smaller stock awards it suggests that even with a massive valuation, [03:12] SPEAKER_03: the board is being cautious about executive incentives [03:15] SPEAKER_03: as they transition into this next phase of hardware production and distribution. [03:20] Announcer: There is also a significant geopolitical angle to consider. In May, Huang mentioned that he sees [03:26] Announcer: the Chinese market eventually opening back up to high-end United States AI chips. [03:32] Announcer: If Nvidia can pair the VERA CPU launch with a renewed presence in China, [03:37] Announcer: it could offset [03:37] Announcer: the competitive pressures they are feeling from domestic rivals here in the United States. [03:42] SPEAKER_03: It feels like a pivotal moment for their product roadmap. Moving from GPUs into CPUS [03:48] SPEAKER_03: and laptop chips suggests Nvidia wants to be the fundamental silicon provider for every layer [03:54] SPEAKER_03: of the AI economy. We will have to see if the third quarter [03:57] SPEAKER_03: production ramp for Vera meets the high expectations set by OpenAI and SpaceX. [04:03] Announcer: We will be watching that rollout closely. Thank you for listening to Model Behavior, [04:08] Announcer: a neural newscast editorial segment. You can find more information regarding our reporting [04:14] Announcer: at mbn.neuralnewscast.com. Neural newscast [04:19] Announcer: is AI-assisted human-reviewed. View our AI transparency policy at neuralnewscast.com. [04:27] Nina Park: This has been Model Behavior on Neural Newscast. Examining the systems behind the story.