英伟达增持1500亿美元回购授权，AI资本开支与竞争格局引关注

Yeah. You know. I mean, when we look back, um, I see that the reaction was initially quite positive. And yeah, there's obviously lots of different ways you can take this messaging. I think a lot of people took it's a view that, you know, that these open I open I and anthropic in particular looking to strengthen the moat they have against, for example, the likes of metals News today that they're uh, uh, taken a new CEO for their own business and that, you know, that actually, uh, as much as there may be some concerns about, um, the pace at which I was developing and our control over it, that actually that this could be a good mechanism for them to kind of protect the moat they have, because it would constrain the whole industry in terms of the pace of development.

Uh, you know, I think broadly speaking, obviously there are still concerns about, you know, the kind of 3 to 5 year outlook for, for AI. But, you know, certainly in the near term and again, reflected in what Nvidia's done today, um, the market is pretty confident that there's enough momentum to sustain the next one two years. And enterprises are still at an early stage of deploying AI into their businesses. And, you know, even if you didn't have any further evolution of, uh, uh, the frontier models, uh, from where they are today, there's still lots of technology capability that French buys to play with and really shift the way that businesses around.

So we're approaching the end of the third quarter. The fourth quarter will begin soon on October 1st, and that means earnings season is kind of around the corner. Um, the big tech names report at the end of October. What are we likely to hear from them when it comes to CapEx? I mean, are they going to continue to increase CapEx at the rate that they've been increasing thus far? Or will all the political discussion over AI and whether we should slow development mean that some of the hyperscalers will start?

We'll start to see the numbers reflect that slowdown. Um. Great question. My my sense would be that, uh, uh, the market is still anticipating growth in spending next year, obviously nowhere near the pace we've seen, uh, this year, but I think we're still expecting to see it increase than that. Yeah. As much of there's been a lot of fanfare around slowing the pace of model development. There's still a huge amount of implementation spend to come through.

Probably right now, the biggest constraint on, um, the pace of capital spending growth is the supply side all the way from, you know, the kind of fab units to to the chips to the capacity to make the chips, and then also all, all the way through to energy and space and. Yeah, how quickly can you actually deploy data centers, which is where ultimately all of this CapEx ends up. Um, so that's probably the biggest natural constraint, let's say, on spending matter probably doesn't surprise you, but I am way ahead of both John and Scarlett on I am way out there.

I've got wind so runs on my phone. Yeah, that's got the apps, I got ChatGPT you got Gemini, I've got Claude, I've got he added muse. I added muse somewhere, but I'm guessing I'm not the monetization that these billion dollar CapEx are being spent for. Talk to us about enterprise integration of AI. Is that happening? Because it's not cheap to deploy AI and use AI in an enterprise? It certainly isn't. I mean, it's definitely happening, you know, across every sector.

And if anything, you know, this year, um, enterprises are kind of running into some spending, you know, kind of constraints or kind of, um, obstacles they probably probably hadn't anticipated because, you know, that they've kind of largely let their engineers run loose with playing around with this stuff, and then they suddenly realize that the consumption is ahead of the budget they put in there, and they're now having to scratch and go, okay, do we increase our dollar budgets, uh, to continue with at a pace development, or do we have to start to kind of find ways to kind of moderate, uh, or change the way our engineers, uh, uh, are using AI tokens to kind of develop things?

And, you know, the answers so far been a bit of a mixture of the two. Uh, but yeah, the enterprise market's huge and it's a massive, massive, massive multi-trillion dollar market. Uh, when you look at technology as a whole, um, and I think it's, it's, it's becoming the new, you know, kind of theater of competition, if you like, between the big players. And yeah, this is where matters coming in with this enterprise strategy and historically matter very much a consumer focused, ad funded business and looking to kind of, uh, uh, compete with OpenAI and anthropic in that space.

I think it's going to be a really interesting space over the next 12 to 24 months, to see how that competition evolves, but it's kind of really a huge opportunity for all of them. Yeah, and Matt is doing that, as we mentioned, by, uh, tapping MongoDB as president and CEO to lead this new AI platform for enterprise customers. Has meta ever done anything at an enterprise level targeting enterprise customers? I mean, is this the first um, I think it if it isn't the first and it's kind of close to a first.

Yes, certainly. I think that kind of biggest and most significant push into this. I mean, they've been, you know, working and engaging with corporates indirectly, I'd say in, in the eye space so far because you know that their models are open source. And so a lot of companies do use them. But I guess that's kind of a bit of an arm's length interaction because there's not so much direct involvement. This is, you know, much more significant, almost like kind of consulting, um, evolution of, you know, um, yeah, I guess kind of for deployed engineers, that's the kind of phrase that we never use to talk about, you know, kind of helping organizations to kind of figure out, um, how they can use AI in, uh, specific enterprise use cases to kind of improve efficiency or create new opportunities.