00:03Yeah, I guess there's a question of why is NVIDIA doing this and then what does Synopsys get out of it?
00:07You know, it's like Synopsys makes the software by which chips are designed but also validated.
00:13So before you send a chip through the fab to be manufactured, you want to know that that design works.
00:18And that's a big part of what Synopsys does with all kinds of tech companies.
00:22And so part of the interpretation of why is NVIDIA doing this is it's a sales channel.
00:27Because Synopsys is basically saying, imagine how good our software would be if it was underpinned by NVIDIA's GPUs and NVIDIA's catalog of software, which largely is called CUDAX.
00:39It's basically a library of building blocks for AI that just make existing software platforms better.
00:45And so the next thing you're going to ask me about is probably circular financing.
00:49But from NVIDIA's perspective, you know, it gives them entry to a tool or a platform that lots of people are using.
00:57So, circular financing?
00:59I mean, come on.
01:01Carol did it.
01:02Is it just getting crazy?
01:04Or is it real?
01:05I don't know.
01:06The parties on this specific deal, NVIDIA and Synopsys, would say this is not circular financing in the same way that the graphic you're showing to our video audience illustrates.
01:15There is no requirement that Synopsys purchases from NVIDIA NVIDIA's GPUs.
01:23Additionally, and this is all according to NVIDIA's CEO, Jensen Wang, who was speaking this morning.
01:27Additionally, as I said, Synopsys does business with all kinds of chip makers.
01:32And NVIDIA also does business with other makers of chip design software.
01:38Siemens is one, for example.
01:39And none of that's going to change.
01:41Everyone is free to carry on as they want.
01:43And it goes back to the why is NVIDIA doing this?
01:45You know, this is an area of the technology world we call electronic design automation, EDA software.
01:53And a lot of people feel that those systems which have historically been run on CPUs, old engineering platforms, could be a lot better if they were run on GPUs or AI accelerators.
02:04They're just better pieces of software as a result.
02:07But there are loads of players.
02:08Now everyone's free to have better software is how NVIDIA would put it.
02:12But why?
02:13I mean, Synopsys wants to have NVIDIA as a customer, right?
02:17Like, so why did they need to?
02:20Like, what does that $2 billion investment do?
02:24Right?
02:24So this is like highly analogous with the Intel situation.
02:28And it's highly analogous with the Nokia situation.
02:31And so this year, I've had two specific opportunities to ask Jensen Huang, what's the rationale behind taking a stake?
02:38Like, why do you need to own between 2% and 3% of these companies?
02:43And his answer is always, I thought it would be a great investment and that in the future that investment will pay off.
02:50So NVIDIA got these Synopsys shares at $414.79 a piece.
02:56The stock's now higher than that.
02:58I can't see it on the screen.
02:59$440 a piece.
03:00But I remember back in October when we were in D.C., I asked the same question of Jensen, why did you have to take almost 3% of Nokia?
03:07And he was like, well, it looks like a pretty genius move now, doesn't it?
03:11Because Nokia shares are up 20%.
03:13And I just like, in the moment, I couldn't believe that was his answer.
03:16But he genuinely believes that it's going to be a great investment because he sees those engineering partnerships changing the respective fields that they're going into.
03:24OK, Ed, we only have three minutes left, but there's two more stories we want to hit with you.
03:28One is DeepSeq because China's DeepSeq unveiled two new versions of an experimental AI model that it released weeks ago.
03:34It adds fresh capabilities the startup said would help with combining reasoning and executing certain actions autonomously.
03:39I'm old enough to remember earlier this year when a DeepSeq update comes out and tank shares of NVIDIA.
03:46We don't see that and we haven't seen that.
03:48Where is DeepSeq when it comes to competition?
03:51OK, so DeepSeq v3.2 has been out for a number of weeks now in the form of DeepSeq v3.2 EXP, experimental.
04:00They just dropped the EXP bit.
04:02And it's an open source model that, against benchmarking, performs very well against GPT-5, OpenAI's latest model.
04:10Where they've made a difference in this open source world is that, like OpenAI, DeepSeq wants to make the tool more useful.
04:16So they have the human reasoning element.
04:18It parrots the behavior of a human, but it replies the tool in a more useful way.
04:22Markets reacted in January through April the way they did because they didn't understand how DeepSeq had achieved that performance for a model at the low-cost point that it did.
04:35Does the world need to be worried?
04:39Or I guess, like, how does this fit in?
04:42How competitive is it, real quickly?
04:44It's competitive.
04:45Models at this scale have tens of billions of parameters or more.
04:48And what DeepSeq did is called mixture of experts or MOZ designs, where if you have a 100 billion parameter model, when you run the inference, you actually run it.
04:57You don't need to access all 100 billion parameters.
04:59They found a way of just accessing, say, 10 billion parameters or 60 billion parameters, which makes the cost of running it in the compute much more efficient.
05:07So the model is as performant.
05:09It's just easier to run.
05:10I do feel like everything's now about efficiency in terms of power use and just getting things done.
05:14Hey, what do we need to know about Masayoshi's son, the SoftBank founder, who said he wouldn't have sold off NVIDIA shares if his company had unlimited money to bankroll its next investments in artificial intelligence?
05:26Is there something important here that we should acknowledge?
05:30All you need to know is that he wishes he hadn't sold it, but he didn't have enough money.
05:34He needed more money.
05:35So he sold the thing that could get him money, NVIDIA shares, and then he used it to invest in the thing he thinks will make him more money in the future, OpenAI.
Be the first to comment