Loading summary
Transcript21 lines
- [00:01]
AWS Announcer
How does F1 turn data into insights at 200 mph? AWS is how fans get inside the strategy. AWS powers next level innovation for millions of businesses.
- [00:19]
Julie Chang
Welcome to Tech News briefing. It's Tuesday, October 14th. I'm Julie Chang for the Wall Street Journal. Last week AMD rocked the market with a multi billion dollar partnership with OpenAI. Of the most direct challenges yet to industry leader Nvidia. We'll introduce you to the old school CEO leading the underdog in the race for AI chips. Then we'll tell you about Elon Musk's high stakes race in Memphis to build out two colossal data centers in order to stay in the AI game. But first, Advanced Micro Devices groundbreaking deal with OpenAI sent its share price so soaring by over 33%. And it positioned AMD as a formidable challenger to its longtime rival Nvidia. Behind the agreement was AMD's CEO Lisa Su, whose leadership is now being watched by the market for future game changing partnerships.
- [01:19]
Lisa Su
Certainly there's a lot of engineering work, but our teams are working together on hardware, software. We're ensuring the supply chain. All of those elements are set up and ready to deliver on this massive commitment.
- [01:32]
Julie Chang
That was sue speaking with Bloomberg Tech last week. WSJ reporter Robbie Whelan joins us to discuss how Hsu's leadership led to this pivotal moment for the company. So Robbie, tell us about AMD's Lisa Su. What's she like?
- [01:45]
Robbie Whelan
Lisa Su has been CEO of Advanced Micro devices for about 11 years now. And when she took over in 2014, the company had a market cap of below $3 billion. Today, at the end of last week, it was nearly $350 billion. Really stunning turnaround. Lisa Su was born in Taiwan. Taiwan has emerged as the source of a lot of some of the greatest minds in the AI infrastructure world. Jensen Huang, who's actually her distant cousin, is also from Taiwan. She immigrated to the US when she was three. She describes herself as an engineer's engineer. She has three degrees from MIT in electrical engineering. She really understands the technology that she is selling to customers backwards and forwards. That's kind of what sets her apart from a lot of other CEOs in the tech world. She's just a very talented engineer. She won something called the Robert and Noyce Medal in 2021. She was the first woman to ever win this prize. And people I've talked to who work with her say that that really comes through when you deal with her. Her deep understanding of how the chips that AMD makes, how they can help People, how they can be attuned to customers needs. And more than that, and this is probably the real secret sauce, is that she also is a very smart businesswoman and she just makes everyone who works for her think not only about how to optimize products in a technology sense, but also think about them in sort of an economic sense. Does this product make sense? Are the margins high enough? How can we design it in a way that selling it is going to get us a great return, not just a good return.
- [03:20]
Julie Chang
Going back in time a little bit, you mentioned that Su became AMD CEO over a decade ago and the company was not doing so well at that time. Why was the company struggling back then?
- [03:30]
Robbie Whelan
When Lisa Su took over amd, AMD had fallen really badly behind on chips that are called CPUs that are used in data centers. And they basically had 0% market share in data center CPUs. And when she took over, she said, look, we've got to really revamp our product line. We've got to make sure that we're making the things that people want and not just the bulk things that anyone can make, but really the differentiated kind of more advanced computing solutions, higher powered chips that are the ones that really are desirable and that people will pay a lot of money for. And so that's what she did. She completely redirected the company's approach to product development, and in a few short years, they started gaining market share.
- [04:11]
Julie Chang
Okay, so AMD's turnaround has been years in the making. Tying this back to the news from last week, how did sue get this deal with OpenAI?
- [04:20]
Robbie Whelan
Yeah, she sort of laid the groundwork for what we're seeing in the last few months, within the first 10 years that she spent as CEO working on this incredible turnaround effort. And what I mean by that is that going really hard into the data center space positioned AMD to really be there when AI took off. AI developers, people like OpenAI and Anthropic and Meta and Microsoft, all these companies that are working on developing AI products, they get the computing power to train and operate those products from cloud computing through data centers. So the real genius of Nvidia was that they figured out how to take graphics cards that are used in games gaming and apply them in really specific ways with a really big software ecosystem so they could be used instead in AI data centers. AMD didn't exactly do the same thing, but they had a big presence in data centers around the time when AI really took off. And so in late 2023, AMD released a line of chips. And these chips were their equivalent of Nvidia's GPUs that run data centers. And so just being there in the mix and talking to customers about what they need in their data cent hardware stack was what allowed Lisa Su to really position the company to go hard into AI chips.
- [05:39]
Julie Chang
AMD is still the underdog to Nvidia. What challenges does sue have ahead of her?
- [05:45]
Robbie Whelan
Yeah, I mean she's got to do more deals, she's got to sign up more big customers, she's going to have to make sure that her software that she sells alongside her chips is updated constantly. And then she's got to keep on rolling out new generations of chips that compete with Nvidia's. Nvidia is clearly still the dominant player in this market. I mean, they're about 10 times larger than AMD or more. But you know, she does have a chance to kind of really keep on grabbing market share.
- [06:14]
Julie Chang
That was WSJ reporter Robbie Whelan. Coming up, Elon Musk is gambling billions on a huge facility in Memphis, Tennessee. To catch up in the AI race, we dig into the challenges he faces after the break.
- [06:36]
AWS Announcer
Millions of players, one world, no lag. How's it done? AWS is how epic games turn to AWS to scale to more than 100 million Fortnite players worldwide so they can stay locked in with battle tested reliability. AWS is how leading businesses power next level innovation.
- [07:02]
Julie Chang
I've been rain Elon Musk has been at the forefront of innovation in electric vehicles and rockets. But when it comes to AI, he's in the unusual position of playing catch up to rivals like Sam Altman's OpenAI. To bridge this gap, Musk is making Memphis the epicenter of his ambitious and costly venture into the AI arena. His company XAI has already built a massive data center there dubbed Colossus, which houses over 200,000 Nvidia chips and powers the tech behind the AI chatbot Grok. Now Musk is on the verge of completing Colossus 2. WSJ reporter Alexander Saidi has been following the money that Musk and XAI are raising to secure for the project. Alex, do we know why Musk chose Memphis for his two data centers?
- [07:57]
Alexander Saidi
Yeah, he looked at eight other locations for a possible data center. And one important thing that he was looking at was that he essentially needed an already constructed facility or warehouse. Musk has been in the unusual position of playing catch up in the AI wars. He co founded OpenAI, but he left the company over disputes with Sam Altman and others over how they were going to run OpenAI and get capital to build out their technology. So he didn't incorporate XAI until 2023, well after OpenAI had launched ChatGPT and was on its way to growing a lot. So they liked Memphis because there was an abandoned factory there that a former Swedish appliance manufacturer called Electrolux had once occupied. And in this facility they saw the ability to rapidly bring in chips, bring in power, bring in cooling systems and water, and get a system online in rapid time. On top of that they really had a lot of favorable interactions with the Chamber of Commerce, the local utility company, which essentially assured him that he would have all the resources he would need to pull off this, this really fast, high stakes bet on artificial intelligence in Memphis.
- [09:22]
Julie Chang
What was the process of setting up that first data center?
- [09:25]
Alexander Saidi
Like, you know, Musk and his team, they sent about 42,000 chips there to start. They needed essentially to get power to the area fast and they needed a lot of power, more than the electrical grid was capable of providing them at that time. So they imported a lot of portable natural gas turbines to be installed on the site. And on top of that, Musk had a farm of Tesla battery packs brought to Colossus One that would actually help smooth the power demand because AI and these like training modules, power demand is very spiky. It goes up and it goes down based on the work that it's doing. So he sets all this up and in about 122 days they went from nothing to a fully functional standalone AI training facility, which was much faster than any company has ever been able to do in constructing a facility like this and at this size.
- [10:30]
Julie Chang
How much is this costing Musk exactly?
- [10:32]
Alexander Saidi
The company has been fundraising quite a bit. They raised 10 billion alone in debt and equity this year. We've reported they're working on a securitization facility through one of their partners, Valor, that's planning on bringing in more money to help them buy chips. But the sums we're talking about here are huge. There's a new Data Center, Colossus 2, that's under construction. We reported they're going to need at least $18 billion just to buy the chips that they need to finish their first stage. Plans to get 550,000 GPUs online there. And some things we're seeing to show like, wow, Musk is really pulling on every lever he can to get money into XAI is how he's relying on his own companies to invest in and supply XAI SpaceX invested 2 billion in XAI, which was very unusual. That company rarely makes outside investments. Tesla has a shareholder proposal before it that would authorize the board to invest an undisclosed amount into xai. Tesla and xai, as I mentioned already do business with each other with the battery packs. So it is really, really, really expensive to do this. And grok, the XAI large language model, still has a little ways to go to catch up with the likes of ChatGPT. So the big question is, okay, they've raised all this money to make it a powerful AI. Like, are enough people going to use it to justify the investment long term?
- [12:10]
Julie Chang
That was WSJ reporter Alexander Saidi. And that's it for Tech News Briefing. Today's show was produced by me, Julie Chang with Deputy Editor Chris Zinsley. We'll be back later this morning with TNB Tech Minute. Thanks for listening.
- [12:31]
AWS Announcer
How does F1 turn data into insights at 200 mph? AWS is how fans get inside the strategy. AWS powers next level innovation for millions of businesses.