Loading summary
Transcript6 lines
- [00:01]
A
If you're an H Vac technician and a call comes in, Grainger knows that you need a partner that helps you find the right product fast and hassle free. And you know that when the first problem of the day is a clanking blower motor, there's no need to break a sweat. With Grainger's easy to use website and product details, you're confident you'll soon have everything humming right along. Call 1-800-GRAINGER clickgrainger.com or just stop by Grainger for the ones who get it done.
- [00:31]
B
If you're the purchasing manager at a manufacturing plant, you know having a trusted partner makes all the difference. That's why, hands down, you count on Grainger for auto reordering with on time restocks, your team will have the cut resistant gloves they need at the start of their shift and you can end your day knowing they've got safety well in hand. Call 1-800-GRAINGER click granger.com or just stop by Granger for the ones who get it done.
- [01:00]
C
Welcome to the podcast. I'm your host, Jaden Shafer. Today on the show we are talking Nvidia news. There's a lot going on. Nvidia is tightening their China sales terms. They're demanding people pay up front. They're also expanding their footprint, whether that's construction sites to fusion reactors. There is a lot going on with Nvidia and AI. So today on the show we're going to talk about the top stories, basically everything happening with Nvidia and, and spoiler alert, there is a lot of drama when it comes to their Beijing H200 deal that's going down right now. So before we get into all of that, I wanted to mention if you want to build AI tools without knowing how to code, go check out my startup, which is AI Box AI. You can describe a tool you want to create whether that's doing a workflow linking together multiple AI models. And our no code AI app builder will create it for you. It will chain together different AI models, pull, put in prompts and create something really cool that you can publish and use. So if you want to go check it out, there's a link in the description to AI Box AI. You also get access to over 40 of the top AI models to test them out. So you don't need subscriptions to Claude and Gemini and Grok and OpenAI and everything else. It is $20 a month. Go check it out. Link in the description to AI Box AI. All right, let's get into the story with Nvidia right now they are tightening their grip on how their most advanced chips are sold in China. There is obviously a lot of demand for their hardware. It's surgin across basically every industry from construction to nuclear fusion. But they're now requiring Chinese consumers or customers, basically a lot of these big organizations to pay the full cost upfront for their H200AI chips. This is according to routers they were reporting on an anonymous source. But this new policy essentially leaves very little room for flexibility. There is no refunds or order changes allowed once a payment is made from China. This goes into, I think an issue where Nvidia is concerned. They're going to make a lot of chips and then Beijing, the government is going to shut down the, you know, the flow of the chips in and they're going to be out this money because maybe they built these special chips that are only really useful in China because of export construct restrictions. They're not going to be wanting to be purchased in the United States and there could be a lot of issues there. So they're saying pay up front, it's currently approved, we'll get you your H2 hundreds but there are no refunds, no cancellations. Some of their customers are going to be permitted to use commercial insurance or asset backed collateral. But, but the terms are way stricter than Nvidia's previous policies, especially for American companies. And I think like in the past in China definitely, but for American companies they could do, you know, partial deposits and so this is kind of definitely getting a lot stricter on China. This move is coming as regulatory approval for the H200 is sort of unclear both in the United States and in China. Bloomberg previously reported that Beijing's expected to allow Nvidia to sell the chips, but there's some restrictions aimed at preventing their use by the military or state owned enterprises or sensitive infrastructure operators. It's kind of interesting. It's the same way Chinese technology, we don't really want it in our military in the United States. So it seems like it's kind of back and forth. But Nvidia is, you know, the most advanced hardware and if China doesn't want to fall behind in the hardware race, I think they know that this is, you know, kind of the direction they need to pursue. But of course in a government perspective it's way more complicated than that. So despite all of the political friction, the demand for the H200 is really strong. Chinese companies have reportedly placed orders for more than 2 million of the GPUs for delivery in 2026, which has made Nvidia ramp up production. But I think once Nvidia sees that they are ramping up production, they really want to make sure that they get their money. And this is coming as in the past. Nvidia, on their public earnings, literally wrote off China as a company last year and said, we don't expect to make any money from China. We're not making any forecasting, any predictions for making money out of China. Because the regulatory environment was so complicated, it seems to have loosened up. It might be softened just enough where Beijing wants to buy chips and Trump administration is going to allow chips to be exported. And there's kind of this middle ground now. Who knows if that will last forever. But it seems like they're going to try to sell, I mean, over, over 2 million orders already for the GPUs. So it seems like they may have found a moment where they're going to be able to send more to China. They've been navigating a really tricky relationship right now between essentially meeting global demand but also managing geopolitical risks. Um, that came at a very high cost when U.S. regulators required them to obtain a license to export its H20 chips to China. That forced Nvidia to write down $5.5 billion in inventory. So its data center center business right now has a lot of additional scrutiny coming. Nvidia is also pushing aggressively into what it calls physical AI, where it's applying its hardware and software stack to machines that operate in the real world. Um, so that's another bright spot for the company, even though the GPU story is a little bit strange for them. So, in any case, one example of this is a new partnership they just made with Caterpillar, which they're piloting an AI assistive system that essentially goes inside of their midsize Cat 306Cr mini excavator. The system is called Cat AI Assistant, and it's built on Nvidia's Jetson Thor platform. They were demonstrating at CES this week. But Brandon Hoopman, who's Caterpillar's vice president of data and AI, said that this new system uses a fleet of AI agents to answer operator questions. It's going to bring up, you know, if there's some sort of safety concern or guidance, it's going to provide access to documentation, schedule maintenance, and it can do that directly from inside the machine. So for Caterpillar, I think a lot of the value is in the data. The company's machines already transmit about 2,000 messages per second. To its systems. And that data can be now turned into real time insights without requiring operators to step off of the job site. Caterpillar is also using Nvidia's Omniverse simulation tool to create digital twins of construction sites which allow it to test scheduling scenarios, better estimate material needs before work actually starts on a construction project. And the company is already operating fully autonomous vehicles in mining. A lot of the executives at Caterpillar think that these pilots are going to be kind of the next step towards broader augmentation and automation. So I think this collaboration fits really nicely into Nvidia's long term vision. Bill Dali, who's Nvidia's chief scientists, last year he was talking and he said that physical AI represents the company's next major kind of frontier, the next thing that they're going to focus on. And during its CES keynote last week, Nvidia outlined the full stack approach that now includes OpenAI models, simulation platforms, developer kits which are designed to move AI from screens into machines. Deepu Tala, who's Nvidia's vice president of robotics and edge AI, also argued that physical AI is not limited to robotic companies. In his view. He says that almost every industry is now building machines that are going to, you know, sense, reason, act in the physical world. So I think Nvidia is going to power a lot of that. At least that's what Deepu has been saying. I think that vision extends into another area that I'm, I think is actually mind blowing. A lot of people are interested, excited, some are skeptical of, and that is in the energy space, specifically in fusion. So this year at CES 2026, Commonwealth Fusion Systems said that they installed the first of 18 massive superconducting magnets inside its spark fusion reactor. They said that when completed, the magnets will form a donut shaped magnetic field designed to confine plasma hot enough to produce more energy than it consumes. Which of course is kind of the golden ticket if anyone could achieve that, if this plasma can produce more energy than it consumes. Because right now, yes, we have some really incredible technology. We can produce plasma. It's hot, but it takes more energy to make it than it consumes. If we can get it the other way around, it's basically an infinite money glitch. Not really, but this is, this is kind of the concept, right? If you put a dollar in and get $2 out, that's, that's kind of the, the golden ticket and that's what they're trying to do with energy. So each Magnet, though, with this weighs more than 24 tons and can generate a 20 Tesla magnetic field, which is roughly 13 times stronger than an MRI machine. So in order for them to actually reach that performance, the magnets are cooled to nearly absolute zero. While they carry more than 30,000amps of current inside the reactor, the plasma is going to reach temperatures exceeding 100 million degrees Celsius and CFS. The company is expecting to install all 18 magnets by the end of the summer. According to their co founder and CEO Bob Mumgad, the company hopes to turn on Spark. This is their product next year, putting on track to demonstrate energy net gain and move closer to commercial fusion power in early 2030. So this is exciting, but also I'm going to put a big caveat on to say, you know, they're hoping to have this done by the end of the summer. This is not currently completed, so this project isn't currently, you know, successful, profitable, net energy generating, but we'll see. And I have high hopes for, for fusion. I think this is an exciting place. Of course, this is an interesting topic, but why bring it up in a podcast episode about Nvidia? And the reason is because in order to do, to reduce the risks, they're actually working with Nvidia and Siemens to build a digital twin of the reactor using the Omniverse simulation tool, Nvidia's Omniverse platform. And then Siemens is providing design and manufacturing software, while Nvidia's platform is going to help run continuous simulations alongside the physical reactor, which is just the most fascinating concept. Nvidia has basically built a platform where you can have digital clones of large infrastructure things or construction sites and you can run these simulations of, you know, real world simulations of. It's interesting because it's not just like an item, it's an environment. Right. It could be a whole construction site or in this case an entire reactor. But so unlike earlier simulations that modeled individual components in isolation, I think this was more common. The digital twin is going to mirror the full system in real time. Engineers are going to be able to test parameter changes virtually before they actually apply them to the real machine. And I think that accelerate is going to accelerate learning. It's also going to make mistakes, cost a lot less. According to CFS, they have already raised nearly $3 billion to build Spark, including a $863 million Series B2 round that include investments from Nvidia and Google. So Nvidia puts money in. They also help them because they know that, you know, they're going to become a customer. And also this is just a really interesting company. I think it's going to be good PR for Nvidia. Google also invested in this, and while $3 billion is obviously a lot of money, its first commercial power plant that they're planning on building called Ark, is also expected to cost several billion dollars more. So they're not out of the woods. They need to keep raising money. Their CEO said advances in AI in simulation could help bring fusion power to the grid faster than they previously thought. I think there is definitely a growing urgency around energy because AI consumes a lot of it and we all love our AI. And so the CEO of CFS said that he believes digital twins and machine learning is going to play a central role in turning fusion from a scientific milestone into a practical power source. Nvidia obviously has a lot of tightening controls right now on where its chips go and how they're paid for. The company is simultaneously embedding themselves deeper into machines, infrastructure, infrastructure, construction, the energy systems that are going to define basically the next industrial era. So there is a lot of excitement going on. Nvidia's got some struggles, some challenges, but they also have a lot of exciting things happening. So this is going to definitely be an exciting company to follow this year. Thanks so much for tuning into the podcast. If you enjoyed the episode, make sure to leave a rating and review wherever you get your episodes. It helps us show a ton to get found by amazing people like yourself. Make sure as always, go check out AI box AI20amonth for over 40 models and image, text, audio. It's all there. You could try everything side by side. It's amazing. I hope you love it. I'll leave a link in the description and I hope you have an amazing rest of your day.
- [12:54]
A
If you're an H Vac technician and a call comes in, Grainger knows that you need a partner that helps you find the right product fast and hassle free. And you know that when the first problem of the day is a clanking blower motor, there's no need to break a sweat. With Grainger's easy to use website and product details, you're confident you'll soon have everything humming right along. Call 1-800-granger. Click granger.com or just stop by Granger for the ones who get it done.
- [13:25]
D
Hey Sal, Hank, what's going on? We haven't worked a case in years. I just bought my car at Carvana and it was so easy. Too easy. Think something's up? You tell me. They got thousands of options. Found a great car and a great. And it got delivered the next day. It sounds like Carvana. Just makes it easy to buy your car, Hank. Yeah, you're right. Case closed.
- [13:48]
A
Buy your car today on Carvana. Delivery fees may apply.