Reshoring Chip Production: U.S. Hopes and Hurdles

0:00The U.S. semiconductor industry is something that has gotten a lot of attention recently. Number one, we just had Trump with his big AI plan come out and there's a huge focus on how to grow semiconductor. or deregulate that industry data centers much faster in the United States. We know that there is a bit of a competitive advantage. We have Taiwan Semiconductor Company coming into the United States and building out fabrication in Arizona as people have questions about if China is going to launch a military strike on Taiwan to capture the island. There's a lot of rules about the chips that they can ship to China, they can't ship to China.

0:32There's a lot going on in this industry. And I wanted to take a podcast to basically break down the last six months of 2025, what has happened over the industry, the big pieces of news, because it paints a very interesting picture of where we're going in the future. A lot of huge news, different things have slipped through the cracks. I feel like if you don't see this full picture, it's a little difficult to understand what the state of the industry is. So I want to break down the top stories from the beginning of the year till now with the semiconductor industry ending in some of the big news that's been announced.

1:04And of course, everyone is talking about this. There's a$500 billion deal with OpenAI and Oracle and a lot of other crazy stuff. So we're going to get into all of that. Before we do, I wanted to mention, if you want to try out any of the top AI models I talk about on the show, I'd love for you to go check out AIbox.ai. This is my own AI startup, and we're in beta right now. We'll essentially allow you to access the top 40 AI models all on one platform for one subscription price. You don't have to have subscriptions to every single model or every single company. For$20, you get access to OpenAI, Gemini, Cohere, DeepSeek, Google, Meta, Microsoft, Mistral, NVIDIA, Quen, XAI, image generators, audio generators, a ton of really cool features.

1:42And you can chat them all in the same thread. So it's incredibly useful. And it's only one subscription of$20 a month. So you can go check that out. There's a link in the description. All right, let's get into what is going on with the US semiconductor industry. So I wanted to start out in January. So the CEO of Anthropic, Dario Amadeo, he co-authored an op-ed in the Wall Street Journal. This was quite famous. A lot of people talked about this. But basically, he was advocating for stricter US chip export controls. Now, we know that obviously, this is a bit of trying to play to his strengths. He's an American company.

2:16So he's like, look, don't ship chips to China, then he knows he's got less competition from some competitors out of China, etc, etc. So in fact, to the point where Nvidia CEO Jensen Huang kind of called him out later this year and said like, hey, you guys just work on your innovation, stop telling us who we can and can't ship chips to. So anyways, that's some funny drama. But in any case, January 13th this year, in his final week in office, President Joe Biden, he made an executive order basically making a bunch of new restrictions on US made AI chip exports. So kind of following through with what Dario had done.

2:47He also established a three tier system on who they could ship it to. But none of that really matters, because I think Trump repealed that and a lot of other things. And it kind of replaced some of it, although that did still have some restrictions. On the 27th of January this year, of course, We had DeepSeq that had their, basically their R1 reasoning model, and that kicked off a huge firestorm from US companies on basically what was going on, how they were going to compete with China, and why the semiconductor industry was so important. It kind of built, made this like, us against you, American versus China, and that's kind of where we are today.

3:24All right, so let's get into what's happening. So in March, starting specifically kind of in the semiconductor industry, we have the story of Intel. So Intel announced that Lip Bhutan, who was a former board member, it was going to take over as the CEO. Intel had honestly kind of been not seeing a huge growth from AI that everyone expected. They'd been dropping, they'd been struggling, bringing on the new CEO. And of course, Tan had a really renewed focus on engineering and what they were doing. So that was kind of the big thing. In April, on the 15th, the US imposed a bunch of new export license requirements on NVIDIA's H200 chips.

3:59NVIDIA wasn't particularly happy with this, but basically NVIDIA said that they expected a$5.5 billion hit in their Q1 fiscal 2026 charges. They're like, look, we're going to lose a lot of money, not be able to ship basically their H20 AI chips to China. Then we had Jensen Huang, who went and had dinner at Mar-a-Lago with Trump and helped negotiate a partial exemption for the H20 chip, basically. But we'll see that was sort of short-lived, but I'll get into that. Earlier on in April, we also had Intel that disclosed they were going to spin off all of their non-core units. So their networking and edge division, basically to reinforce their high-end R &D.

4:39This is all going to be under TAN's leadership. There is also some rumored discussions that came out of a joint venture between Intel and TSMC. TSMC was going to take a 20 % stake in this kind of new chip fab. Neither of the companies confirmed that, but it was it was kind of the rumors at the time. So then we get to May. So in May, multiple people reported that the Trump administration was not going to proceed with enforcing President Biden's AI chip export framework. It was supposed to start on May 15. Then on May 13, the Commerce Department formally rescinded the AI diffusion rule, which was originally issued under Biden.

5:16They also basically said, like, they promised some future guidance on Huawei's Ascend chips and other matters. And ultimately, they decided and they told everyone, like, look, we're rescinding Biden's thing, but no one is allowed to use Huawei's Ascend AI chip. That's against, like, U.S. rules. And they kind of made a whole thing about it. So on May 20th, there was a whole bunch of reports that came out basically saying that the Intel CEO Tan was looking at the sale of Intel's networking and edge infrastructure units. This made up about$5.4 billion worth of revenue. So it feels like, I hate to be so pessimistic on Intel because it typically has been a great American company, but it feels like as they're spinning off what they're calling non-core units, it feels like the company is sort of struggling and they're just going to try to focus on what their core business is.

6:05they're spinning off these other things that are not really their core focus because they basically can't sustain them. So on May 20th, that's kind of when that news broke. And then a day later, China's commerce secretary came out openly and condemned the US guidance, which warned against using Huawei chips anywhere in the world. They said that this could constitute an export violation. And they, of course, threatened some legal retaliations against the US. And they didn't like the US kind of coming in and messing that up for them. On the 28th of May, NVIDIA confirmed that licensing restrictions on the H20 chips had led to$4.5 billion charge in Q1.

6:43It estimated up to$8 billion revenue loss in Q2 due to the constrained shipments. Also on that day, AMD acquired a company called Enosemi. So this was basically a startup that was focused on silicone photonics, which was really something that's very important for basically some important innovation in semiconductors. And it's interesting because now we're starting to see these acquisitions by AMD and some other big players to become more competitive in the market. Then we have June. So in June, AMD acquired another company. They were kind of on a bit of a tear. They acquired Brium, which is basically a software startup specializing in computer and AI inference optimization.

7:19The deal is essentially positioned to reduce their reliance on CUDA. And they're also trying to kind of take a challenge at NVIDIA's ecosystem dominance. I guess not the problem with NVIDIA, the strength of NVIDIA is that they don't just have chips, but they have software. They have this whole ecosystem of tools that make it really hard for other companies to kind of break in. So on June 6th, just two days later, the semiconductor company also Aqua hired the entire AI chip team from Untethered AI and basically folded their expertise and shut down the startup's support operations. So like I mentioned, on an absolute tear of acquisitions.

7:53and then that brings us to july so on the 13th of july there's a whole bunch of rumors that nvidia planned to reapply for export licenses to start shipping their h200 chips to china again this was obviously a really strategic pivot at the same time they announced a china specific chip which they're calling the rtx pro basically they have restrictions on like how powerful of a chip they can send to china so they made less powerful chips but they used ai to kind of optimize those chips and make them better. So technically, they're, you know, following the export restrictions, but at the same time, they're able to deliver a product that would be a competitive and interesting inside of China.

8:31Then a day later, Malaysia implemented a new export regulation. Basically, any export transit or transcriptment of US origin AI chips required a 30 day notification and trade permit. This is basically part of an effort to stop smuggling into China. We saw some pretty funny news where people were i think this is probably not super accurate but there was funny news where like basically singapore made up for 30 of nvidia's revenue there's a number of reasons why this happened but a lot of people at the time were saying look nvidia singapore makes up 30 of nvidia's revenue this is probably just because people in china are getting stuff shipped to singapore and then smuggling it into china so basically singapore represents china in reality i think that singapore just had a lot of purchasing offices and so nvidia would bill from singapore And so that's kind of where the revenue got marked, but it wasn't necessarily revenue going to, it wasn't necessarily chips getting shipped to Singapore.

9:24So anyways, that was kind of a funny thing. But on July 17, a bunch of news came out that a multi-billion dollar UAE deal was coming to purchase NVIDIA AI chips and that it was on hold due to US national security concern over basically they're worried that the chips are then going to get rerouted to China, even though the UAE said they weren't going to do that. So the U.S. is really trying everything they can to stop smuggling, just stop these chips from getting to China. They didn't want China to get AI supremacy. On July 23rd, the Trump administration and also the 24th, so basically this week, the Trump administration unveiled their official AI action plan, which was releasing more than 90 policy actions, including, of course, export control, reform, deregulation to accelerate AI infrastructure, and a push to promote AI exports to U.S.

10:11allies. The plan that was rolled out basically told agencies to implement location monitoring of chips and ease permit hurdles. So basically, if you have a chip, it's got to be able to tell you where it is. So it's not just like who NVIDIA can sell to, but they just don't want these chips to ever get to places that are banned. Then we come to the latest news, which is that on July 24th, Intel actually reported Q2 results that showed$12.9 billion in revenue. but a$3 billion net loss. The CEO announced layoffs, which was about 24 ,000 rolls. He's been trimming the workforce to a 75 ,000 by the end of the year.

10:51That's kind of their goal. And at the same time, Intel canceled its Germany and Poland manufacturing project, and they further delayed their Ohio fab plans as part of a cost cutting consolidation. So really, it feels like NVIDIA is on a tear, but they're worried about kind of export controls to China and trying to get their chips there as soon as possible. And so NVIDIA seems fine. They're kind of crushing it. Intel, on the other hand, while they had$12 billion in revenue, which is fantastic, they had$3 billion in net loss. Intel is not doing great. It doesn't feel like they really are one of the big winners in the AI race.

11:27It seems like NVIDIA has taken that spot, and they really could have been. So it's going to be interesting to see how this plays out in the future. In any case, I'll keep you up today on everything going on in the chip fabrication industry with semiconductors. This is really important because basically it's a backbone of AI working or not working and whoever controls the semiconductors is going to control AI. So very, very interesting. Thank you so much for tuning into the podcast today. And like I mentioned before, if you want to try any of these top AI models that I have mentioned, including DeepSeek from China and all the latest from OpenAI, go check out AIbox.ai I have the top 40 AI models on there and you can try all of them for one monthly subscription of 20 bucks.

12:08We also have image, audio, tons of other cool stuff. So go check it out. The link is in the description. All right, thanks so much and I'll catch you next time.