# AI Hardware Pivots and Semiconductor Supply Chain Shifts

**Podcast:** The AI Engineer Podcast
**Published:** 2026-06-19

## Transcript

Welcome to the podcast today.
Midjourney is pivoting to hardware.
They've created a full-body ultrasound scanner.
This is one of the biggest AI pivots I think I have seen since the start of the AI wave.
I'm excited to get into that.
Anthropics' federal ban is becoming a bit of a marketing asset for it.
Maybe this isn't such a bad thing.
At the same time, Amazon's MGM, they are dropping Luca Guadrino's Sam Altman film called Artificial.
We're going to tell you a little bit about the drama behind why that's happening.
And at the same time, AWS is...
in talks right now to sell their Tranium chips to outside buyers.
So before it was kind of just inside of AWS.
Now they're going to be selling them to outside buyers.
This is directly competing with NVIDIA.
Intel, another person in the chip space, has jumped 9%.
Their stock is up after Trump announced that Apple had a chip design deal with them.
We're going to get into all of the details with that.
If you want to empower Claude to generate images, audio, and video, I'd love for you to try my platform out, which is AIbox.ai.
We have a new MCP.
So you grab a little link from us.
Once you create an account, you give it to Claude in the connector section.
You log into AIbox.
And after that...
Any tool that you build over on AI box, whether that's an image generator, an audio generator, a video generator, you can call it inside of Claude.
So you can say, Hey, Claude, use AI box to generate an image of XYZ thing.
It's going to go find the correct box that you've created, the correct tool you've created.
And it's going to actually use Claude to generate.
to go use that and generate images.
You get the link will pop up right inside of your Claude chat to the image that was generated.
You can go download that and use it in any project or you can tell Claude to, you know, go and just do whatever you need it to do.
So it's completely hands-free.
This is something that has helped me a ton.
If you've, you know, if you've been using Claude lately, you know that it's incredibly capable.
It can control your computer screen.
It can do anything, but it cannot generate.
images, audio, and video like other competitors can.
So it's not just those three capabilities.
You also get access to over 80 different AI models and you can use that same MCP inside of chat GPT.
You can use it inside of Gemini, all of the different models and agents that you use.
You can port this full library of AI models into them with the AI box MCP.
So it's $7 and or $8 and 99 cents a month.
There's a link in the description.
You can go check that out if you want AI box dot AI slash MCP.
I would love for you to try it.
I hope this unlocks a ton of incredible creativity for you.
Let's continue with the episode.
So let's kick it off with MidJourney.
They have launched their very first hardware product.
This is something I wasn't expecting when I opened up Twitter yesterday or X yesterday.
It is a full body ultrasound scanner.
It's going to capture detailed images in 60 seconds.
So it's using...
40 semiconductor transducers and two petaflops of compute.
The company is planning to open a spa in San Francisco's Union Square that is going to have 100 scanners.
It's all going to be next year, 2027.
And they are positioning this as a preventative health tool that could rival MRI quality without radiation or hospital logistics.
Okay, this is just like...
Coming from the company that was previously doing AI generated images, absolutely not something I was expecting.
But if you go look up like pictures of this thing, it's incredible.
There's all of these tiny microscopic speakers basically that are blasting little sounds at you.
In order to get this full body scan, you actually have to go into a vat of water.
So you walk into a vat of water, you're surrounded by all these tiny little speakers that are blasting at you, and it can get every single nanoparticle of your whole body scanned.
And it's...
pretty impressive because it's not just externally, it's internally.
So it's, it's able to scan your lungs, all of your, um, all of your organs.
It's really, really incredible.
And you can do the whole thing in 60 seconds.
So if you've done an MRI scan before, you know, you're, you're super claustrophobic.
You can't move.
You're in this giant tube.
You have to wear, uh, you know, a lead vest for radiation.
This is very cool.
So the way this whole thing works is that AI is going to handle the 3d reconstruction of muscle, fat, bone, and organs.
So basically, It's going to be fed back all of these basically like bat echolocation, sonar, sound bounce detection that it's done.
And then the AI is going to put that back together and decode this into a way that shows a full 3D reconstruction of your whole body.
A dozen people have been scanned so far.
Midjourney hasn't published any kind of comparative imaging data against MRI baselines.
And the company is actually not going to go do FDA diagnostic clearance.
They're actually going to start with body composition maps.
This is a much lower regulatory category.
They're also kind of floating the idea of a new FDA device class for consumer health tracking.
So it's not like something that's like, hey, go in this thing and it's going to diagnose if you have any issues.
They're more like, hey, this is for preventative.
This is for giving you information.
It's kind of just like a body tracker just because FDA takes so long.
But this is really incredible.
Basically, Majornine is betting that this kind of real-time full-body imaging is going to become a consumer wellness staple rather than a clinical tool.
The way that they're thinking about this is that gym memberships are going to be able to put this in, spas, etc.
And their goal is to build 50,000 of these that they put around the world.
And because you can do the whole scan in 60 seconds instead of, you know, hours in an MRI scan, their goal is that the whole world could be able to get full body, high quality.
scans, internal, external, in 60 seconds with just 50,000 of these.
So very cool, very bullish, you know, huge congratulations to the whole team over at MidJourney for doing something so focused on health that is so productive to humanity beyond just generating AI images, which is what they're famous for.
Okay, let's talk about Anthropics Federal Ban.
Some people are saying this is becoming a bit of a marketing asset.
Because the US government forced them to pull their newest model, Fable 5 and Mythos 5, last week.
And this is all after Amazon researchers found that there was some jailbreak vulnerabilities.
And apparently Anthropic said, hey, look, why do we have to fix these jailbreaks?
Because you can do the same jailbreaks on chatGBT 5.5.
So they're like, why are you targeting us?
And there was kind of this whole back and forth.
And the government's like, well, there's a jailbreak.
And you said this was super dangerous.
So we got to pull it.
Now, some people think this is really bad for Anthropic, but over at TechCrunch, the hosts of their equity podcast said that this incident has positioned Anthropic at the center of AI policy.
And this debate that is happening this week was done without the company spending any marketing dollars.
They're saying, look, it's actually making Anthropic look like they're fighting the government.
And beyond that, it shows them as, number one, a strong company.
But if they're getting kind of signaled out...
by the government for what could be an industry-wide problem, that might actually strengthen their pre-IPO narrative because investors are saying, look, there's something different about Anthropoc.
They're stronger or they're better, and they're being targeted because their AI models are so much better than ChatGPT and anyone else's.
So we'll see how the chips fall with that, if it's actually benefiting their company or not.
Okay.
Amazon's MGM is dropping the new Sam Altman film Artificial.
This was created by Luca Guadino, and the film is basically dramatizing Sam Altman's five-day firing and reinstatement as opening eye CEO.
This all happened back in like November 2023.
I know it seems like a million years ago, but it wasn't that long ago.
They made a whole movie about it.
But Amazon doesn't want to distribute it anymore.
And so they're actually shopping for new distributors.
This is coming just a few weeks after Amazon announced a $50 billion investment in OpenAI, which is kind of a weird...
conflict between, you know, the studio's parent company and also depicting OpenAI's internal crisis.
So it's almost like, look, they're putting $50 billion into OpenAI.
Maybe they don't want to be the people that distribute a video that might not put Sam Altman in the best light.
So that's going to be interesting to see who else picks that up.
I think we all want to see the film.
Come on, we all have to live through the drama of him getting fired and wondering what was going to happen.
We all want to see the movie, whether it's good or not.
I love these types of movies.
I love the one they made about Uber.
I love the one they made about BlackBerry.
I love all of these kind of like Silicon Valley insider movies.
I know they're not all perfect.
The We Crashed one was phenomenal.
Highly recommend it.
Apparently, Andrew Garfield is playing Altman.
The thing that I'm happy about is that the film is far enough along that they're just shopping, sending this to a different distributor, and they're not canceling it out altogether.
It's basically near finished of a product.
So I think they just have some editing and stuff, but I think it's mostly all been filmed and shot, which I'm excited to see what it actually looks like.
Okay, AWS is in open talks, more Amazon-ish news, to sell Tranium chips outside of just being...
built inside of AWS.
So they're actually going and targeting outside buyers.
This is something that NVIDIA is probably feeling particularly threatened by.
The AWS CEO, Andy Jassy, he pegged the chip's business at about $50 billion annual run rate, which is enough to rival Intel's revenue and create the first credible alternative to NVIDIA for AI training in Silicon.
I mean, that's pretty wild for them to be like, hey, look, we have this new chip.
We've been using it internally and we think this thing is worth...
you know, the entire company of Intel.
Tranium 4 capability is already completely sold out, which is more than a year before this chip is actually shipping and they've completely sold it out a year ahead of time.
That I think has forced Amazon to choose between serving AWS customers or external buyers.
Amazon is going to need to secure basically additional capacity from TSMC, which is currently dominated by NVIDIA orders and has become Apple's replacement at the foundry's largest customer.
NVIDIA is simultaneously expanding into a 200 billion CPU for AI market, which is pushing into Intel and AMD territory.
So it's kind of interesting because we have Amazon moving into Nvidia's territory.
We have Nvidia moving into Intel and AMD's territory.
The reason why I think this matters and why it's particularly interesting to me.
is because right now we have kind of this hyperscaler supply crunch.
And in a way, it's actually forcing Amazon's hand because if they sell their Tranium externally, it's going to sacrifice this kind of a bundled cloud revenue that Amazon makes when it owns the full stack.
So they have this kind of bundled revenue when they're selling it inside of AWS.
But...
you know, for every single chip that they sell or every single chip that they ship, it's a five-year revenue stream lost, right?
Because if they'd taken this Tranium, they put it in their own data center and then they sold it to people, they would be able to keep using that chip for five years and selling it.
It's usage to people, kind of renting it out, right?
Instead, if they go and sell that chip to someone, it's a one-off deal.
But right now, the reason that Amazon is doing this is because if they don't do, you know, kind of sell it one off when there's this massive demand, they're basically leaving billions of dollars on the table because there's certain customers that are not going to rent servers from Amazon.
They're going to go buy them.
They're going to try to get them from NVIDIA, but they're going to try to purchase them for themselves one way or another.
So Amazon's like, look.
Let's try to make some money here.
They're kind of trying to bet that this is a good NVIDIA alternative, that people are going to be interested because this is backed by a $50 billion business.
And I think right now that is, you know, meaning a lot more to them than kind of the short-term cloud business bundle losses that they might be.
missing out on.
Okay, into more chip news.
Intel's stock has jumped 9% after Trump announced an Apple chips design deal that's going to be happening.
This, you know, bumped them up in pre-market trading.
The move, I think, shows that there is a major shift in semiconductor manufacturing, pulling Apple away from Taiwan.
And marking Intel's biggest foundry win since their CEO, the new one, LitButan, refocused the company on some outside partnerships.
So their stock has gained 464% over the last 12 months.
If you guys remember, Intel stock has been sliding for many, many years.
This is lifting their market cap.
to $608 billion a bid, a much broader 90% year-to-date rally in the semiconductor sector.
Elon Musk's TerraFab, also described as basically the world's largest chip factory, is being designed with Intel's technology team, which is the first major outside foundry commitment for Intel's very capital-intensive manufacturing business.
And Apple currently designs all of its own M-Series and A-Series chips, and they have a fabrication.
unit and all that is done in Taiwan, which means that any meaningful volume shift to Intel is going to be the company's most significant outsourced order to date.
The reason why I think this is a big deal is because Intel has gone from manufacturing and they've kind of been, they're turning into sort of a geopolitical asset in the last 12 months.
The Apple deal, if it includes real volume, is going to validate what the CEO has been betting on that US founder capacity.
which is, of course, backed by government capital and a lot of President Trump's attention.
They're saying that this can compete with Taiwan.
And so I think if we kind of watch whether this is going to open up the floodgates for other design wins or if this is going to kind of be like a one-off political kind of like moment where...
Trump just wanted to make a big announcement.
And so we kind of got this deal put together.
I think that's remaining to be seen, but overall I'm excited to see this American company succeed moving some manufacturing to the United States.
And I think this is great for diversifying our supply chain overall.
So we'll see what happens there.
Thank you so much for tuning into the podcast.
If you enjoyed this episode, make sure to leave a rating review wherever you get your podcasts.
And I hope you have a fantastic rest of your day.
I will leave a link in the description to AI box.ai.
If you want to add image, audio, and video models all into Claude, Gemini, and ChatGPT, all with the AI Box MCP.
There's a link in the description.
I would love to have you try it out and hope that unlocks a lot of capabilities for you for only $8.99 a month.
All right, catch you all in the next episode.