---
type: raw-transcript
source_type: youtube
category: raw
domain: ai-infrastructure
status: captured
captured: 2026-05-12
published: 2026-05-05
confidence: medium
tags:
  - memory
  - micron
  - hbm
  - inference
  - ai-infrastructure
  - memory-wall
  - ssd
  - power-efficiency
source: https://www.youtube.com/watch?v=IRhuHt51AXg
x_source: https://x.com/t_h_e_circuit/status/2051659983415296192
channel: The Circuit
duration: "52:31"
---

# Breaking the Memory Wall: Micron’s Strategy for the AI Era - Transcript

Source: https://www.youtube.com/watch?v=IRhuHt51AXg  
X source: https://x.com/t_h_e_circuit/status/2051659983415296192  
Channel: The Circuit  
Published: 2026-05-05  
Duration: 52:31

> Transcript captured from YouTube subtitles via yt-dlp on 2026-05-12. Cleaned lightly for readability; may contain subtitle errors.

## Transcript

If ultimately the amount of power that we have is limiting the growth, then we have to find a way at the data center level to deliver much more efficient performance in a fixed power budget.
And that's where so much of our Hello everyone, welcome to another episode of the circuit.
I am Ben Beharin.
>> Hello world.
I'm Jay Goldberg.
>> Well, we have an excellent guest today.
We have Jeremy Wernern from Micron and everybody has constantly asked us to have more folks talk memory on this podcast and so your wishes and demands are fulfilled.
Jeremy, thanks for coming on.
just give a brief uh uh highlight and quick intro on yourself and we'll jump into a discussion.
>> That's a lot of lot of pressure if I have to carry everyone's expectations here today.
I'm Jeremy Warner.
>> I lead the core data center business unit at Micron.
So we deliver SSDs and DRAM products to data centers all around the world.
[snorts] >> A company I think our uh our listeners are are following uh closely.
Um, so, uh, so there we go.
Um, A Jay's going to kick us off with the first, uh, sort of question and then let's, uh, we'll roll with it.
>> So, I I've heard memories doing well.
>> Not bad.
>> Um, >> can always do better.
Well, so that's that's my question is like memory is famously stereotypically hugely hugely cyclical and this seems like you all just have to wake up every morning and pinch yourselves and go is this really happening?
So, and you you've been in memory a long time.
Can you help us think through compare this cycle to past?
Is is this one is this one different?
How is it different?
>> Well, there's a lot of interesting things going on.
I'm I'm a little scared to give you the answer after um the way you talked about TSMC's earnings call.
Uh but it is different because of AI is driving really a change in the way that memory adds value to the data center.
uh and I I'm sure we'll talk a lot more about that today, but it memory has become the key strategic asset in breaking the bottleneck for enabling inference in the data center and also uh enabling the training of the world's most advanced models.
So um it's a really incredible time for for memory and I don't see that trend slowing down.
>> [snorts] >> I'm I'm curious like do you guys like were you guys surprised by this or was it like look you know we know we know what we do at some point like is going to unlock or you saw like AI and deep learning early on because I feel like a lot of well obviously a lot of people didn't necessarily see AI coming but especially like the memory side of this like why like so I'm just curious like was there yeah we kind of knew this or was it we caught this early on and like what was sort of just the context on how this how this evolved.
>> Well, I think the whole world was a little bit surprised right when the model got advanced enough to really be able the LLM model got released and everyone figured out what the capabilities were.
So along with everyone I think we were uh surprised by the capabilities of the technology taking such a incredible leap uh once compute and training capability got kind of over the hump.
Uh but once the once we saw that and I think you know chat GPT their their launch uh and seeing those capabilities really I think woke the world up a few years ago to what was happening.
Um, but for sure I don't think that we and I I'm sure we didn't anticipate just how explosive the demand growth would be.
So we knew the tech technical details of how important memory and storage were going to be, but we didn't anticipate how fast going to grow.
>> Yeah.
I mean, Jensen has actually credited your CEO for being an early advocate.
You think you guys, so you guys probably knew about it before.
Is that safe to say?
Yeah, we we knew early on and we've been working on a lot of technologies that support AI from closest to the GPU to farther from the GPU whether it's our HBM LP DDR5 and the SOCAMS we're talking about are high performance SSDs our high-capacity SSDs.
So these products take 3 to 5 years plus to be able to develop and productize.
So it's not like we woke up one day and decided, hey, let's just throw this thing together.
Uh we heard AI had a good idea.
And I think Sanjay for a long time has done an amazing job of building an infrastructure that allows us to have a long-term vision, plan for the future, explore technologies, build huge scale operations, execute on our road map, work deeply with our customers.
And you know, all of those things come together to enable Micron to enjoy a lot of success in terms of helping to enable uh an incredible time.
And I think, you know, one of the things I keep, you know, trying to bring back into the the context of this conversation of of of what is different is, you know, not just TSMC, right?
TSMC has sort of said this all the foundaries but but you guys as well and other players in memory are are investing in capacity increases that you just would not do if the belief is that this just returns to deep cyclicality and even just the numbers right as you know Jay and I have sort of you know constantly talked about on the on the podcast right if this if this if semiconductor industry does go to let's just say it's trillion dollar mark this year or or next year which will probably happen for sure and then probably to two trillion very quickly after that.
It doesn't just go back down to a $700 billion market and then up to like the floor has changed, the shape of the industry has changed.
Those dollars are now larger in TAM and that's what everybody is planning right scaling and and growing into because the belief is that that is a a sustainable right AI is a sustainable innovation for the entirety of the industry.
Yeah, it's a sustainable innovation and I I truly believe we're just scratching the surface of where we're going with AI, right?
So for the last few years, training these models to be more complex was really the biggest driver of the infrastructure buildout in the data center.
Uh, of course there were a lot of users using AI, people using it in interesting ways at work, you know, help me answer this question faster or um, edit my document.
But now, I mean, just in the last six months, I've had to completely reform what's possible and how fast in terms of using these models.
I mean, Jay, I was listening to your podcast, uh, and you were talking about running 20 cloud agents in parallel, right?
Or only five when you're on the show, but 20 normally.
>> Yeah.
I Yeah, right now I have zero, but um, it's early, but >> you ran out of your token budget or something.
>> I was I was up late last night.
I had a lot running.
>> Okay.
>> So, yeah.
I I mean now with the introduction of Agentic AI um individuals are using the inference companies are just starting to figure out how to use it and what's possible with it and physical AI hasn't even begun really in a huge meaningful way and that is that's a road map that paves many years in the future of incred incredible expansion and transformation of human society.
I mean, the world is going to look totally different in 10 or 20 years.
I don't know, maybe it'll be like an Asimov novel.
Um, you know, I think if you want to look at the future, look at the past.
Look at what look at what he wrote.
And I think, you know, obviously I I don't think that we're going to have a robot living in the middle of the moon, but um it it it does it is um precient in terms of what's possible from autonomous vehicles, flying autonomous vehicles, robots that do everything, incredibly automated production and manufacturing.
I mean, we just have to the capabilities of this technology.
Uh, we're just at the beginning of scratching the surf surface of the transformation that's going to happen in the world.
>> Yeah, absolutely.
All of that is going to drive a lot of demand for memory and it's going to generate an incredible amount of data that needs to be accessed rapidly which bodess well for our SSD products as well.
>> Yes.
All right.
So, let's get into the to the inference point you made, right?
And I think this is probably the best observation and framing for sort of what's happened, right?
We're coming out of really the training era and all of the infrastructure.
silicon decisions architectures around AS6 and GPUs were largely built right with the training era in mind and now we're seeing this disagregate to the inference era good examples right is that you know even Nvidia is starting to talk about inference specific products with the gro lpus uh you know Google launched with TPUs a dedicated inference as all all things we think will happen you will see training architecture designs and you'll see inference ones but the inference one has a very different memory memory problem like there's a memory wall as Google calls it.
So, so maybe walk us through the workload and and why these inference AS6 or inference accelerators are are are a different demand on memory and how that could just sort of now play out for for the memory industry.
>> Sure.
Sure.
So training uses memory to learn and then forget and then it spits out a model.
But inference uses memory to remember.
And inference you can very broadly think of in two kind of key stages.
One is this prefill stage and that's where all the prompt tokens are processed and then you have a decode stage.
The reality is a little more complex than that.
There's a number of decode stages, but in the decode stage, each token is iterated um to get to a better answer.
And in the decode stage, everything that happened in the future, all the knowledge, you really want to be able uh I'm sorry, in the past, you really want to be able to uh feed that in to get the best result, the best answer.
And that's where a lot of the intelligence comes from.
So um in the decode stage that's where of inference that's where the memory wall comes in because there's really kind of two ways to do this.
Okay.
And so until now with the traditional architectures um there's something called KV cache.
KV cache is more like a concept which is as you go through the decode stage you calculate your tokens you save the KV cache in memory and then you feed it back in and then you calculate the token again and you iterate in that way many times.
Now, the longer the context window, uh, and for for those not familiar with context window, you can think of it as like the more information that you've been putting into your question, think of it the length of your chat, for instance, um, the more iterations that you need.
And so if you don't have enough memory to store what happened in the past, then you have to start at the beginning and recomputee everything.
And so if you think of every cycle and recomputing, what that means is that you get an exponential growth in the amount of compute that you need to go through.
In other words, every cycle takes as much compute as all the the prior cycles combined effectively.
Whereas if you're able to save the state from the last one, each cycle just is it's linear one more.
So if you don't have enough memory and to store enough context, then you're taking a square times as much compute.
And so having enough memory that's fast enough to feed the compute and to store enough context.
It's not just context, it's also the size of the models, the number of parameters has grown.
So it's gotten and that's important for having more intelligent models and the more parameters, the more tokens uh per iteration and then the number of concurrent users agentic AI running per GPU has grown.
So all of that drives an incredible amount of KV cache needed per GPU in order to avoid having to go back and recomputee everything.
And if you're successful at delivering enough memory and storage, then in theory, you can get n squared the amount of compute out of a GPU than you otherwise would if you were Maybe maybe can we take a step a step back because we have a fair number of non-memory non-technical listeners.
>> Sure.
>> Because I totally understand it.
It's not that I'm completely ignorant about memory.
But maybe for those people, could you just walk us through the sort of the the memory hierarchy at large?
>> Sure.
>> There's lots of different kinds.
There's lots of acronyms.
It >> there's a lot going on there.
So, um, the GPU or TPU or XPU or whatever you want to call it that does the AI calculations, very close to that is the memory that's probably gotten the most press.
So that's called HBM or high bandwidth memory and it's used for training but also for token generation in the inference stack.
So the typical amount of KV cache that's stored very close to the GPU in HBM is maybe 10 to 100 gigabytes.
If you don't have enough HBM to store everything you need, then the KV cache moves a little bit further away from the GPU.
So then it goes out to what we would call main memory.
Main memory typically is attached to a CPU.
So that could be in like an H100 system.
It was often times attached to an Intel or AMD x86 CPU.
If you look at something like from Nvidia now with great Blackwell, that main memory is attached to race the CPU and the amount of the size of that memory might be anywhere from 4 to 20 times the size of the HPM that you have on the GPU.
So you get more memory there, but it's slower.
It's further away.
As you go down the stack beyond that to date, that's pretty much where KV cache has stopped for inference.
If you don't have enough memory there, then you start recomputing.
But now that inference is getting so complicated um and the context window is growing, people are looking for ways to expand that memory footprint.
And so then moving down the stack, you have a number of uh established or new ideas that are coming into play.
So right after main memory, there's a concept called expansion memory.
This hasn't been deployed really in any meaningful way in production yet.
But the idea is that you would take a lot of memory typically let's say very high capacity DIM modules and you would connect them maybe through optics in a separate box connected to all the GPUs.
And so if you run out of main memory you can go there for some very fast storage.
If you don't have enough there, and this is one of the most exciting parts that um Jensen talked about this year, then you go into what's called context memory storage and that's where people start using SSDs to store more.
So now when you get to context memory storage, you might be your your latency is longer, your bandwidth is less, but you're getting a thousand times the amount of capacity that you have say in HBM.
M >> and then finally at the very very bottom you just have exabytes of network data lakes in the data center of these massive SSDs.
So there's this whole hierarchy of where we're going with this.
>> And and so when we see people today complaining or you know up against constraints, what are we talking about today where we're seeing the most the the tightest of all this?
like what's really the bottleneck at this moment >> from a technology perspective or from a production perspective or what what's >> we well yeah where where is the pain most acute today >> um I mean both DRAM and SSD all through the stack I think you see um people want more people need more as soon as we can release product, they'll consume it.
As soon as we increase the capacity and the performance of the products, they'll find a way to deploy it.
So, um, >> everywhere.
>> It truly is.
>> Okay.
So the and the thing I think that you know boggles my mind in all of this right just coming from training and now moving to inference is you know in a training solution if you're a frontier lab or you're you know a hyperscaler or somebody it's a small number of people who are just throwing training workloads at x you know x y and z rack of compute right and to your point like that happened very distinctly in memory the goal was just train so that I've got an output that I can then go serve.
Where this like really boggles my mind and I'm sort of setting this up for like how does how does what specifics of this memory hierarchy help solve this problem is we're about to move to tens of millions eventually hundreds of millions and then billions of people attempting to use context windows at the exact same time.
Like we're talking many many people simultaneously needing massive context windows to go do the things they're going to do.
And I get like this is why like with TPU AI right we're building very different architectures to do this but to me it's like okay I've got you know a rack full of chips and there's this memory and then as you you know pointed out we might have memory appliances.
Sometimes I just feel like though like even that feels hard to solve this problem where you know tens of millions billions and you know of people literally are just going to need everything they're interacting with to have this crazy amount of memory about what they're doing.
So just like where is like what helps solve that in in the memory hierarchy or what is the demand on that kind of simultaneous hitting of all of these people for for their workloads?
So I mean speed right uh I if if the bottleneck is not flops and it's really memory bandwidth then we need to increase the bandwidth that all of that history can make it to the GPU.
So I a lot of it comes down to speed.
That's why we've been innovating really rapidly in for instance our HBM products.
Uh we just announced the HBM4 product.
Uh it's two over two times the bandwidth of our HBM 3e product which was cutting edge a year before.
Um and of course like when you deliver all that performance you do start to run into other bottlenecks especially at the data center level and this is where I think most people's heads [snorts] have been at for a long time on the AI roll out which is power.
So >> uh >> can you get enough power to drive all this uh compute capacity and can you make the most out of the compute capacity?
Can you make the most out of the power?
So if you deliver twice as much performance but you use twice as much power and you have fixed amount of power then you don't really deliver more to users, right?
So really what it's about is delivering if ultimately the amount of power that we have is limiting the growth then we have to find a way at the data center level to deliver much more efficient performance in a fixed power budget and that's where so much of our innovation comes in.
So it's delivering more performance while finding ways to do it without scaling the power consumption equivalently.
And you know it's always look in memory people talk about bandwidth a lot just kind of like how much data is moving at what rate.
But the reality is that it it's it's getting more complex than that.
And that's where there's some really interesting dynamic shifts that are happening in the stack with Micron because in order to become really truly more efficient in your power use, you have to look at into the depths of what's happening and how these how the inference is being processed.
And when you look deeply into it, there's it drives a significant amount of code design requirement and opportunity or where do you do certain things in the GPU on the memory side um and at different layers of the stack.
So that is really um a necessity for driving innovation in an area that is one of the other kind of transformative elements of the memory industry at this time.
>> Let me let me ask a question some something that sort of piqu my interest when we were preparing for this earlier and you you've touched on it here.
I I get the demand for memory, right?
Things closer to the GPU, as much bandwidth as possible for the calculation, but it sounds like there's at large, like not just on the compute, not on the GPU tray, but sort of across the whole data center ecosystem.
Why why are we seeing that?
Sure.
So um AI it does a few things.
First of all it generates a lot of data itself.
Uh I anyone who uses Grock or is on X will know how much image generation is happening with AI much faster than you know even the most adept meme creator would be able to create images and all of that data.
I, you know, I always like to say that um most people are digital hoarders.
We tend not to throw our bits away.
>> Guilty, >> right?
Oh, maybe I'm >> charged >> in 10 years.
Uh, and the interesting thing is that not only is all of these AI models enabling all of us who maybe didn't have the physical capability to bring our creative ideas to fruition, but now we have these inc this incredibly powerful tool to make our thoughts reality or at least digital reality.
um it it makes so many it is a creative revolution for people.
So you we're creating a lot more data and then businesses as well are able to >> create and use their data.
And one of the most important things for getting the most out of AI is actually having all of your data in a place that's accept accessible.
And so, you know, AI, it doesn't just create a lot of data, but it it accesses that data to provide insights, solve problems, give better answers like you said, Jay.
Uh, and so what that means when you need to access all that data is we call it warming.
Warming of the data.
So, you know, we have this term in storage.
So, sometimes we call some data hot and some data cold.
And hot just means someone's likely to come look at this thing soon.
And cold is uh you're hopefully your tax returns from 10 years ago.
No one wants to see that, right?
Uh so but with AI, you ask it a question, it's looking through everything to try to get an answer.
So the things that used to be cold, they're getting warmer.
Everything's getting warmer.
And when things get warmer, then you need a lot faster storage because you're accessing >> Yeah.
All this >> that that's the past and there is kind of this exciting future growth as well, which is we're out of we don't have enough memory for all the KV cache.
And so now there's a big future growth of SSDs in the data center to store query scheduling and all the multi-turn workflows that in the past if we tried to do it with today's architecture you'd be you'd have to go through that recomputee cycle I talked about.
>> Yeah.
I I I will I I'll note that like when in in a lot of a lot of the AI stuff I'm playing with, I'm spending a lot of time reading up on different harnesses, right?
Do I use, you know, Open Claw, there's Hermes, there's all these new things coming out all the time.
Um, and a big a big part of that is load balancing across models, but an even more important part of it is creating persistent memory for the AI.
Because one of the big problems now is you use an AI agent and it doesn't ne it doesn't remember from session to session what you're doing.
>> Does that wouldn't that drive you crazy?
It's like teaching a class of students and every day you have to start from the first page of the textbook.
>> Yeah.
Yeah.
And and so every so many of the tools out there that you know people are saying, "Oh, you need to use this, you need to use that." It it all comes down to creating a file structure that gives your your AI agent memory.
They're all hacks around that.
It's it's super interesting.
Um it >> and so so what'll happen is, you know, if you're giving it memory, then your history is sitting there in your memory.
If you go away and do something for a week, either you have to leave it there and if you do a lot of things, then you're gonna run out or you can send it down to SSD and then bring it back when you pick back up on the task.
>> Yeah.
Yeah.
Yeah.
So I have like multiple tasks and projects in in the sidebar of claude co-work say and I'm very deliberate now of like oh I'm doing this thing I need to switch to the right project or the right context window because I don't want to blend them together and all those things have to be stored discreetly and then I'm worried about like each one of those at some point is going to max out the context window and constantly trying to juggle that around like people and I have conversations with people like I've all this week I've been hearing oh you got to use this this tool or that tool or superpowers or whatever and I think what what will where this will lead is someone is going to create a harness that manages this a little bit more uh less bespoke to sort of automatically manages it and everyone will love that.
It'll be fantastic because it'll be so much simpler to use, but it's going to come at the cost of lots of extra memory, right?
Because because right now I can sort of manage it on a bespoke basis and it's I'm very efficient, but I very very gladly give give up that bespokeness just to have something that manages it and I know that will come at the cost of me meeting needing a lot more memory.
It'll be so our what we find is the context length what you're talking about is [snorts] >> Geez.
The you know the other >> sounds like that sounds like a challenge.
the the the other the other thing that like boggles my mind about like like this conversation we're having is right now like my usage with AI and let's say that's right a chat with with with chat GPT or with claude is really fragmented memory with fragmented storage like it's only looking at a folder it you know what what's to your terminology like what's warm and what's cold like it's it's context related like I don't have the entirety of my data available to me and I recently learned that I can't live on a computer without a tire of storage because that's how much stuff I have on my device and it's mostly cold, right, in an AI context.
Like I would love to be able to just be like, "Hey, what was this?
Is this this go find it?" But it just can't do that yet.
And so like you said when enterprises bring all of their data right online both with access to the edge like this feels like a even another step function of AI's value when you really do have all of that data available to you which is again simultaneously a memory and a storage problem that we still have to solve a lot of problems for >> we do so uh along those lines for enterprise is scale data.
Uh and talking One of the new products that we've just announced is a 245 terabte SSD.
This thing is barely bigger than a deck of cards.
So when we introduce a product like that, we're able to massively bring down the amount of power that's going into storage at the data center enterprise scale and also reduce the footprint of that storage by over 80%.
So that al you know with one of the big bottlenecks being power with one of the big bottlenecks being the square footage of data centers >> right >> getting data closer to the GPU.
That is one of the big future trends.
Unlocking the capabilities of the data center, giving more performance, more storage, closer to the GPU in a lower power footprint.
So, that's one of the things I'm really excited about um in kind of file and object storage.
M on the PowerPoint.
Is that just an innovation that happened in storage or is it because you can put so much more in one footprint versus to having to divide that up across two or three different spots?
Like maybe just talk a little bit about like what goes on there from a power stand because I think that's it's the most important point, right?
And you look at what I think everybody has sort of consensual consensually bought into with Jensen is essentially like if you have a power budget, you want the most compute and this includes all the infrastructure we're talking about within that.
So obviously any innovation in in bringing down cost for power power matters.
So maybe just talk a little bit about like what is it exactly about this that's leading to some of that better power.
>> Yeah, sure.
So SSDs are inherently a lot more power efficient in terms of delivering performance than a hard drive.
Like >> right >> from a read perspective, you're getting >> you can get a thousand times the read performance depending on your workload.
So there's this inherent there's no moving part in an SSD power benefit just right off the bat.
But then another huge benefit when you think about the power consumption is what it takes to deliver all that capacity in the data center.
So like I said 245 terabytes in a a space that's about a quarter of a hard drive.
And the hard drive capacities that are being deployed today are in the low 30 terabyte range.
So that means that you need a lot less networking.
You need a lot less connectivity.
You need a lot less uh boxes and power supplies and fans and everything that goes around deploying >> has a real cost has a real power consumption.
So, you know, you just eliminate all that waste, consolidate down, and now you're just paying for the performance that you need.
And that performance is delivered at a at a much more efficient um gigabyte per watt.
>> Yeah, I like gigabyte per watt actually as a as a model framing um to the analysis.
That's that's interesting.
So So you guys have done this like let's talk a little bit about more of sort of the innovations that um that you guys are working on across the stack and and and I want to sort of look at this like two ways.
Well, one as we talked a little about historically like do do you do you feel the cadence in innovation for memory like there's pressure to increase that because I'd always been under the expect understanding and you can correct me if I'm wrong this historically that memory was historically and maybe to some degree storage a little bit more conservative in pushing the limits right because so much had to go right be consistent you couldn't have failure rates right on wafers Now, that doesn't go away.
I'm not saying that that all of a sudden you become logic and you're like, "Sure, we're happy with 50% yield." Like, no, that doesn't happen.
But I but I feel like like this is one of those moments where like it put like we feel we need to put the our foot on the gas.
And I'm just curious like as you guys think about this and areas where you can go solve problems and and again like coming from what I think must be also super interesting is like we're in a period of time where to be an engineer right especially at a company that's always is one of those like the most like there's problems all over the place like just a giant map board of problems that we have to go stall particularly too for for memory and storage.
So just how do you guys think about right that the problem at large accelerating innovation and just some of those areas you guys are focusing on uh really across the whole stack.
>> It is accelerating.
Our timelines are accelerating.
The pace that we need to innovate is accelerating.
The amount of intelligence we need to build into our products is accelerating.
The complexity is accelerating.
the rate that we're building mega fabs is accelerating.
Uh, and it's it's a real challenge, but it's exciting.
I mean, who doesn't want to come to work every day and at the end of the day look at the clock and say, "Wow, like I barely breathe today." Um but and we're running at an incredible pace and it's I think it's fun and it's energizing for people at Micron and honestly one of the things that we're doing is embracing the AI technology.
So how do we go faster?
Well, there's this incredible tool that's been I don't want to say dropped in our laps because we helped build it, uh, but that we've helped create had a meaningful role in creating this incredible technology and we're using it to accelerate what we can do and run faster, improve yields faster, design and develop faster, find issues faster, you know.
So, all those traditional problems that uh we've faced, they still exist.
Now, we just have to do everything faster, better, more efficiently, turn chips faster, um go through process technology faster, ramp and install tools and more fabs around the world faster, faster, faster, faster.
And you know, there's only seven days a week.
So, at some point, um, you have to find ways to innovate.
And we've always done it.
Now, AI is an amazing tool that I [snorts] think we're doing some really interesting things with to take us to the next level.
>> Yeah.
And on that point, I imagine like you you did talk about the the code design that's happening.
I can imagine like customers are like, "Guys, we need you to do this.
Like, can you do this?" and you're like like just engaged in helping solve right their problems two three years away in kind of that deep co-optimization with both memory and storage and and Micron alongside their partners there.
Yeah, the depth of engagement.
We've always had a really solid deep technical engagement with our customers.
And one of the interesting things about we what we do is we kind of have to work with everyone from software providers to CPU and GPU hardware designers, the fab technology, the um system builders, the data center builders.
So, we get to work with everybody up and down the stack.
But the depth of the engagement with what's happened now is beyond anything that we've ever seen before.
And so that deep deep co-design is really something that I think you know coming back to the beginning of the the discussion is different this time for memory.
>> So that's a lot.
That's a lot.
What What do you think people the market in general?
>> You guys don't have memory guys on, right?
We go too deep or >> No, no, no.
This is perfect.
This is perfect.
>> What What are we still missing?
Like what do you think the market still misunderstands about what's what's happening?
people are seeing the increase in capex from the big CSPs and other data center um companies and they're they're worried about um where going into ustainable and I think that those businesses are also going through a tremendous revolution that you know I mentioned earlier is going to transform human society in so many ways going to solve so many problems people without access to medicine or surgeons are going to be able to get it people uh we're going innovate at a much faster rate.
We're going to be able to automate, you know, production that's going to raise the quality of life for [snorts] billions of people around the globe.
So not all is obvious necessarily at all times of exactly when and how does the monetization of these models translate into revenue streams and you know are they spending more than their capability and I would say no.
I think the the potential is still greater than what most people can imagine.
And it's funny because in Silicon Valley, I feel people are so excited about this.
It's easy to get in your kind of bubble here, >> right?
where everyone really kind of probably understands the deep depth of the technology and they hear about all these cool things that people are working on.
Um, when I talk to a lot of my friends who aren't necessarily in the industry or in other industries, you know, a few of them are saying, "Wow, I can completely do incredible things in my industry in AI." And they are innovating.
I have a friend who's doing some incredible stuff um in an industry I don't want I can't say too much about it but um other friends they you know they see what's happening on chat GPT LLM um they're not necessarily they see what happened on the stock market they're they're not in tune with the potential of what's going to happen over the next 20 years and I fully believe it's going to happen.
>> So, so let me let me ask the the flip side of that question.
We I think we we would all agree there's something here broad so the broad number of people in the world don't quite see it yet.
Um so that implies that over time they'll they'll catch up.
They'll start getting more informed.
They'll realize chat GG chat GBT is more than just like coming up with silly memes.
you can do real work with it.
Demand keeps accelerating, but in the near term, are we gonna be able to keep up?
Is, let me be more specific.
Is the memory industry gonna be able to keep up?
Or do we get to a point somewhere next year where we're just like, that's it, guys.
We're we've built everything we could build.
The new fab isn't ready yet.
It's just just hang out for six months until we get up we get there.
>> Well, we're already there from a keep up from a production perspective.
Yeah, unfortunately um we didn't build enough fab in the world.
And turns out building fabs is is not the easiest thing.
I'll give you some context.
You know, we're we're building um five fabs around the world right now.
We've announced uh in Boise, Idaho, we have are building a 600,000 square foot clean room.
We've announced construction in upstate New York.
So, you know, we're bringing memory to the US uh augmenting our memory fab in Virginia.
And that's going to make the US a huge producer of memory.
And each of those abs in Boise and New York and we have, you know, plans beyond the first one.
As soon as we finish one, we start on the next one.
Each of those is the size of 10 football fields.
That's how big the factory floor is, just to give you some context.
Uh, and then we're building fabs.
We've announced uh groundbreaking on the Nanfab in Singapore to expand our production there.
We've announced um expansion of our DRAM facilities as well in Japan.
We just bought um a fab from PSMC in Taiwan.
So, we're doing a lot of construction to bring things online.
So, right now, um, the industry is construction, uh, clean room space limited, and it'll probably be that way for a while.
When we talk about being able to meet demand, that is um, the biggest uh, challenge that we have.
We do a lot of other things to push out more technology running more efficiently.
The we move to newer technology that gives us more bits per square foot of fab.
But ultimately um we've already failed to keep up keep up with demand >> a a as everybody else has, right?
You've got, you know, Intel on earnings, you've got, you got Nvidia, you got TSMC going like, "Look, guys, like we're at we're at capacity." Like fabs don't grow on trees like you like you just pointed out.
And uh but I you know, I think just just wrapping on the whole conversation as a whole, like what's interesting is well, one how much has changed, right?
So like even a year ago, we weren't having these same conversations and there's a good chance that a year from now, right?
All of this might be very very different.
None of it changes the demands on compute, but all of these things that we're going to solve problems for.
And sort of even just coming back to like who's adopting, right, AI, it's just going to get more capable because of more compute, more memory, more storage.
You're going to be able to do a whole lot more, right, than you I like this.
I forget who said this, but somebody, you know, I like on Twitter said like that what you're using today is the dumbest AI is ever going to be.
And it feels pretty darn smart, you know?
So it's like all of this and like you know problems you guys are solving are part of what makes this more useful, more valuable and the economics come back to the hyperscalers like they monetize it as it gets better, right?
As more people can use it and find value and and and memory and storage is is a huge part of that.
>> Yeah.
>> Well said.
>> Thank thank you.
Well, Jeremy, we we we really appreciate your time and for for coming on the circuit.
You're a welcomed guest anytime you want to come in and and talk memory.
Um, thanks for listening everybody.
We hope you uh you found this insightful and we will talk to you next week.
>> Thank you everybody.
Tell your friends
