Nvidia, Siemens to Build Industrial AI Operating System

Jensen Huang
2 followers
4 Video Views·Mar 3, 2026
Nvidia and Siemens announced plans to build an industrial AI operating system together. CEOs Jensen Huang and Roland Busch speak to Ed Ludlow about this partnership, the need for energy, Nvidia's new chips, China and Siemen's potential deals in the operations software space. They sit down at the Consumer Electronics Show.
Over 175 years Siemens has been at the forefront of I'd say several industrial revolutions.
NVIDIA is at the forefront of the latest AI industrial revolution and it's a pleasure to
have you both here with us gentlemen. Roland I want to try and understand how real this is.
You call it an industrial AI operating system. We're going to talk about how you're joining
forces on the software side and the hardware side but I think the most useful place to start would
be could you outline a timeline for when Siemens goes beyond its own footprint around the world
using this operating system to customers actually doing things at scale because I think when you're
on stage that's the kind of sense I was getting. You want to accelerate to scale. Well I mean this
technology is already in place and it's working. I mean we see there are many examples and I brought
some on stage today where you have customers starting with a digital twin of a product
they want to manufacture a digital twin of the manufacturing side and bring that all together
and not before that you have it optimized then build in the real world and AI is already working
on the shop floor. They call it machine learning but the point is with the new models you can bring
that to the next level. That means you go away from giving an advice to somebody with technology
but really act on your behalf. This is when things go more autonomous or adaptive and you see that
already starting. The big thing is how can we scale it because it requires a lot of skills from
our customers, a lot of technology and it's still not that easy to implement. We are making it easy
to deploy and easy to use but you see picking up momentum in this each step we do and we saw some
great examples as a chip building is sure but even startups using our technology like
Jensen, we've discussed the five layer cake often. With NVIDIA it started with the
GPUs but it's now software, the element of simulation. On stage you talked about integrating
the software side in particular into EDA. Again a similar question for you, more of a timeline.
What is it you think you'll do first in how you guys work closer together? The first thing that
let me just say very quickly we're announcing a big partnership between us. We've known each
other for a long time but this the partnership we're announcing is really a big deal.
One, we're accelerating their EDA software. We're accelerating their simulation software.
We're integrating AI technology, physical AI and agentic AI into their team center and their factory
automation operating system and so we're working together across this entire spectrum. When we
software then we'll get to use it to design our chips and systems. When we accelerate their
simulation software we'll use it in our AI factories to simulate the thermal properties
of our AI factories. When we integrate our automation and agentic systems into their AI
industrial operating system we can then use it in our factory floors with our partners for example
Foxconn and so we're working across this entire spectrum together and we're going to put the
technology to use basically as soon as we can. What's the net effect for you Jensen? Is it
improves margins, efficient capital allocation? I know that might sound a bit dry but actually
right now that's the answer everyone's searching for. How is this investment in AI and use of the
technology actually change things in the real world for me? Well we announced yesterday Vera
Rubin. It takes six different chips to integrate into this incredible system called Vera Rubin
and when you're done each one of these Vera Rubin GPUs is 240,000 watts and it is 10 times more
energy efficient than the last generation. It is 10 times more cost efficient than the last
generation but it's still the technology is insanely complicated. 15,000 engineering years
came together to build this system and so when we accelerate EDA tools, when we accelerate
simulation tools and when we can eventually and I'm hoping very soon design entire Vera Rubin
systems inside a Siemens digital twin the chance the ability for us to create much much more
complex systems will scale. We'll do it much more efficiently and so this is really about being able
to do the impossible and being able to do it impossible the impossible right the first time.
And once we then realize that AI creates real world impact this is where it really deploys the
full power and also the economic power. It's not only in the data centers or in AI factories which
we see but also on the edge because once you start inferencing with low latency you bring this
technology to the edge. There's a huge potential for our customers too to deploy this technology
and this includes of course it includes hardware where we come from the jibs. It goes into the
controllers some of our controllers run on GPUs and then it goes all the way to the industrial PC.
Exactly. Now these industrial PCs are AI industrial PCs. We supercharge it and they can now run
algorithms trained in the cloud they can run it on the shop floor and do all that trick what we
talked about it in real-time optimization and running in a plant and that makes a huge difference.
It drives economic growth. The question that's being searched for is how is this going to manifest
in the real world you know the emphasis at this year's CES I think is physical AI. Is not the
manifestation of the the final stage of physical AI just one giant robot that you guys call a
in the manufacturing context is that where you're seeing demand from actual customers
you know that they need a factory that is automated 100 percent and genuinely autonomous
in some degree. So number one is if you want to build a factory and very often you're missing
out on labor and and we talk about the unskilled labor it's hard to find number one number two is
once you build it autonomous and and automated then you have a much much higher yield what you
can generate you use less energy by the way at the same time before you optimize it in a way
so therefore there's a lot of benefits and if you want to talk about United States ramp up
manufacturing United States you need to go as digital and as automated and AI supercharged as
possible. A factory is robotic and it is orchestrating robots that are building systems
that are also robotic like for example self-driving cars with robotic system
and the reason why it's so hard to deploy robots today is because it's hard to program these
robotic systems the the software expertise that's necessary the customization necessary is really
intense it's it's just too much and so the fact that we could now apply artificial intelligence
physical AI technology to these robotic systems make them easier to teach you show it a few
demonstrations and the AI learned it by itself. Would you go as far as say Jensen that you've
solved for that that software limitation? That software limitation is is the Chad GPT moment of
it is now here I think over the course of the next couple two three years we're going to make
some really big breakthroughs. Let's please talk about energy electricity power supply call it what
you will in turn for each of you how worried are you about it as a bottleneck and what is your
experience day to day and running both companies in in that respect? Energy should always be a
bottleneck for any industry and this is a new industry that's growing incredibly fast as you
know AI is both a technology that's going to revolutionize many applications and we're talking
about some of them here but the AI industry itself to take the manufacturing of the artificial
intelligence takes energy it takes energy it takes AI factories it's exactly the reason why
from Hopper to Blackwell we increased energy efficiency by 10x from Blackwell to Rubin we
increased energy efficiency again by 10x and that translates directly to our customers revenues
because in the case of an AI factory whatever factory size you have you're limited by the power
and that power within that power constraint you want to have the most tokens or most AI per watt
that you can possibly generate and so every time we improve energy efficiency we're effectively
improving both the AI capabilities for our customers and their revenues because they're
always constrained by power. The response to your keynote and forgive me Roland one minute it was
you know Vera Rubin 10x throughput really important tokens generated at one tenth
but they'll still say hey Jensen what if the electricity is just not there it's not available
well there's electricity there's never enough electricity and my question is what is the
anxiety level for you both about that issue there's always energy there's never enough energy
and this is every in every industrial revolution will be energy constrained and this industrial
revolution is also energy constrained with now one of the most important things though
here speaking about the United States if not for President Trump's pro-energy growth agenda
we would have a very hard time growing at all. In order for a new industry to emerge you need
energy and so I think it's safe to say that we wish we had more energy in the United States
Europe wish it had more energy I think the world all wish we had more energy and so we have to
invest in all sorts of different forms of energy but whatever energy you have you have to make it
as energy efficient as possible and that's one of the reasons why both of us drive our technology
roadmaps so hard because every single new generation of technology is more energy efficient
than the last that's kind of the nature of it. You are in the AI factory I won't say data center
because data center is where you store data but you are in the AI factory supply chain you must
see it. So what we see is number one is you see that the energy demands is roughly scaling with
GDP growth it's decoupling because exactly of this effect that we are providing more and more
efficient technology so that means you're decoupling but still it grows along with
economic growth and then on top comes a demand for data centers they demand high quality of
energy at the same time which in in some cases creates bottlenecks is it now from power generation
and you talk is it either renewables or gas turbines there's a huge bottleneck for gas
turbines as we speak it goes all the way to high voltage transformers to medium voltage
and to switch technologies so therefore you see along the whole supply chain obviously there's a
huge demand we are serving that there's a reason why this business of ours is growing extremely
fast we can keep up with the demand of our customers but in some cases you might end up
in bottlenecks if you keep on going as fast as possible it's a regional dependence in some cases
where you have good programs good policies and you see that it's catching up along with the
demand and others you you might end up in a gap but you see is on is on the demand is is higher
than supply chain and we are ramping up it's been the case for a while now which is which is
ultimately the condition we want yeah we want a condition where the demand is very strong because
there's you know great demand for new technology specific to you jensen how severe is the memory
bottleneck right now well the memory bottleneck is severe but we're fortunate to have worked with
all you know first of all nvidia is the only company that works with all three of the hdm
suppliers and and all three of them are major customers and major suppliers of us and and so
so i mean we have we have the good fortune of working with them for a very long period of time
and we've got things all planned out it's gonna be okay you've been asked a lot about china in
the last 24 hours i'm not gonna ask you a slightly different question what is the attitude of the
chinese government in allowing h200 to go into the country i understand the position on licenses
the administration here in the united states position but but what does china's government
say to you about how they want to approach it i haven't spoken directly to them but but ultimately
the way they'll communicate through us to us will be through the companies if the companies are are
allowed to buy nvidia products in china then there'll be strong demand and we're seeing strong
demand and so i think indirectly they they've communicated with their companies i've learned a
lot about siemens's work in software and i've always thought the word siemens to be synonymous
with with industrial manufacturing physical things it's the same lesson we've been through recently
with nvidia trying to understand actually the software competencies is that an area where we
might see you be active in mna think about what you else you might need that you might not get
from the jensen and nvidia partnership so you might because we will we will increase our software
competence our software portfolio actually it's it's a little bit short of 30 billion which siemens
invested in mna to build our software competence which brings us to the point that well what's
missing today exactly today we can build the most comprehensive physics-based digital twin of
whatever you want to build still there are pieces in operational software software which runs plans
where you can look into that ai gives new spaces obviously we invested in dogmatics it's life
science so it's about molecules we can imagine doing more there also in simulation but also
building a data backbone for this life science industry it's super relevant we did it for other
industries with team center we repeat that now with luma for life science which is again
shortening cycle times and reducing costs for life science and drugs and medicines so um and i mean
the beauty of it is that we live in both worlds we have the domain know-how we know how to run
things how to operate um how to automate plans buildings grids trains and and we have the software
competence too and and that makes us brings us in a unique position then and then we need strong
partners which are which have complementary technology which is so amazing and if that
comes together you see matching things happen when nvidia will have access to new technology
on the infant side through grock um i'm going to do my classic jensen could you please clarify
for me question which is is this an acquisition or is it a licensing deal spread over time
because jonathan and sunny joined we we hired 400 or something like that a little bit less uh
incredible engineers and we also licensed their technology they they designed an architecture
that's very very different than than um what we've done and it's focused uh on low latency
token generation um nvidia is incredibly good at inference and we're we're great at training
post training as well as the inference phase and test time scaling of of ai and so so we've got
that that that uh space covered i'm excited about some of the work that we might be able to do
together to invent a new segment that might be able to address future use cases and i haven't
described it to people but but the time will come so you know jonathan um is kind of behind the tpu
and the lpu and you know before all of this happened there was um grock v2 and v3 and and
you know all of that and it's leading to me to ask what would that new platform or segment be
what is what is it that you're trying to build on with those i haven't told anybody yet but i'm here
jensen you can tell me you know i'm gonna look in the camera and tell you that that uh come to a
gtc conference in the future and i will tell you all the secrets i'm gonna make a hard pivot because
i have to data centers in space i reported just before the holidays that the reason spacex wants
bear with me bear with me the reason that spacex wants to raise 30 40 billion dollars whatever that
is needs to buy the gpus have you discussed data centers in space with with elon and spacex jensen
i can't discuss what i haven't told everybody i've discussed with anybody you think it's a
viable technology platform sure sure i mean there's lots of energy in space right and and
and the cooling is is abundant in space and so the the the challenges of of um ai factories
are are different out in space it would be ai factories in space yeah forgive the ignorance
almost but are we literally talking about the same architecture for the gpu that goes into
an ai factory still just being able to go into kind of a satellite form factor yeah sure but
the way that you would cool it and power it would be very different and so the system design
will be radically different the chips will be the same you're pretty concerned with data centers away
i factories here on earth to see even see the need for or viability of such a technology i don't know
whether we explored all our potentials on earth yet but there's one beauty in this idea think
about any kind of manufacturing which one you want to bring to the space everything what you produce
you want to bring down to earth it's hard if you do energy how to bring it down to earth it's hard
if you produce any kind of hardware how to bring it down but tokens intelligence i mean you can
transfer easily to earth so therefore if i would start doing produce something in space i would
start there we should talk about autonomous driving have you seen elon musk's response to
to your keynote yesterday no what do you say part one in simple in short was well we were already
doing that smiley emoji on x the the the other part of by the way i wouldn't be surprised um
first of all i think i think the tesla stack is the most advanced av stack in the world and and
i think the tesla av operations is the most advanced in the world and and i i'm i'm fairly
certain that that they were already using end-to-end ai yes and whether whether their
their ai also did reasoning or not um it is somewhat secondary to to that first part i think
his point was that the first 99 is hard enough but the long tail thereafter is is important you
know it's very difficult the question i got from the audience that was at your keynote was
on a dollar per mile basis what is the fundamental difference on your stack and your software
approach to tesla's vision-based approach um ours is also vision-based you know of course we
have we have in addition to to vision we also have radar and lidar um but um but otherwise
otherwise the approach is rather similar i i think i think elon's approach is about as state
of the art um as anybody knows of autonomous driving robotics and so it's it's a it's a stack
that's hard to criticize i wouldn't criticize it i i would just encourage them to continue to do
what they're doing they're doing a great job physical ai for you is and forgive my pronunciation
but erlangen manufacturing site when's that real it's the same question i mean in contrast in
contrast to the side you was talking about before this one exists i mean we are we are ramping up
we are automating it as we speak using technology i showcase some of that and we are we are really
deploying step by step now on that very same side of the technology we was talking about
the ai brain the digital twin composer uh real worlds uh in real time connectivity to what
happens um on the shop floor so you can act and and drive basically dynamically what's
happening on the shop floor so um and we will talk about more along uh two thousand so cool
they're using siemens using video technology in the siemens factory yes and beauty is using
siemens technology in our foxconn and video factory and so you know it's a great partnership
jensen you are one of the most uh important and biggest employers in silicon valley i've
been down to santa clara c you are a leader of the technology industry right now the industry
and those of us that live in california are reviewing the billionaire's tax the question
that a lot of people submitted to me to ask you is how does that impact that talent pool
and the industry and silicon valley is it something that's of concern to you
or it is not you know i got to i haven't thought about it even once
uh we work in silicon valley because because um that's where the talent pool is
and and um we have offices all over the world wherever there's talent we have offices we have
office in germany we have office in you know all over the world and so so um we chose to live in
silicon valley and and um uh whatever whatever taxes i guess they would like to to um
apply uh so be it and you know i'm perfectly fine with it it didn't never cross my mind once
i appreciate the answer and again you know it's right now in the technology industry it's what
people are reflecting on uh we've had really not this person not this person this person
is trying to build the future of ai
