Video summary
The global landscape is currently undergoing a historic shift as the world prepares to invest an unprecedented 31.6 trillion dollars in data center construction by 2050, a figure that dwarfs all previous infrastructure booms combined. This massive surge in spending, driven entirely by the artificial intelligence boom, represents a demand for computing power and storage that exceeds current capabilities, particularly in developed nations where connecting new facilities to the power grid can take up to eight years. While annual investment is projected to jump from 800 billion dollars this year to 1.8 trillion by the end of the decade, the industry faces significant logistical challenges, with hyperscalers demanding new centers be built within six months—a timeline that often conflicts with the lengthy requirements for grid integration and physical infrastructure development.
The momentum behind this investment extends far beyond major technology giants like Nvidia, creating a complex supply chain where manufacturers of power and cooling equipment across Asia are capitalizing on the rush to avoid bottlenecks. Experts argue that this level of expenditure is fully justified by the transformative value AI brings to every sector, noting that modern development teams no longer write code manually but instead leverage large language models to accelerate product creation. As applications evolve from simple text generation to advanced robotics and agentic systems capable of interacting through natural language interfaces, the need for robust infrastructure becomes critical to meet consumer expectations for products that continuously improve rather than stagnating at capabilities seen just a few years ago.
Despite concerns regarding environmental impact and the speed of construction, industry leaders emphasize that every technological advancement comes with costs that must be weighed against the profound benefits they unlock. The acceleration of scientific research, healthcare innovations, and medical breakthroughs is expected to yield significant returns that outweigh the initial heavy investment in power and cooling systems. As physical AI becomes more integrated into daily life, the infrastructure supporting it will grow increasingly essential, ensuring that society can harness the full potential of artificial intelligence while managing the necessary resources to sustain its rapid expansion.
Read the full video transcript
Now, the world is about to spend more
building data centers than it ever spent
on railways, electricity grids, or the
entire internet combined. And it's all
thanks to AI. That's right. For anyone
unfamiliar, a data center is essentially
a massive warehouse packed with computer
servers that store, process, and power
everything from your Netflix stream to
chat GPT. And a new Price Waterhouse
Coopers report says that global
investment in these facilities will hit
an almost unimaginable 31.6 trillion
dollars by 2050. And it's being driven
entirely by the AI boom. Annual spending
alone is set to jump from around 800
billion dollars this year to 1.8
trillion dollars by 2050. And it's not
just Nvidia making money off of this. An
entire supply chain of power and cooling
equipment makers across Asia is cashing
in as everyone is racing to avoid
infrastructure bottlenecks. But here is
the catch. Hyperscalers want new data
centers built within 6 months, and
getting them actually connected to the
power grid can take up to 8 years in
developed countries, which means the
demand for AI infrastructure is
massively outpacing the world's ability
to actually physically power it.
And on that note, joining us is Tomer
Galy. He's global CTO at Deloitte's
Nvidia alliance.
Tomer, I think the first question is, is
the world actually capable of building
this much infrastructure this fast?
>> I think Jensen Huang, the CEO of Nvidia,
said it. He said that the world we need
we need more electricians, more
plumbers. So, he referred to these kind
of aspects.
And like you said, the demand is
extraordinary.
It's basically like endless, because
right now the use cases
you have use case almost for everything.
And this is today.
>> Right.
>> Meaning in the future, once you'll have
more robotics, because today, you know,
robotics
is advanced, but not yet. And when
you'll have these kind kind use cases
driving further AI adoption. Anyway, we
see this everywhere.
>> Right. You know, we're talking about
over 30 trillion dollars in spending by
2050.
This is a number that just simply dwarfs
previous infrastructure booms that we've
seen in the past. Railways, electricity
grid, you heard me name them.
You know, is that level of investment
actually justified by AI's projected
value?
>> I would say completely yes. Completely
yes, because if we look at it today,
the basic use case, I think in every
high-tech company, every startup
company, developers don't write code
today.
In the teams that I lead, no one writes
code. Everyone use AI, whether it's
OpenAI, whether it's Anthropic. Everyone
uses that or open-source large language
models.
So, we have these kind of use cases. And
of course, eventually they're developing
products. And today, the clients also
expect to see AI in the products, right?
You don't want to see products that you
were capable of doing two, three years
ago. Today, everyone is expecting to see
agents. You want to see interfaces with
the natural user interface. So, you
speak to it, you write to it, but in
free text language. And you have the
agentic systems doing whatever they were
designed to do. So, yes, this investment
uh proves itself. And we see this also
in the stock market. The way that
everything shifts.
>> Right. I mean, there's obviously a lot
of fear, you could say, amongst a
certain sector of the general public
about the environmental concerns with,
you know, the speed at which we're going
to be seeing these data centers built.
Um you know, a lot of this spending is
essentially flowing into power and
cooling infrastructure that most people
have never heard of. I mean, a lot of
this is just, you know, terminology that
the average viewer right now who's
watching doesn't understand.
What is your response to that?
>> I would say everything comes with a cost
and that cost you know has a certain
weight and look at all of the benefits
that we are getting and we are going to
get the improvements and acceleration of
the pace of science, of research, of
health care, of medicine. All of these
aspects that we are all going to gain
from this because we do have AI
basically everywhere and it's just
growing and the capabilities are just
improving.
And like I said, that's for this aspect
and as it comes more to the aspect of
physical AI
we're going to see this more in the
day-to-day.
>> Well, you know, I have to say that it's
just absolutely fascinating to
to see what AI can already offer the
public. Every single person we know we
know today is using AI in some form,
obviously. Uh working together with
NVIDIA, you're witnessing a lot of the
behind-the-scenes action that it takes,
you know, when it comes to actually
rolling out AI globally. I'm sure that
we're going to have you back on for
continued updates about what happens
with these these uh data centers more
specifically. Tomer, thank you so much
for joining us.
>> Thank you very much.