Moore’s Law turns 60 this year, but chip scaling is slowing while AI demand keeps rising. What guiding principle do you think will define computing’s next era?
Answer: Accelerated computing is becoming the defining principle. For decades, Moore’s Law gave
us abundance by making chips faster and cheaper. But as that slows, the real gains are coming
from re-architecting the computing stack for performance per watt. At NVIDIA, we say accelerated computing is sustainable computing, because moving workloads from traditional servers to GPU-
accelerated systems delivers orders-of-magnitude improvements in efficiency. Over the last decade, that’s meant a 100,000-fold boost in energy efficiency, which is what makes modern AI possible.
AI is quickly proving to be the most powerful sustainability tool we’ve ever had. It saves materials
through simulation and digital twins, improves grid reliability, and drives efficiencies across
industries. While AI demand is rising, much of its load is replacing less efficient computing.
The next era of computing will be defined not just by raw scale, but by how much more we can achieve with every watt of energy we use.
With AI capacity expanding at 33% a year, which regions are you betting on as the unexpected leaders in the “AI factory” race?
Answer: That’s an open question. We know that access to energy, especially clean energy, is the
primary factor in siting for AI factories right now. Regions and countries eager to establish
innovation hubs recognize that and are working to expand energy production, modernize grids and
increase access to clean energy. Because regulatory delays can be significant in the energy sector,
it is likely that regions that are effective at streamlining approvals will see the most growth in the
near term.
And that geographic flexibility is actually one of the exciting things about AI factories. We say “AI
doesn’t care where it goes to school,” meaning AI models can be trained anywhere–having a fast
connection to the Internet doesn’t matter for AI training like it does for traditional data centers.
And even for AI inference, latency is much less of an issue than for typical Internet traffic. Regions
outside of today’s data center hubs could see new AI factories—if energy is available—helping to
ease grid strain typically seen with concentrated development.
Data centres may use 4% of the world’s power by 2030. What concrete steps do you believe could change that trajectory?
Answer: Actually, I would disagree that having AI consume a noticeable portion of the world’s
energy is a bad thing–for sustainability or for humanity. AI is displacing and reducing energy
consumption in other sectors–it excels at improving efficiency and enabling innovation and
discovery, from manufacturing to buildings to transportation. If we didn’t see AI becoming a
relatively larger share of energy consumption, I think that would be a sign that we aren’t using it
effectively and aren’t taking full advantage of its benefits.
Data centres are inherently going to need more energy to power this next era of AI and the
groundbreaking innovations and research it is propelling forward. AI is also a tool that can be used
to optimize the energy grids that it needs to run on. Data centres can and are being run on
renewable energy sources. NVIDIA purchases carbon-free electricity to cover 100% of our leased
data center footprint as of 2025. This is increasingly becoming a standard practice in this sector.
AI’s ability to modulate power usage can help mitigate rolling blackouts, stabilize energy prices, and
create the opportunity for a smooth transition to a renewable energy grid. Using the right kind of
energy along with integrating AI into the energy system grid for optimized efficiency, is how we
ramp up AI usage and data centres in a responsible and sustainable way.
Over the long run, the evidence is clear–AI can be a net contributor to efficiency and sustainability
across industries. We’re already seeing measurable savings in energy-intensive industries like
manufacturing, transportation, and construction. Done right, AI will expand energy abundance while
lowering intensity.
Generative AI spend is forecast to hit $644 billion next year. Which new business segment do you expect to see the fastest growth as a direct result of that surge?
Answer: We’re already seeing AI drive transformation in sustainability and industrial sectors, and I
expect that to accelerate fastest. From climate modeling with platforms like Earth-2, to
breakthroughs in carbon capture and new battery chemistries, to digital twins that optimize
factories and buildings, AI is unlocking massive efficiency gains.
One example I love to talk about is Foxconn’s new facility in Mexico, where AI-powered digital twins
are forecasting a 30% lifetime reduction in energy use. As generative AI matures, the ability to
model, optimize, and reimagine physical systems will be one of the most powerful engines of
growth–especially in energy, manufacturing, and climate solutions. This is where AI can help make
the physical world more predictable, efficient, and sustainable.
But this is ultimately difficult to predict, in part because AI is probably the most democratizing
technology ever developed–which means small teams and even individuals are empowered to do
more with less, and to test new ideas. We’re really living in the Cambrian explosion of tech, and
there are more opportunities than ever for positive, innovative disruption.
What do you consider the most exciting current debates in the conversations around the
evolution of AI and HPC (high performance computing)?
Answer: One conversation I find interesting is around whether AI data centers can serve as flexible
load on the grid. In the near term, during this period of rapid AI innovation, the infrastructure is so
valuable that it’s hard for customers to turn it down, even temporarily. But over time, that flexibility
is a huge opportunity.
AI factories are firm resources, but they can also be designed to dial back selectively – shifting non-
urgent workloads while maintaining critical services. We’ve already seen promising demonstrations of this, like our work with Emerald AI, where a cluster of NVIDIA GPUs cut power use by 25% for
three hours during a grid stress event, without sacrificing service quality.
That kind of flexibility could make data centers not just consumers of energy, but partners to the
grid – helping cities avoid blackouts, lowering costs for communities, and making it easier to
integrate renewables. As the AI revolution matures, I think we’ll see this debate turn into a real
solution for building cleaner, more reliable power systems.
