The next frontier of high-performance computing is emerging not in traditional fields such as weather prediction, astrophysics, or nuclear simulations, but within the critical search for next-generation therapeutics. 

A recent study published in Nature Biotechnology (https://www.nature.com/articles/s41587-026-03217-x) introduces an AI-enhanced computational platform capable of navigating vast chemical libraries through a synergy of machine learning, adaptive virtual screening, and extreme-scale cloud computing. This research addresses a significant challenge in computational science: how to navigate a chemical universe containing tens of billions of potential candidates without the necessity of exhaustive computation. 

The findings suggest that the solution lies not in merely expanding supercomputing capacity, but in optimizing computational efficiency. To this end, researchers developed AdaptiveFlow, an open-source platform engineered for ultralarge virtual screening and cross-platform deployment. In a landmark demonstration, the system’s ligand-preparation workload achieved near-perfect linear scaling across more than 5.6 million Intel virtual CPUs on Amazon Web Services. 

This breakthrough fundamentally shifts the paradigm of computational drug discovery. By transforming chemical search into a massively parallel process and utilizing AI to prioritize the most promising regions of chemical space, AdaptiveFlow effectively optimizes resource allocation where it is most scientifically impactful.

When the database becomes a supercomputer problem

Modern drug discovery increasingly begins not in a laboratory but inside a computational representation of chemical space.

The theoretical universe of drug-like molecules is estimated to contain more than 10^60 possibilities. No conventional computing system can enumerate, simulate, and experimentally evaluate anything remotely approaching that number.

Even the portion of chemical space that can be purchased or synthesized on demand has exploded. The study points to libraries that have grown from billions of compounds to trillions, while one of the largest ready-to-dock collections examined in the work, the Enamine REAL Space, contains approximately 69 billion compounds.

At that scale, virtual screening becomes an HPC problem.

A brute-force strategy would attempt to dock enormous numbers of molecules against a biological target, calculate their interactions, and rank the resulting candidates. But docking calculations are computationally expensive, and multiplying them by tens of billions quickly becomes impractical.

The fundamental challenge therefore becomes one of computational triage.

Which molecules should receive expensive calculations?

Which regions of chemical space are likely to contain useful candidates?

How can a supercomputing infrastructure process millions of independent calculations without allowing scheduling, data movement, and orchestration overhead to overwhelm the scientific workload?

AdaptiveFlow approaches those questions as an integrated HPC architecture.

Six million CPUs, one computational pipeline

The most dramatic demonstration involves the platform’s AdaptiveFlow Ligand Preparation (AFLP) component.

The researchers rewrote the software in Python and designed its workload around extremely fine-grained parallelism. Rather than treating a gigantic chemical library as one monolithic job, molecules are divided into collections and computational work units that can be distributed across large numbers of processors.

The system was able to execute ligand-preparation calculations using more than 5.6 million Intel vCPUs simultaneously.

The scaling behavior was described as essentially perfectly linear.

That is an important HPC result.

Perfect or near-linear scaling means that adding computational resources continues to produce approximately proportional increases in throughput. In real-world supercomputing, achieving that behavior at relatively small scales can be difficult; maintaining it across millions of concurrent CPU resources is considerably more challenging.

AdaptiveFlow’s architecture attacks several sources of inefficiency simultaneously.

Work is broken into relatively small subjobs, allowing the scheduler to distribute computational tasks across a huge pool of resources. Approximately 1,000 ligands can be grouped into collections, while subjobs and job arrays help reduce scheduling overhead while maintaining high levels of parallelism.

The architecture supports both Slurm-based HPC environments and AWS Batch, allowing the same computational concepts to extend from conventional supercomputing clusters into elastic cloud infrastructure.

That flexibility is increasingly important as scientific workloads become too large, or too intermittent, to justify running exclusively on fixed infrastructure.

The cloud becomes a scientific supercomputer.

The AdaptiveFlow demonstration also illustrates how the definition of a supercomputer is changing.

The computational infrastructure used in the study was built around AWS services including AWS Batch, Amazon S3, CloudFormation, and containerized workloads. The platform was also designed to take advantage of spot instances, allowing large computational workloads to exploit spare cloud capacity.

During the enormous library-preparation workload, less than 0.1% of the CPU hours used were interrupted by preemption.

That resilience is critical when a scientific application is operating at millions-of-CPU scale.

At that magnitude, failures are inevitable. The question is not whether individual compute resources will disappear, but whether the application architecture can absorb those failures without bringing the entire scientific workflow to a halt.

AdaptiveFlow’s short, independently executable subjobs provide that resilience. A failed unit of work can be rerun rather than forcing the entire calculation to restart.

The approach resembles an important principle from traditional HPC: break a large scientific problem into sufficiently independent pieces that the system can continue making progress even when individual computational elements fail.

The difference is scale and elasticity.

Instead of a fixed machine with a known number of processors, the cloud can provide an enormous pool of computational resources when the workload demands it.

AI does not replace HPC; it tells HPC where to work.

The most important innovation, however, may happen before the processors begin their calculations.

AdaptiveFlow incorporates Adaptive Target-Guided Virtual Screening (ATG-VS), which uses an 18-dimensional grid of molecular properties to organize chemical space.

That organization gives machine-learning methods a map.

Rather than blindly docking every molecule, the system can identify chemical subspaces that are more likely to contain useful compounds for a particular biological target. Computational resources can then be concentrated in those regions.

This creates a fundamentally different relationship between AI and high-performance computing.

Traditional HPC often asks:

How can we execute this calculation faster?

The AdaptiveFlow approach adds another question:

Do we need to execute this calculation at all?

That distinction could prove transformative.

If machine learning can reliably identify regions of chemical space that are likely to contain valuable candidates, the supercomputer no longer needs to spend equal amounts of computational effort everywhere.

It can spend more computation where the probability of scientific payoff is highest.

The researchers report that ATG-VS can reduce screening costs by up to 1,000 times for the 69-billion-compound REAL Space compared with exhaustive searches, while retaining strong enrichment of promising candidates.

The computational strategy therefore becomes a combination of AI-guided selectivity and HPC-scale execution.

More than 1,500 ways to search

AdaptiveFlow is not tied to a single docking algorithm.

The platform integrates more than 1,500 docking protocols, including GPU-accelerated approaches and machine-learning-based methods.

That matters because molecular docking is not one universal calculation. Different algorithms make different approximations about molecular interactions, protein flexibility, and binding configurations.

By providing a common computational framework for many docking approaches, AdaptiveFlow can turn the HPC infrastructure into a large experimental platform for computational chemistry.

The architecture also accommodates modern AI-based docking technologies, including deep-learning approaches.

According to the study, deep-learning docking and GPU acceleration can potentially provide an additional 10× to 100× increase in throughput.

Combined with the reduction in the amount of chemical space that must be searched, the resulting computational pipeline is dramatically different from simply attempting to brute-force billions of calculations.

The goal is not merely to build a faster molecular-docking machine.

It is to build a system that continuously decides what should be computed next.

A three-layer computational architecture

AdaptiveFlow is organized around three principal components.

AFLP, or AdaptiveFlow Ligand Preparation, prepares enormous molecular libraries for screening.

AFVS, the AdaptiveFlow Virtual Screening engine, performs the actual screening calculations.

And AFU, AdaptiveFlow Unified Workflow, connects computational stages into complete screening pipelines.

This modular design is significant for HPC because preparation, screening, machine-learning analysis, and downstream processing have different computational characteristics.

Some stages are CPU-intensive. Others can exploit GPUs. Data-intensive stages require high-throughput storage and efficient movement of molecular information. Scheduling-intensive stages require massive numbers of short jobs to be launched and completed efficiently.

The platform therefore treats drug discovery as a workflow-scale HPC problem, rather than simply a collection of individual scientific calculations.

The underlying infrastructure can operate across CPU and GPU systems and, where individual docking programs permit, ARM-compatible computing environments.

That portability could become increasingly important as heterogeneous computing becomes the norm across both cloud and traditional supercomputing centers.

Turning chemical space into a parallel data structure

One of the deeper implications of the work is that chemical space itself can be treated as a computational data structure.

Instead of viewing 69 billion molecules as 69 billion independent records, AdaptiveFlow organizes the library according to molecular properties.

The 18-dimensional representation provides a mechanism for partitioning the search space.

From an HPC perspective, this creates a hierarchy: chemical space → molecular subspaces → compound collections → computational subjobs → CPU/GPU resources.

Each level can be optimized independently.

The scientific problem is therefore transformed into a scheduling and resource-allocation problem that modern distributed computing infrastructure is exceptionally good at solving.

The result is an unusual convergence of disciplines: computational chemistry supplies the scientific models, machine learning supplies the intelligence for prioritization, and HPC supplies the massive parallel execution engine.

The proof is not only computational.

The researchers did not stop at demonstrating computational scalability.

AdaptiveFlow was used to identify inhibitors against biological targets including FSP1 and PARP1, with experimental work providing evidence that some computationally identified compounds bind their intended targets.

For PARP1, the researchers conducted a primary screen of 100 million molecules and synthesized 160 candidates for experimental verification. Protein NMR and X-ray crystallography were used to validate direct binding.

The study also reports cellular experiments involving BRCA1-deficient triple-negative breast cancer cells and describes computationally identified PARP1/PARP2 inhibitors with selective activity.

For FSP1, the computational search produced nanomolar inhibitors, with co-crystal structures helping researchers understand their binding modes.

These results are important because they demonstrate the purpose of the HPC infrastructure.

The objective is not to produce impressive processor counts.

It is to turn processor cycles into scientific discoveries.

The new supercomputing equation

For decades, high-performance computing has largely been about increasing the number and performance of processors available to scientists.

More cores.

More memory.

Faster interconnects.

More GPUs.

More efficient algorithms.

AdaptiveFlow points toward another dimension of performance: Compute less, but compute the right things.

That could become one of the defining characteristics of AI-enhanced scientific computing.

A system capable of processing billions of candidates does not necessarily need to evaluate billions of candidates with the same expensive algorithm. Machine learning can serve as an intelligent filter, reducing the computational search space before the most expensive calculations begin.

Then HPC can apply enormous parallelism to the candidates that remain.

This is particularly powerful for problems where the search space is vast but useful solutions are comparatively rare.

Drug discovery is one example. Similar approaches could potentially be relevant to materials science, catalyst discovery, protein engineering, battery chemistry, and other scientific domains in which researchers confront enormous combinatorial spaces.

From brute force to intelligent force

The achievement of 5.6 million concurrent vCPUs is impressive on its own. But the more consequential achievement may be the combination of that scale with algorithmic intelligence.

The researchers have effectively connected two kinds of acceleration.

Hardware acceleration: distribute computational work across millions of CPU resources and, where appropriate, GPUs.

Algorithmic acceleration: use molecular-property organization, machine learning, and adaptive screening to avoid wasting computation on low-value regions of the search space.

The two approaches multiply one another.

If HPC makes an individual calculation faster, AI can make the overall search smaller.

If AI identifies a promising region, HPC can examine that region at extraordinary scale.

And if the system can dynamically move between those two processes, the boundary between algorithm design and supercomputer architecture begins to disappear.

Citizen-scale possibilities from supercomputer-scale discovery

The ultimate promise of this approach extends beyond the processors themselves.

The study describes an open-source platform that can make ultralarge screening workflows more accessible to researchers using HPC clusters and cloud resources.

That matters because the world’s largest computational discoveries increasingly depend on software ecosystems, not just machines.

A scientific platform that can scale from conventional clusters to millions of cloud CPUs potentially gives researchers a common framework for experiments that once required highly specialized infrastructure.

The result is a new model of scientific computing in which the supercomputer becomes not merely a place where calculations happen, but part of an adaptive discovery loop.

AI proposes where to look.

HPC searches at extraordinary scale.

Experimental science determines what matters.

The results feed back into the computational models.

And the cycle begins again.

The next frontier is not bigger; it is smarter.

The 69-billion-molecule problem illustrates something fundamental about the future of supercomputing.

There will always be scientific problems for which simply adding more processors is not enough.

When the search space grows faster than available computing resources, efficiency becomes as important as raw performance.

AdaptiveFlow demonstrates one possible answer: combine extreme parallelism with artificial intelligence so that computing resources are directed toward the most scientifically promising regions of an enormous search space.

The researchers’ 5.6-million-vCPU demonstration provides the hardware-scale proof point. The 69-billion-compound library provides the computational challenge. The AI-guided reduction in search costs provides the algorithmic breakthrough.

Together, they suggest a future in which the world’s largest scientific computers do not simply calculate faster.

They calculate more intelligently.

For drug discovery, that could mean navigating chemical space that was once computationally unreachable. For high-performance computing more broadly, it offers a compelling vision of what comes next: machines that combine massive parallelism with machine intelligence to transform impossible searches into tractable scientific experiments.

The age of brute-force scientific computing is not necessarily ending.

But it may be evolving into something far more powerful: intelligent force at supercomputer scale.

NVIDIA CEO tells global technology ministers that artificial intelligence is becoming national infrastructure, with data centers, power, GPUs and computing capacity forming the foundation of the next industrial revolution

In Chapel Hill, N.C., NVIDIA CEO Jensen Huang delivered a compelling message to G20 technology ministers: the next great infrastructure race will not be defined by traditional networks, but by the pursuit of artificial intelligence. Huang asserted that nations must categorize AI infrastructure alongside essential utilities like water, electricity, and transportation to avoid falling behind in a historic economic transformation. He emphasized that the AI revolution requires massive investments in computation, memory, energy, and physical data-center capacity. Ultimately, as generative AI becomes a fundamental driver of modern economies, Huang urged global leaders to recognize that building comprehensive supercomputing infrastructure is no longer optional, but a prerequisite for future progress.

The five-layer architecture of the AI economy

Huang’s vision begins with what he describes as a five-layer AI stack.

At its foundation is the computing infrastructure required to execute increasingly sophisticated models. Above that are the software and model layers, followed by data and applications, the parts of the stack where AI ultimately becomes useful to businesses, scientists, governments, and individuals.

Huang argues that countries do not necessarily need to dominate every layer.

Instead, each nation should determine where it has competitive strengths and invest accordingly.

A country might concentrate on semiconductor manufacturing. Another might specialize in energy, data centers, AI models, scientific applications, or robotics.

But there is one layer Huang believes every country must embrace:

AI diffusion.

The objective, he told ministers, should be getting artificial intelligence into virtually every sector of the economy, from education and healthcare to manufacturing and science.

That concept closely parallels the argument explored in SuperComputing News’ recent analysis of Meta’s vision for personal superintelligence. Meta proposed that supercomputing could eventually become an invisible utility, with users interacting with AI agents while enormous centralized computing systems perform the underlying work. 

Huang’s G20 message points toward the same destination from the infrastructure side.

If AI is going to become available to billions of people, somebody has to build the supercomputers.

The data center is becoming the new power plant.

For decades, computing infrastructure was measured in processors, memory, and storage.

The AI era increasingly measures it in gigawatts.

In his interview, Huang described a remarkable escalation in infrastructure economics. He estimated that building approximately one gigawatt of AI infrastructure represents an investment of roughly $50 billion to $60 billion.

He also said NVIDIA expects infrastructure on the order of 100 gigawatts to be built between now and the end of the decade.

Those numbers illustrate how dramatically the economics of computing have changed.

A traditional high-performance computing center might be measured in megawatts. Frontier AI infrastructure is increasingly being discussed in hundreds of megawatts and gigawatt-scale deployments.

The computer has effectively become an industrial facility.

And that facility requires an industrial ecosystem.

It needs electrical generation.

It needs high-voltage transmission.

It needs substations.

It needs advanced cooling.

It needs fiber networks.

It needs enormous storage systems.

It needs thousands, or potentially hundreds of thousands, of accelerators.

And it needs the semiconductor supply chain capable of producing them.

This is why Huang’s comparison of AI to electricity and roads is more than a metaphor.

The infrastructure itself is becoming an economic asset.

GPUs turned supercomputing into an AI engine.

The technical foundation underneath Huang’s argument is the architecture that NVIDIA helped establish decades ago.

Graphics processing units were originally developed for massively parallel workloads in computer graphics.

But the same architectural characteristics that made GPUs effective at rendering images also made them exceptionally well suited to scientific computing.

Fluid dynamics.

Particle physics.

Quantum chemistry.

Image reconstruction.

Numerical simulation.

And eventually, artificial intelligence.

Huang emphasized this broader computational heritage in the interview, noting that GPUs are fundamentally parallel processors capable of addressing workloads extending well beyond AI.

That matters because modern AI workloads are themselves enormous numerical problems.

Training and inference involve vast collections of matrix operations executed across thousands of processing elements. At hyperscale, individual accelerators become components in distributed computing systems in which networking, memory bandwidth, storage, and software are as important as raw floating-point performance.

The result is a new class of supercomputer.

It may be called an AI factory.

It may be called a hyperscale data center.

It may be called an AI cloud.

But architecturally, these facilities increasingly resemble some of the world’s most sophisticated supercomputing systems.

AI is escaping the data center.

Huang’s vision also extends beyond traditional cloud computing.

He described AI as an intelligence layer that can be placed inside digital and physical systems.

An AI agent connected to software tools can become a digital worker.

Connect that agent to a robotic manipulator, and it becomes a manufacturing system.

Put it inside a vehicle, and it can become an autonomous driving system.

Connect it to laboratory equipment, and it can become an automated scientific research platform.

That progression, from model to agent to physical machine, is one of the most important developments in modern computing.

AI is no longer confined to a browser window.

It is moving into factories, laboratories, vehicles, robots, and scientific instruments.

And every physical deployment adds another computational workload.

The supercomputer is moving into the physical world.

From automation to augmentation

Perhaps the most optimistic element of Huang’s vision is his argument about employment.

He rejects the simplistic idea that increasingly capable AI necessarily means the disappearance of human work.

Instead, he argues that AI will automate individual tasks while leaving the larger purpose and context of jobs in human hands.

In his view, workers will become supercharged.

That concept is particularly important when considered alongside the personal-superintelligence model explored in SuperComputing News' Meta analysis.

The fundamental question is not simply whether AI can perform a task.

It is whether access to enormous computational intelligence can allow one person to accomplish what previously required an entire organization.

A researcher could use AI to analyze thousands of scientific papers.

An engineer could generate and evaluate enormous numbers of design alternatives.

A programmer could have AI agents write, test, and debug software.

A small business could gain access to sophisticated financial, marketing, and operational capabilities.

A student could have an individualized AI tutor.

A scientist could connect an AI research agent directly to simulation software.

The interface becomes conversational.

The workload underneath remains supercomputing.

The democratization of computational intelligence

This is where Huang’s vision intersects most directly with the idea of supercomputing for the masses.

Historically, access to advanced computing was concentrated in national laboratories, universities, and major corporations.

A researcher needed access to a supercomputer center.

A company needed to build or rent specialized infrastructure.

A student generally had access only to whatever computing resources were available locally.

AI changes that equation.

The computational infrastructure can remain centralized while the intelligence becomes distributed.

A smartphone does not contain a hyperscale data center. It connects its user to one.

The same model can apply to AI.

The device becomes the interface.

The network becomes the connection.

The data center becomes the supercomputer.

And the AI becomes the intelligence layer connecting humans to the computational system.

That is why the construction of AI infrastructure is so important.

The more people who use AI, the more computing capacity society needs.

Every country needs its own computational capacity.

Huang’s message to G20 ministers was not simply that governments should buy NVIDIA hardware.

His broader argument was that countries need domestic AI capacity.

He urged governments to determine which portions of the AI stack they can develop competitively while ensuring that researchers, students, companies, and startups have access to computing.

That access can have a powerful multiplier effect.

Give a startup a powerful AI platform, and it can develop a product.

Give researchers large-scale compute, and they can test hypotheses that previously would have taken years.

Give students access to advanced AI tutors and the economics of education begin to change.

Give manufacturers AI-enabled robotics and simulation, and production processes can be redesigned.

Huang said NVIDIA has seen researchers and startups become activated once local computing infrastructure becomes available.

That may ultimately be one of the strongest arguments for national AI investment.

The objective isn’t merely to own computers.

It is to create computational capacity for an economy.

The electricity problem

There is, however, an unavoidable physical constraint.

Computers require electricity.

The G20 discussions have already highlighted concerns that power generation and transmission may struggle to keep pace with AI’s rapid expansion. Elon Musk warned during the first day of the meeting that power shortages could become a near-term constraint, while other technology executives have emphasized the need for faster data-center construction. 

This transforms AI policy into energy policy.

A nation cannot build a gigawatt-scale AI facility without a gigawatt-scale power strategy.

That means AI investment could stimulate development far beyond the technology sector.

Power plants.

Transmission lines.

Transformers.

Cooling systems.

Construction.

Semiconductor factories.

Networking equipment.

Advanced materials.

Skilled trades.

Engineering.

Operations.

Cybersecurity.

The AI infrastructure boom therefore has the potential to become an industrial infrastructure boom.

The jobs are not only in software.

Huang pointed to the expanding employment ecosystem surrounding AI infrastructure, from chip fabrication and computer manufacturing to data centers and AI factories.

That is an important distinction.

The AI revolution is frequently portrayed as a race among software engineers and machine-learning researchers.

But the physical AI economy requires electricians, construction workers, mechanical engineers, power engineers, network engineers, semiconductor technicians, cooling specialists, and data-center operators.

It also requires the enormous industrial supply chains supporting them.

The result could be a new form of technological manufacturing economy in which software intelligence and physical infrastructure reinforce one another.

AI creates demand for infrastructure.

Infrastructure creates computing capacity.

Computing capacity enables new AI applications.

Those applications create new economic demand.

And the cycle accelerates.

Safety without surrendering ambition

Huang’s optimism does not mean he believes AI safety should be ignored.

Quite the opposite.

He argued that technology developers have a responsibility to build systems safely and work with regulators.

But he warned against allowing fear of hypothetical harms to become the primary framework for technology policy.

His preferred approach is to regulate actual, measurable harms while allowing emerging technology enough room to develop.

The argument reflects a broader theme in his interview: technological advancement itself can contribute to safety.

AI systems can become more reliable through better models, better grounding, improved reasoning, better evaluation, and more sophisticated software.

For Huang, the answer to uncertainty is not necessarily to stop technological progress.

It is to improve the technology.

That position is now becoming an important part of the international debate over AI policy. Reuters reported Wednesday that Huang urged G20 countries to avoid regulations focused primarily on theoretical harms and instead concentrate on practical problems associated with AI.

The AI industrial revolution

Huang believes the transformation underway is comparable to previous infrastructure revolutions.

Electricity changed manufacturing.

The automobile changed transportation.

The internet changed communication.

Computing changed information processing.

AI could change the production of intelligence itself.

That is a profound shift.

For centuries, societies invested enormous resources in educating humans because human intelligence was the fundamental productive resource.

Huang offered a provocative analogy in his interview: just as schools and universities helped societies produce and distribute human intelligence at scale, AI could increasingly provide a digital form of intelligence at scale.

That does not make human education obsolete.

It makes its potential reach much larger.

A student in a region with limited access to specialized instruction could potentially interact with an AI system capable of explaining advanced mathematics, physics, programming, or chemistry.

A small research team could access computational capabilities that once required a national laboratory.

A startup could rent intelligence rather than build an enormous technical staff.

That is the democratization of supercomputing.

The one-person enterprise

The economic consequences could be enormous.

If AI agents become capable of performing research, programming, analysis, design, marketing, and administrative tasks, the minimum viable size of an organization could shrink.

A single entrepreneur might be able to coordinate a collection of specialized AI agents.

A small engineering firm could perform sophisticated simulation and design.

An independent scientist could automate portions of a research workflow.

A local manufacturer could use AI to optimize production.

The limiting factor increasingly becomes not access to software, but access to compute and the ability to direct it effectively.

This is precisely the issue raised by Supercomputing News' earlier examination of Meta’s personal-superintelligence strategy: the future of computing may not be defined by making supercomputers smaller, but by making their capabilities accessible to vastly more people. 

The supercomputer disappears behind the interface.

This may be the most important transformation of all.

The world’s most powerful computing systems may become increasingly invisible.

A person may ask an AI system to design a battery.

Behind that request, an agent could search scientific literature, generate candidate materials, run molecular simulations, evaluate results, and propose another iteration.

An engineer might request a more efficient aircraft design.

The AI could generate geometries, invoke computational fluid dynamics simulations, analyze the results, and repeat the process.

A scientist might ask an AI system to investigate a biological mechanism.

The system could search databases, construct hypotheses, and launch computational experiments.

To the user, it looks like a conversation.

To the infrastructure, it is a massive distributed workload.

That is the future Huang is describing.

The interface becomes simple because the infrastructure underneath becomes extraordinarily complex.

From supercomputing centers to an intelligence grid

The implications extend beyond NVIDIA.

The AI infrastructure race is creating a new computational ecosystem involving semiconductor companies, cloud providers, national laboratories, universities, telecommunications companies, utilities, and governments.

It is increasingly reasonable to think of this system as an emerging global intelligence grid.

Its components are physical:

• AI accelerators
• CPUs
• high-bandwidth memory
• optical and electrical networking
• distributed storage
• data centers
• cooling systems
• power generation
• transmission networks

Its software layer is equally important:

• operating systems
• AI frameworks
• compilers
• distributed training systems
• inference engines
• agent frameworks
• model-serving platforms
• scheduling and orchestration

And above all of that are the applications that turn computational capacity into economic value.

This is fundamentally a supercomputing architecture.

The race is no longer simply to build a better model.

For much of the AI boom, the conversation centered on model size.

Then it moved toward training efficiency.

Now the strategic conversation is increasingly about infrastructure.

Who has enough GPUs?

Who has enough electricity?

Who can build data centers quickly enough?

Who has sufficient networking?

Who can manufacture advanced memory?

Who can connect new facilities to the grid?

Who has the software ecosystem to keep thousands of accelerators operating efficiently?

And who can put that capacity into the hands of researchers, companies and citizens?

The answers could determine which countries lead the next phase of the industrial economy.

While there is no guarantee that every prediction regarding artificial intelligence will materialize, given potential risks such as infrastructure delays, power constraints, rising costs, model underperformance, and regulatory shifts, Jensen Huang’s message offers a fundamentally optimistic framework. The technology is poised not to replace human ambition, but to amplify it. By augmenting existing intellectual capacity with near-unlimited access to computational intelligence, Huang invites nations to elevate their ambitions, as the technology renders larger goals attainable. Ultimately, his argument transcends corporate interests, focusing instead on the imperative of developing computational capacity. As the next industrial revolution takes shape through silicon, electricity, software, and human ingenuity, the nations that invest in the necessary infrastructure may find that AI becomes the foundational architecture for entire industries. Moving beyond the historical confines of national laboratories and corporate data centers, the next phase of this evolution involves democratizing supercomputing, transforming it into an everyday capability for billions, and establishing the essential computational bedrock of the future global economy.