🎤 Opening and the Moment for Accelerated Computing
At GTC in Washington, D.C., I stepped onto the stage with a clear message: this is a defining moment for accelerated computing. I called the event "the Super Bowl of AI" and meant it. The language I used was intentional because the convergence of hardware, software, and large scale deployments has transformed what was once a niche engineering discipline into a foundational technology for society.
For thirty years we have been advancing a specific kind of computing. We gave it a name. We call it accelerated computing. It is not an incremental step beyond conventional processing. Instead, it is a categorical shift in how we design systems to solve the hardest computational problems, from AI and simulation to scientific research and national security applications.
In simple terms I said that we invented the GPU and the programming model called CUDA. Those two inventions created a platform that allowed developers, researchers, and companies to rethink computing architectures. They enabled a whole ecosystem of tools, frameworks, and applications to flourish. When you consider the scale and pace of AI progress over the past decade, the lineage from GPUs and CUDA is undeniable.
🧭 What I Mean by Accelerated Computing
When I talk about accelerated computing I am reminding people that not all compute is equal. General purpose CPUs are indispensable. However, for workloads that require massive parallelism and high throughput, specialized accelerators are a far better match. That is what GPUs were built to do.
Consider the nature of modern AI. Training a large neural network requires performing trillions of floating point operations, applying optimization techniques across millions or billions of parameters, and doing that iteratively. CPUs were not designed for this at scale. GPUs, with thousands of cores optimized for matrix math, provide orders of magnitude more throughput per watt for these workloads.
But hardware by itself is not enough. CUDA, the programming model we developed, unlocked that hardware by making it accessible to software developers. CUDA turned a hardware innovation into an industry platform. It created a virtuous cycle. As more developers adopted CUDA, more libraries and frameworks appeared, which accelerated the development of new AI models and applications.
Why this matters now
The present moment is not just about faster chips. It is about systems, networks, and software stacks. I emphasized that accelerated computing is the treasure of our company. That word was not hyperbole. When you have a platform that can run across data centers, labs, robotics platforms, and edge devices, you unlock an enormous number of use cases.
AI is moving from prototypes and research papers into production across industries. That shift is why I said this moment has now arrived. Companies and governments are transitioning to AI first designs. When that happens, the infrastructure demands change. They demand dense compute, high bandwidth networks, and software ecosystems designed to take advantage of parallel accelerators.
🏛️ GTC in Washington, D.C. and National Implications
Holding GTC in the nation’s capital was deliberate. I wanted to highlight the role that advanced computing and AI will play in the economic and national security fabric of the United States. Telecommunications, defense, healthcare, finance, and scientific research all depend on reliable, secure, and advanced computation.
Telecommunications is especially critical. It is the backbone of the economy, the lifeblood of industries, and a core element of national security. At the conference I announced a major partnership with Nokia that is fundamentally based on accelerated computing and AI, aimed at keeping America at the center of the next revolution in wireless technology, including the development of 6G.
Investments in national AI infrastructure are not just about raw compute. They are about creating sustainable ecosystems where startups, academia, and large companies can innovate. I stressed the need for the United States to lead in open source so that our country and our startup ecosystem can continue to thrive. This is a strategic priority. Open source is the connective tissue that accelerates innovation across the economy.
🔗 Quantum and NVQ Link: Bringing Quantum Control to a Broader Ecosystem
Quantum computing is another domain where accelerated classical computing plays an essential role. Quantum processors are incredibly powerful for certain problem classes, but they do not replace classical systems. Instead, they form complementary capabilities that must be orchestrated together.
Today I announced the NVQ link, an interconnect designed for quantum computer control and calibration. The NVQ link is intended to enable quantum error correction and to support hybrid simulations where classical GPUs and quantum processors work in concert. In practical terms, this means creating a control plane that can manage delicate quantum hardware while leveraging classical accelerators for preprocessing, postprocessing, and hybrid algorithms.
I was proud to announce that 17 different quantum industry companies are supporting NVQ link. This broad industry support is a validation of the approach, and it reflects a shared belief that accessible, scalable quantum-classical integration is essential to move the field forward. Quantum error correction and control are not trivial engineering problems. They require high bandwidth, low latency, and deterministic behavior across complex systems.
By providing a standardized interconnect, we hope to reduce the friction in developing quantum applications, and to accelerate real world adoption of quantum-classical hybrid workflows. This is an example of how accelerating classical compute can amplify emergent quantum technologies.
⚡ Department of Energy Partnership: Seven New AI Supercomputers
One of the announcements that I am most excited about is our partnership with the United States Department of Energy to build seven new AI supercomputers. These systems will be distributed across national labs and facilities and will serve as engines for scientific discovery, climate modeling, materials science, energy research, and more.
Why is this important? Because AI is a research multipurpose tool. Historically, large scale computing has transformed science. From weather modeling to particle physics, supercomputers have enabled simulations and analyses that were once impossible. AI adds a new dimension. It can accelerate hypothesis generation, discover new materials and drugs, optimize energy systems, and help scientists extract signals from vast datasets.
These new AI supercomputers are designed to be more than just big clusters. I described them as AI factories. The idea of the AI factory is that a system is highly optimized to run AI workloads continuously. It is unlike a data center of the past. In the past a data center ran many heterogeneous workloads. These new systems will run one thing predominantly. They will run AI, and that focus allows us to optimize hardware, software, power delivery, cooling, and network architecture to an unprecedented degree.
When scientific institutions and national labs have access to these systems, their ability to contribute to national priorities multiplies. That is why the Department of Energy partnership is not just a number of machines. It is a strategic investment in national capacity, competitiveness, and resilience.
🏭 Manufacturing, Reindustrialization, and Jobs
At GTC I underscored a message that I feel strongly about. We are manufacturing in America again. The renaissance in advanced manufacturing matters. It has consequences for supply chains, for quality control, for skilled jobs, and for national resilience.
Reindustrialization does not mean returning to the past. It means building modern production lines for silicon, advanced packaging, and the complex systems that integrate chips with optics, cooling, and software. When we invest in domestic manufacturing, we capture the upstream and downstream economic activity. We create high quality jobs that pay well and require advanced skills, and we strengthen our capacity to respond to geopolitical disruptions.
Manufacturing also underpins innovation. When design and production sit close together, iteration cycles shorten. Engineers can prototype, test, and refine hardware in collaboration with software teams, which accelerates product cycles and reduces time to market. That is why I said with pride that it is incredible to see manufacturing momentum in America.
🔓 Open Source and National Leadership
I do not hesitate to say that our country depends on open source leadership, and our startups depend on it too. Open source provides a common foundation for developers and researchers to build upon. It reduces duplication, accelerates adoption, and helps small teams compete with large incumbents.
My commitment is for NVIDIA to continue investing in open ecosystems and to ensure that the essential building blocks for AI remain accessible. This is not simply altruism. It is strategic. When an entire ecosystem is open and collaborative, innovation accelerates across the economy. Startups can build differentiated services on top of shared standards. Research progress propagates faster. And national competitiveness strengthens as more teams contribute to and benefit from shared infrastructure.
Open source also provides transparency, reproducibility, and trust. These qualities are especially important when AI technologies are integrated into systems that affect public safety and national security. Open models, open tools, and open benchmarks help create a robust dialogue about how to deploy AI responsibly and effectively.
🚗 Robotics and Autonomous Vehicles with Uber and Drive Hyperion
Autonomous systems and robotics are practical embodiments of accelerated computing in the physical world. I announced a partnership with Uber to connect our NVIDIA Drive Hyperion cars into a global network. This collaboration is aimed at bringing robotaxi fleets online in cities around the world.
Why is this significant? Autonomous vehicles are not only technical challenges in perception, mapping, control, and safety. They are also system engineering challenges that require robust compute, powerful neural networks, low latency decision making, and reliable communications.
Drive Hyperion is a platform that integrates sensors, compute, and software stacks specifically for autonomous driving. By working with Uber, we aim to move beyond isolated pilots to a global network where robotaxi cars can operate at scale. The goal is to reduce cost, improve safety, and enhance mobility options for urban populations.
These systems also serve as testbeds for other robotic applications. Technologies developed for perception, sensor fusion, and compute efficiency will translate into logistics, warehousing, and industrial robotics. The road to fully autonomous robotaxis is long, but networked deployments and partnerships accelerate the path to meaningful scale.
📡 Telecommunications, Nokia, and the Path to 6G
Telecommunications infrastructure is entering a new era driven by AI and accelerated computing. I emphasized the strategic partnership with Nokia, rooted in a shared belief that AI will be central to future wireless systems. The next generation of networks, often referred to as 6G, will require intelligent orchestration, real time optimization, and edge compute that can process complex models near the point of data creation.
Nokia brings decades of expertise in network engineering, spectrum management, and global operations. By combining that domain knowledge with accelerated computing and AI, we aim to create networks that are more adaptive, efficient, and secure. Use cases include low latency services for industrial automation, enhanced public safety communication, and new classes of consumer and enterprise applications.
Building the next generation of wireless networks is not just about radio frequencies. It is about end to end systems that integrate baseband processing, network slicing, orchestration, and AI driven optimization. That is why partnerships between compute companies and telecom equipment vendors are so important. They bridge the gap between silicon and systems, and they help ensure that new networks can meet the demands of future applications.
🧠 AI Factories Versus Traditional Data Centers
One of the metaphors I used during the keynote was the term AI factory. This concept captures a transformation in how organizations think about compute infrastructure. A traditional data center is a multipurpose facility that runs diverse workloads from storage to web services to batch processing. An AI factory is optimized for one dominant workload: AI.
When an entire data center is specialized for AI, every design choice can be tuned to maximize throughput, reduce latency, and increase energy efficiency for AI tasks. This includes choosing accelerators that deliver high matrix multiply performance, optimizing interconnects for large model parallelism, and designing cooling and power systems to handle concentrated thermal loads.
These factories will look different than the data centers built for general IT. They will be denser, with more power per rack. They will use high bandwidth interconnects to support distributed training. They will incorporate specialized storage tiers for huge datasets. And they will need software stacks that automate model training pipelines and lifecycle management.
For companies that depend on AI, this shift will enable shorter development cycles, more accurate models, and ultimately better products. For research institutions, AI factories can turn months of experiments into days or hours. The ability to iterate faster accelerates discovery and improves outcomes.
👥 AI and Augmenting Labor
I made a point to highlight the economic dimension of AI: this technology augments labor. That is not an abstract claim. It is a practical observation based on how AI is being applied in industry. Rather than replacing humans across the board, AI often enhances productivity by automating repetitive tasks, providing decision support, and enabling workers to focus on higher value activities.
Consider practical examples. In manufacturing, AI can inspect parts for defects faster and more consistently than human inspectors. That improves quality and reduces waste. In healthcare, AI can help analyze medical imagery to flag areas of concern, which allows clinicians to spend more time on patient care and less time on routine analysis. In customer service, conversational AI can handle basic inquiries while routing complex issues to human specialists.
Augmentation also opens opportunities for workforce development. As AI systems handle low level tasks, workers can be retrained for roles that require creativity, judgment, and interpersonal skills. The economic outcome is not zero sum. With the right policies, investments in training, and emphasis on human centered design, AI can contribute to economic growth and improved job quality.
🌐 Security, Resilience, and National Competitiveness
AI and accelerated computing are central to national security and economic resilience. Advanced compute enables better modeling of threats, more precise simulation for defense technologies, and faster analysis of intelligence data. That is why partnerships between industry and government are so critical.
Security is not just about protecting networks and systems. It is also about ensuring that critical supply chains for semiconductors, advanced packaging, and specialized components are resilient. I discussed the need for domestic manufacturing and strategic investments that reduce single points of failure in the global supply chain.
National competitiveness in AI depends on a healthy innovation ecosystem. This includes strong research institutions, accessible computational resources, supportive open source frameworks, and startup-friendly policies. The Department of Energy AI supercomputers and the partnerships we announced are pieces of a broader mosaic that can sustain long term leadership.
🧩 Building Ecosystems: Partnerships and Industry Support
Technology advances fastest when ecosystems form. The announcements I discussed reflect a commitment to building those ecosystems. The NVQ link has the support of 17 quantum industry players. Our work with Nokia ties compute to telecom. The Department of Energy systems integrate national labs. The Uber partnership connects Drive Hyperion into real world mobility services.
These partnerships have tangible effects. They reduce fragmentation, deliver interoperable systems, and encourage shared standards. When companies collaborate rather than reinventing the same components, the pace of innovation accelerates. I believe this collaborative approach is essential for tackling big multidisciplinary problems that no single organization can solve alone.
🧾 What This Means for Startups and Developers
I wanted to be explicit about the implications for startups and developers. Open source and accessible infrastructure lower the barrier to entry. When a small team can access robust libraries, pre trained models, and affordable compute, they can build competitive products faster.
For startups this reduces capital intensity. They can prototype, iterate, and scale without needing to make massive up front investments in hardware. For developers it means better tools, more shared knowledge, and an ecosystem that rewards creativity. My commitment to open source is part of this broader vision to ensure that innovation remains distributed and inclusive.
🔭 Looking Ahead: The Future of AI and Accelerated Computing
In the keynote I tried to strike a balance between immediate practicalities and long term vision. The short term requires building AI factories, partnerships, and systems that deliver concrete value. The long term is about a transformation that affects science, industry, and society.
We are entering an era where computing accelerators are embedded across the economy. From edge devices to national labs to quantum systems, a coherent architecture will emerge that enables seamless hybrid workflows. These systems will have far reaching impacts, from discovering new materials and drugs to optimizing energy grids and making cities safer and more efficient.
I am optimistic, and I am realistic. There will be technical challenges, ethical considerations, and policy debates. But the trajectory is clear. The combination of GPUs, interconnects, standards, and collaborative ecosystems is unlocking capabilities that were once speculative.
🗣️ Memorable Lines and Core Messages
I want to highlight a few lines that capture the tone of the event and the priorities I emphasized. They serve as shorthand for the strategic pillars I described.
- "This is the Super Bowl of AI" captures the scale and intensity of the moment. It is a moment of competition and collaboration where winners will define future industries.
- "Accelerated computing, its moment has now arrived" is a statement about timing. We are no longer preparing for a distant future. The technology is here and being deployed at scale.
- "We invented the GPU. We invented the programming model called CUDA" is a reminder of the lineage of innovation that enabled the current AI boom. It also frames the responsibility to continue building foundational technologies.
- "AI factories" is a metaphor for specialized infrastructure optimized to deliver AI at scale.
- "We are manufacturing in America again" underscores the importance of domestic production and the economic benefits of reindustrialization.
📈 Economic and Societal Impact
The economic implications of this technological shift are profound. When industries adopt AI as a core platform, the productivity gains are substantial. This is not a simple story of automation replacing jobs. It is a more nuanced story where AI augments human capabilities and creates new opportunities.
Healthcare can be transformed by accelerating drug discovery and diagnostics. Climate science can benefit from faster simulations and better models for mitigation strategies. Supply chains can become adaptive and resilient. Manufacturing can become more efficient and nuanced, with AI optimizing complex production flows.
At the societal level, we must focus on equitable access to the benefits of AI. That is why open source, accessible infrastructure, and investments in training matter. If the benefits of AI are concentrated in a few companies or regions, we amplify inequality. My message was clear. We must make a concerted effort to democratize access to these capabilities, so that the benefits spread across society.
🔄 Practical Steps and Roadmap
What actions did I outline, explicitly or implicitly, as necessary for the roadmap ahead? They can be summarized as follows.
- Build specialized AI infrastructure The concept of AI factories implies building systems optimized for AI workloads at scale.
- Invest in open ecosystems Support open source tools, models, and standards that lower the barrier for innovation and ensure interoperability.
- Strengthen domestic manufacturing Reindustrialize critical supply chains for semiconductors and advanced packaging to reduce geopolitical risk.
- Partner across sectors Combine expertise from telecoms, quantum companies, government labs, and mobility providers to create end to end solutions.
- Focus on workforce development Ensure that workers have access to retraining and education so they can benefit from AI augmented roles.
- Prioritize security and resilience Make sure critical infrastructure is designed to be robust against adversarial threats and supply chain disruptions.
📚 Examples of Cross Domain Impact
To illustrate how these pillars play out, consider a few cross domain examples that tie together multiple announcements I discussed.
Example 1: Climate modeling and materials discovery. AI factories at national labs can train models that simulate complex physical processes. Quantum accelerators connected via NVQ link could handle specific subproblems that classical simulation struggles with. Together, these systems accelerate the discovery of new materials for batteries or carbon capture technologies.
Example 2: Smart cities and telecommunications. Networks built with Nokia and enhanced by AI can adapt to traffic patterns, optimize energy usage in buildings, and improve emergency response. Distributed AI inference at the edge reduces latency for real time services such as autonomous vehicles, while central AI factories continuously train and refine models using city scale data.
Example 3: Autonomous mobility. Drive Hyperion fleets connected to a global orchestration network can share learning across cities. An incident in one location can inform safety improvements elsewhere. Large scale datasets and shared models reduce the time to deploy robust automated services and make the systems safer and more reliable for passengers.
🧭 Governance, Ethics, and Responsible Deployment
Throughout this transformation we must keep governance and ethics front and center. Powerful technologies require careful thought about deployment. I emphasized the need for responsible design, testing, and regulation that balance innovation with public safety and trust.
Transparency is a critical component. Open methodologies, reproducible benchmarks, and audit trails are necessary to build confidence in AI systems. For systems that touch critical infrastructure, such as telecommunications or transportation, rigorous safety standards and independent verification should be prerequisites for deployment.
Regulation should not be a blunt instrument that stifles innovation. But neither should it be absent. Policymakers, industry, and civil society must work together to create frameworks that encourage innovation while protecting people and critical systems.
📢 Final Reflections and Call to Action
Standing in Washington, D.C., I felt both sentimental and proud. Sentimental because these innovations represent decades of work by engineers and researchers. Proud because the United States remains a fertile ground for invention, and because I believe the partnerships we announced will strengthen our national capacity.
My call to action was simple. We must continue to invest in foundational technologies, collaborate across domains, and adopt an open and strategic approach to building the next generation of computing infrastructure. Whether it is building AI factories, integrating quantum systems through NVQ link, partnering to accelerate telecom modernization, or connecting autonomous vehicles into global networks, the work ahead is collaborative and multidisciplinary.
There will be tough engineering challenges, policy discussions, and societal questions to address. But the path forward is compelling. With focused investments in infrastructure, open ecosystems, and workforce development, we can ensure that the benefits of the AI revolution are widely shared and that the United States remains at the forefront of this transformation.
"Accelerated computing, its moment has now arrived."
I mean that. The technologies we build today will power discoveries and industries for decades to come. My responsibility, and the responsibility of the teams I work with, is to continue building platforms and partnerships that enable these possibilities. At GTC in Washington, D.C., I shared that vision, and the announcements we made were concrete steps on that journey.
✅ Summary of Key Announcements and Themes
Let me close with a concise recap of the most important announcements and the themes they represent.
- GPU and CUDA lineage Recognition that decades of innovation in GPUs and programming models underpin the AI revolution.
- NVQ link for quantum-classical integration A standardized interconnect to handle quantum control, error correction, and hybrid simulations, backed by 17 industry partners.
- Department of Energy partnership Building seven AI supercomputers to accelerate national science and research, described as AI factories.
- Telecom partnership with Nokia A commitment to integrate AI and accelerated computing into the future of wireless networks, including 6G research.
- Manufacturing in America A renewed emphasis on domestic manufacturing, reindustrialization, and job creation.
- Open source leadership A dedication to open ecosystems so that startups and the research community have a strong foundation.
- Drive Hyperion and Uber A partnership to connect autonomous vehicle platforms into a global network of robotaxis.
- AI for economic growth A message that AI augments labor and can drive growth when combined with policy and workforce investments.
🔎 Closing Thoughts
The moment I described at GTC is one of both opportunity and responsibility. Accelerated computing and AI are not just technological trends. They are platforms that reshape how we solve problems, how we create value, and how we ensure security. The announcements made are practical steps to build the infrastructure needed for an AI driven future.
I am optimistic about what we can achieve when industry, government, and academia work together. From quantum to telecoms to robotics, the future will be built by teams that collaborate, share tools, and focus on solving real world problems. That is the spirit I conveyed in Washington, D.C., and it is the path I intend to follow going forward.
