Introduction

Those who have followed us for a few years have heard us talk about AI agents, or the digital janitors. They have also often heard us say "slowly, slowly, suddenly." The reason we're creating this note is because we're now singing the last verses of "slowly." The power that AI agents can bring will create disruptions in most areas.

Our purpose, as always: to present our perspective from where we sit on the edge watching these technologies develop. We are not technical experts or academic minds. We look at technologies and their forces from a financial perspective. If you want absolute truths, stock recommendations, and academic precision, you should stop here. We have nothing to give. But for the rest of you who are curious and seeking perspectives on the future, you get ours here. At least as we see it in the summer of 2025.

We use calculators, Excel, Word, language models, websites, and conversations with companies and analysts as the foundation for our "Disruptive Perspectives" notes.

Understanding the Concept of "Agents" in Philosophy, Economics, and Sociology

Philosophy

In philosophy, an agent often refers to an actor with the ability to act purposefully and make decisions based on intentions, rationality, or free will. Within agency theory, an agent is one who can initiate actions and influence their surroundings. Philosophers like Kant and Sartre have emphasized the agent's autonomy and moral responsibility, while newer perspectives, such as posthumanism, challenge the distinction between human and non-human agents (e.g., machines or AI) by attributing agency to technology.

Economics

In economics, particularly in principal-agent theory, an agent is a party that acts on behalf of a principal, often under asymmetric information. For example, a business executive (agent) acts for shareholders (principal). Economic models often assume that agents are rational and maximize utility or profit, but behavioral economics (e.g., Kahneman and Tversky) challenges this by including irrationality and cognitive biases. In game theory, agents are modeled as strategic actors who make choices based on others' actions.

Sociology

In sociology, agency concerns the individual's or group's ability to act independently and shape social structures - as in Giddens' structuration theory, where agency and structure are mutually dependent. Agents can be individuals, organizations, or technologies that influence social systems. For example, an AI agent in a social network can shape communication or behavioral patterns.

In our world, the concept of "agent" is about the ability to act purposefully, influence surroundings, and make decisions, whether as a human, organization, or technology. Philosophy emphasizes autonomy, economics focuses on rational choices and incentives, and sociology underscores the interplay between agents and social structures.

NVIDIA's Use of the Term "Agents"

NVIDIA defines agents as autonomous assistants, either digital (AI) or physical (robots), that can perceive their surroundings, make decisions, and perform actions to achieve goals.

NVIDIA's agents build on principles from AI and machine learning, where agents learn through data, sensors, and algorithms to optimize actions. This connects to philosophy's idea of autonomy (agents as independent decision-makers), economics' focus on optimization (maximizing efficiency as a goal), and sociology's perspective on how agents influence systems (e.g., robots in workflows or digital agents on social platforms).

The AI agent is what you and I as users encounter. How we encounter the AI agent depends on what we call the interface, which can be a PC, TV, mobile phone, and car. And soon in the AR glasses on your nose.

Historical Development, Current Status, and Future Use

Historical Timeline

• 1940-50s: The beginning of AI with Turing and early computers, laying the foundation for autonomous agents
• 1980s: The emergence of multi-agent systems in AI, where agents interact in complex environments (e.g., game theory and simulations)
• 2000s: NVIDIA's development of GPUs revolutionized parallel data processing, enabling advanced AI agents
• 2010s: Deep learning and NVIDIA's CUDA platform accelerated the development of digital agents for image processing, language models, and simulations
• 2020s: NVIDIA's Omniverse and Isaac Sim introduced platforms for developing both digital and physical agents, used in everything from simulating virtual worlds to autonomous robots

Current Status

Future Use

Digital agents are expected to dominate in virtual worlds (metaverse), customer service (chatbots with human-like interactions), and simulations for research, such as climate/weather models or health. Google and Meta's strong focus on AR glasses in recent years can be seen in light of the development of digital agents.

When we were first introduced to the internet, it was via a browser. Those who controlled the browser controlled the user's gateway to the web. This time, the battle will move from hands and keyboard up to the nose tip, controlled by your voice commands. This will, unlike the browser, be intelligent. It will remember who you are, it will know you, and you will never replace it. We think. This battle, which we call "The Battle of Metaverse 2.0," probably starts with launches from Google and Meta in 2026-27. And it's a fight against Apple's hegemony over the mobile phone.

Physical agents: Autonomous robots will play a larger role in industry, transport, and healthcare, with improved sensing and decision-making capabilities. In spring 2025, the first humanoids began working on factory floors with repetitive tasks. Companies developing, designing, and selling humanoids will start pilot production during 2025. Our new colleagues and buddies will probably become both more numerous and many over the next five years. Their goal is to do exactly the same as us, only better, cheaper, and faster.

Categories of Digital Agents

By Technology

By Use Cases

• Customer Service: Virtual assistants that handle inquiries
• Content Production: AI that generates material for marketing, articles, or design
• Data Analysis: Agents that analyze large amounts of data and assist as decision support
• Virtual Simulations: Agents used in games, metaverse, or research

AI Agents in Your Glasses

By 2027, glasses will have gained a new function. Not just for vision, but as an interface to a reality where the physical and digital merge. Mixed Reality (MR) glasses, also called AR/VR glasses, have developed into the next big thing for tech giants, startups, and pure AI companies. A big question that arises then is: who owns the user experience and who gets access to it?

The Hardware-Platform Connection

Hardware and the platform (software) are tightly connected. One determines what the user sees and feels, the other determines which digital actors get to contribute functionality and intelligence. For AI agents, it's crucial whether they get access to sensors, processing power, and context, and to what degree they can operate freely.

The Major Players

Platform Strategies and Agent Access

It's not enough to look at the glasses' specifications. The platforms behind them determine to what extent AI agents can contribute to the user experience:

• Apple: Maintains its closed ecosystem, with limited access and strictly curated interfaces
• Meta: Offers greater flexibility, with open SDK and support for third-party AI but still within the framework of its own priorities
• Google: Focuses on technological openness, but with binding to its own infrastructure
• Startups: Open the door completely, but lack depth in both software and distribution

Looking Further Ahead

Toward 2029, we expect a more open structure. Standardized protocols and regulatory pressure may force AI agents to move between platforms (called interoperability) without loss of functions. Perhaps agents will create a common language, like HTML is for web pages. Then smart, self-learning assistants can work on all glasses - regardless of brand or platform.

If xAI establishes its own platform and builds from the ground up with AI at the center, they could become a real challenger. The ideal winner has high technical performance on a completely open platform. Maybe that's what xAI is trying. Time will tell. The question is no longer just who builds the best glasses. It's also who opens the door for future AI agents, and who keeps them out.

Leading Companies in the Digital AI Agent Ecosystem

Winners and Losers

These are of course rough categories where the idea is to sketch possible consequences of the digital agents' march of conquest.

Winners

Among companies, it's particularly the tech giants NVIDIA, OpenAI, Google, and Microsoft that lead, while many startups build niche AI agents for various use cases. At the individual level, there's high demand for data scientists, AI developers, and specialists on ethical questions around technology. Roles related to digital innovation and change management are also becoming increasingly important as AI changes how we work and make decisions.

Losers

This can provide higher efficiency and lower costs for businesses, but also creates challenges. Unemployment, need for retraining, and greater inequality in access to opportunities may be the consequences. In addition follow important ethical questions about surveillance, algorithmic bias (AI making unfair decisions, often driven by "bad data"), and accountability for decisions made by digital systems.

Physical Agents: The Robot Revolution

While digital agents transform our virtual world, physical agents - robots and autonomous systems - are reshaping the physical world. The convergence of AI, advanced sensors, and mechanical engineering is creating a new generation of machines that can work alongside or replace humans in various tasks.

Categories of Physical Agents

Leading Companies in Physical Agents

• Tesla: Developing Optimus humanoid robot alongside their autonomous vehicle technology
• Boston Dynamics: Advanced robotics with Atlas, Spot, and Stretch robots
• ABB & FANUC: Industrial robotics leaders with millions of robots deployed globally
• Intuitive Surgical: Da Vinci surgical robot system used in hospitals worldwide
• Amazon Robotics: Warehouse automation with over 750,000 robots in operation
• BYD & CATL: Chinese leaders in manufacturing automation and battery production

Challenges in Collaboration Between Digital and Physical Agents

As digital and physical agents become more prevalent, new challenges emerge in how they collaborate with each other and with humans. These challenges mirror classic organizational theory but with unprecedented complexity.

Key Challenges

Parallels to Organizational Theory

The challenges of agent coordination mirror classic organizational problems. Just as Max Weber described bureaucratic hierarchies, we're now seeing algorithmic hierarchies emerge. The principal-agent problem from economics takes on new dimensions when both principals and agents can be artificial.

We're moving from "System of Record" (passive data storage) to "System of Engagement" (active interaction) to what might be called "System of Agency" - where systems don't just record or engage but act independently on our behalf.

Digital and Physical Agents in AI Training and Inference

Understanding the distinction between AI training and inference is crucial for grasping how agents operate and evolve.

The shift toward edge computing means more inference happens locally - in your phone, car, or AR glasses - rather than in distant data centers. This reduces latency, improves privacy, and enables real-time applications. By 2029, we expect most AI agent interactions to happen at the edge, with only complex tasks requiring cloud connectivity.

Where Are AI Agents Located?

AI agents exist across a spectrum from cloud to edge, each location offering different capabilities and trade-offs.

Everyday Tasks AI Agents Can Solve

• Schedule optimization and calendar management
• Real-time language translation in conversations
• Personalized health monitoring and recommendations
• Automated financial management and investment
• Content creation and editing
• Home automation and energy optimization
• Personal shopping and price comparison
• Learning assistance and tutoring

From Reactive to Proactive

The evolution of AI agents follows a clear trajectory:

By 2029, we expect most personal AI agents to operate at the proactive level, anticipating your needs based on context, history, and patterns. Your AR glasses might book a restaurant when detecting you're discussing dinner plans, or your health agent might schedule a doctor's appointment when detecting concerning symptoms.

Moore's Law and the Agent Revolution

Moore's Law - the observation that transistor density doubles approximately every two years - has been the drumbeat of the digital revolution. But for AI agents, we're seeing something beyond Moore's Law: the convergence of multiple exponential improvements.

This compound exponential growth means that capabilities we think are 10 years away might arrive in 3-5 years. The "slowly, slowly, suddenly" pattern we mentioned at the beginning is driven by these overlapping exponentials reaching inflection points simultaneously.

Our Disruptive Investment Strategy in a World Where Agents Are Coming

As agents reshape every industry, our investment strategy focuses on identifying companies positioned to benefit from or enable this transformation. We look for businesses that either build the infrastructure for agents, deploy agents effectively, or solve problems that agents create.

Investment Categories

Investment Principles for the Agent Era

Timing the Agent Revolution

We're in the "late slowly" phase of agent adoption. Key indicators we're watching for the "suddenly" phase:

• AR glasses reaching 10 million units sold annually (expected 2026-2027)
• Humanoid robots in regular production use (starting 2025)
• Agent-to-agent transactions exceeding $1 billion annually (2026-2027)
• Regulatory frameworks for autonomous agents established (2025-2026)
• Consumer AI agents managing >$100 billion in assets (2027-2028)

Conclusion: The Agent-Driven Future

Digital and physical agents represent one of the most significant technological shifts in human history. We're not just automating tasks - we're creating a new form of intelligence that can perceive, decide, and act on our behalf. This isn't science fiction; it's happening now, accelerating from "slowly" to "suddenly."

The investment implications are profound. Traditional business models will be disrupted. New monopolies will emerge. The companies that successfully deploy agents will achieve unprecedented scale and efficiency. Those that don't will become obsolete.

But beyond the financial opportunities, we're facing fundamental questions about the future of work, the nature of intelligence, and the structure of society. How we answer these questions - through our investments, regulations, and choices - will determine whether agents become tools of liberation or instruments of a new digital feudalism.

The agent revolution isn't coming - it's here. The armies of digital and physical agents are already being deployed. The question isn't whether they'll transform our world, but how quickly and completely. As investors, we must position ourselves not just for the world that is, but for the world that's rapidly becoming.

Remember: "Slowly, slowly, suddenly." We're approaching the suddenly. Are you ready?