Agentic AI: The Next Investment Layer Beyond Nvidia

Introduction

Nvidia CEO Jensen Huang called it "the turning point for agentic AI" during a recent earnings call, and this time, the hyperbole may be apt. Chapter One of the AI investment cycle—the buildout of GPU clusters, data centers, and network infrastructure—is not yet over, but Chapter Two has quietly begun, with an investment logic entirely different from the conditions that made infrastructure the obvious choice in 2023–2024. Agentic AI refers to AI systems capable of autonomously executing multi-step tasks—they can plan, reason, take actions in software environments, and adjust based on new information without step-by-step human guidance. These are not chatbots; they are software agents that can browse the web, execute code, interact with APIs, and manage workflows, completing complex tasks that previously required skilled labor. Jeff McMillan of Morgan Stanley described the Silicon Valley scene at a 2026 internal research meeting: "Overnight, every company became an agentic company." He added that many remain visions, but the direction is clear, and the investment impact is already showing in revenue figures.

Event Background

The AI infrastructure layer has already delivered exceptional returns: Nvidia captures roughly 90% of AI accelerator spending, at an annual run rate of approximately $1.8 trillion. Demand for data center operators, power infrastructure, and networking companies has surged. PineBridge and MetLife describe data center equipment growth as "basically locked in for the next four to five years," with annual growth of about 25%. The structural demand for computing is beyond doubt. However, the market is increasingly focused on: Can the application layer generate enough revenue to justify infrastructure investments? Sequoia's $600 billion revenue gap analysis captures this anxiety. If an answer exists, it will largely come from enterprise software: AI-native applications, AI-enhanced workflows, and agentic systems that can significantly reduce enterprise labor costs or increase revenue. The companies that build and distribute these applications are the logical next beneficiaries in the AI investment cycle.

Digital Economy Analysis

The economic significance of agentic AI lies in its transformation of AI from a productivity tool into a labor substitution solution. Compared to copilot or assistant AI, the autonomy of agents gives them a much larger value proposition. A copilot that helps knowledge workers write emails or summarize documents marginally enhances individual productivity; an agent that manages a complex multi-step workflow—analyzing customer accounts, identifying issues, generating solutions, drafting communications, executing recommended actions, and recording results—replaces or substantially augments an entire class of knowledge work. This means enterprises are willing to pay significant software subscription fees for agents that truly deliver, a stark contrast to MIT's Project NANDA finding in 2025 that $30–40 billion in enterprise AI spending generated no measurable P&L impact. When agentic AI works, its P&L impact is easy to measure and hard to dispute.

Business Model ObservationsFrom a revenue model perspective, agentic AI has spawned several new business models: - Subscription-based agent services: Charged per seat or per task, such as Microsoft Copilot achieving $37 billion in annualized revenue (Q1 2026, up 123% year-over-year). - Outcome-based pricing: Revenue sharing based on labor cost savings or revenue increases from the agent. - Platform add-on model: Existing enterprise software (e.g., ServiceNow, Salesforce) adds agentic capabilities on top of core platforms, charging additional fees.

The core of these models is value-based pricing—companies are willing to pay for quantifiable savings, which was difficult to achieve during the AI pilot phase (2023-2024).

Market Competition Analysis

The application layer of agentic AI is more fragmented than the infrastructure layer. Nvidia, with approximately 90% market share and a nearly insurmountable moat, has become a perfect vehicle for infrastructure investment. In the application layer, dozens of companies compete across different segments for enterprise deployment. The highest-conviction investment targets typically have three characteristics: 1. Distribution moat: The ability to deploy AI products at scale through existing enterprise relationships (e.g., Microsoft, Salesforce). 2. Vertical specialization: Deep domain expertise in specific workflows being automated, making products difficult to replicate by general-purpose agents (e.g., clinical documentation in healthcare, compliance monitoring in financial services). 3. Proven revenue momentum: Not just pilot projects, but genuine annual recurring revenue (ARR) growth from paying enterprise customers.

Companies that meet these three criteria are not necessarily the largest or most well-known names. Some of the most attractive agentic AI investment opportunities exist in mid-sized enterprise software, vertical SaaS, and cybersecurity—companies that have not yet achieved Nvidia-level valuations but show early and genuine revenue growth, indicating they have found product-market fit in the agent layer.

Data and Regulatory Impact

• Agentic AI presents new challenges for data governance. When autonomous agents access enterprise systems and external data, risks around data privacy, security, and compliance increase. Regulations such as the EU AI Act may require transparency and auditability reviews of agent actions. In cross-border data flows, agents calling APIs in multiple countries may trigger data localization requirements. Future regulation is likely to focus on:
• Interpretability of agent decisions
• Application of data minimization principles in automated workflows
• Human oversight and intervention mechanisms

These will affect the deployment speed and cost of agentic AI.

Global Trend ObservationAgentic AI is not a short-term event but a deepening of the long-term trend in the AI economy. It sits at the intersection of the "AI Economy" and the "Platform Economy." In vertical domains such as finance, healthcare, and software development, agentic AI is creating economic value at a substantial pace. Infrastructure investment will continue, but over the next 18-24 months, risk-adjusted opportunities increasingly lie in companies that convert infrastructure into revenue. Identifying them before the market fully prices them in is both the challenge and opportunity that defines the second chapter of the AI investment cycle.

DigitalEcoNews Insight

The rise of agentic AI marks a shift from AI as a "tool" to AI as "labor." For enterprises, this means reassessing cost structures: the tipping point for the human-machine collaboration cost ratio has arrived. For investors, the fragmentation of the application layer demands more refined screening, but potential returns may surpass infrastructure investment. For the global digital economy, agentic AI will accelerate the automation of knowledge work, reshape the white-collar job market, and foster new platform ecosystems. Over the next decade, companies capable of deploying trustworthy, auditable agents at scale will dominate the next wave of value creation in the digital economy.