Deloitte's 17th annual Tech Trends report, published in December 2025, identifies five forces reshaping enterprise technology: physical AI and robotics, agentic AI, the AI-first infrastructure reckoning, IT operating model transformation, and cybersecurity in the AI age. The common thread is a shift from experimentation to scaled deployment. This article examines the infrastructure and agentic AI dimensions and explains what they mean for U.S. enterprises.

What Is the 'AI Infrastructure Reckoning' in Deloitte Tech Trends 2026?

Deloitte uses the phrase "AI-first infrastructure reckoning" to describe a specific problem: enterprises built their compute environments for pre-AI workloads, and those environments are now misaligned with production-scale AI demands. The report notes that while AI inference costs have dropped roughly 280-fold over the past two years, total enterprise AI spending has grown because usage has outpaced cost reduction. Some organizations are seeing monthly AI compute bills in the tens of millions of dollars.

The root cause is structural. Cloud-based AI services worked for proof-of-concept projects, but become cost-prohibitive when deployed across enterprise operations with continuous inference requirements. Deloitte frames the response as a three-tier hybrid architecture: public cloud for elastic training workloads and experimentation, private on-premises infrastructure for predictable high-volume inference, and edge computing for latency-critical applications requiring split-second decisions.

The report also calls attention to non-cost drivers behind this reckoning: data sovereignty requirements, intellectual property protection, and resilience for mission-critical applications. Deloitte argues that many enterprise data centers lack the specialized processors, advanced networking, and cooling architectures required for AI workloads. Retrofitting is often more costly and slower than building purpose-built "AI factories" from scratch.

How Is 'AI Goes Physical' Shaping Enterprise AI?

Deloitte's "physical AI" trend refers to AI systems embedded in machines that perceive, reason about, and interact with the physical world in real time. This goes beyond pre-programmed robotics. The report describes adaptive systems that learn from their environments, powered by multimodal vision-language-action models and on-device neural processing units.

The practical applications are concentrated in asset-heavy, task-intensive sectors. Amazon has deployed over one million warehouse robots. BMW uses AI-guided vehicles that drive themselves through production routes. Healthcare organizations are testing smart sensors for real-time patient monitoring. Deloitte's survey of more than 3,200 global business leaders found that about 58% are already using physical AI to some extent, and that number rises to 80% when accounting for planned deployments over the next two years. UBS projects two million workplace humanoid robots by 2035, representing a $30 to $50 billion market. Deloitte warns, though, that the convergence of IT and operational technology significantly expands the cyber attack surface, requiring early investment in security architectures and fleet orchestration.

Who Are the Biggest Players in Agentic AI?

Deloitte tracks agentic AI as a distinct trend, separate from physical AI. Agentic AI refers to software systems composed of autonomous agents that reason toward outcomes, make decisions, and execute tasks without continuous human direction. The report identifies this as a market projected to reach $8.5 billion in 2026 and potentially $35 billion by 2030, with Deloitte noting the total could climb to $45 billion if enterprises address orchestration and governance challenges more effectively.

The companies shaping this space span several categories. Foundation model providers like OpenAI, Google DeepMind, and Anthropic are building the reasoning engines that power agents. Cloud platforms including Microsoft Azure, Amazon Web Services, and Google Cloud provide the deployment and orchestration backbone. Enterprise software companies like ServiceNow, Salesforce, and SAP are embedding agentic capabilities into their products. On the startup side, companies such as Adept AI and Covariant are building agents for specific enterprise and physical-world tasks. Emerging interoperability protocols, including the Model Context Protocol (MCP), Agent-to-Agent (A2A), and Agent Communication Protocol (ACP), are becoming essential for multi-agent coordination.

Defining Multi-Agent Systems and Agentic AI

A multi-agent system is a configuration in which multiple autonomous AI agents interact to solve problems, either collaboratively or through structured competition. Each agent has its own capabilities and objectives. The system's value comes from coordination, where tasks are distributed, executed, and results synthesized without requiring a human to manage each step.

Agentic AI, as Deloitte defines the term, is the deployment of these multi-agent configurations within real-world enterprise and physical environments. The distinction from earlier AI tools is significant. A standard large language model responds to a prompt. An agentic system interprets a goal, decomposes the goal into subtasks, selects tools and data sources, executes across enterprise systems, and adapts when conditions change.

Deloitte's data on adoption underscores both the interest and the difficulty. Among surveyed organizations, 30% are exploring agentic options, 38% are running pilots, 14% have deployment-ready solutions, and only 11% are running agents in production. The report identifies three blockers: legacy systems not designed for agent interaction, the absence of formal agentic strategy in 35% of organizations, and the challenge of redesigning processes for AI-first operations.

Practical Implications for U.S. Enterprises in 2026

Deloitte emphasizes that the organizations pulling ahead are those rebuilding operations from the ground up with measurable outcomes. The report frames this as a widening gap between organizations that redesign work and those limited to automating existing routines.

For U.S. technology leaders, several actions emerge from the data. First, evaluate the cost tipping point where cloud AI expenses reach 60 to 70% of equivalent on-premises hardware costs; at that threshold, private infrastructure becomes economically rational. Second, treat agents as a new category of worker with dedicated frameworks for onboarding, performance tracking, and cost management. Third, address architectural prerequisites: microservices, modular APIs, and modern data pipelines are table stakes for both agentic and physical AI. Fourth, embed governance into AI operations from the start; Deloitte found that organizations where senior leadership actively shapes AI governance achieve greater business value.

Cybersecurity demands parallel attention. The same AI capabilities that create business value also introduce new vulnerabilities, including shadow AI deployments, adversarial attacks, and expanded attack surfaces from physical AI. Deloitte recommends combining defensive AI (automated threat detection, red teaming with agents) with security frameworks designed specifically for AI systems.

Why Deloitte's 2026 AI Trends Matter

The 2026 report marks a turning point in Deloitte's 17-year Tech Trends series. The central shift is from possibility to production. With 64% of organizations increasing AI investments and only 1% of IT leaders reporting no significant operating model changes underway, the direction is set. The competition now centers on execution speed and quality.

Infrastructure maturity and agentic AI readiness are the two areas where U.S. enterprises face the most pressing decisions. Organizations that invest in purpose-built compute, adopt hybrid architectures, and redesign work for human-agent collaboration will hold structural advantages over those still running pilots on infrastructure built for a different era.

Frequently Asked Questions

What are the five trends in Deloitte Tech Trends 2026?

The five trends are physical AI and robotics, agentic AI, the AI-first infrastructure reckoning, IT operating model transformation, and cybersecurity in the AI age.

What does 'AI infrastructure reckoning' mean?

The term refers to the mismatch between existing enterprise compute environments and the demands of production-scale AI. Organizations need to rethink compute strategy across cloud, on-premises, and edge tiers.

How large is the agentic AI market expected to be?

Deloitte projects the global agentic AI market at $8.5 billion in 2026, growing to $35 billion by 2030, or up to $45 billion if governance challenges are addressed effectively.

What percentage of organizations are running agentic AI in production?

According to Deloitte's 2025 Emerging Technology Trends survey, only 11% of organizations are actively using agentic AI systems in production. Another 14% have deployment-ready solutions, while 38% are running pilots.

What is physical AI?

Physical AI refers to AI systems that enable machines to perceive, reason about, and interact with the physical world in real time. Applications include adaptive robotics, autonomous vehicles, and IoT sensor networks.