AI Agents and the Companies That Will Build Them
From Chatbot to Co-Worker
In 2025, you typed a question and got an answer. In 2026, you describe a goal and an AI agent goes and does it. A chatbot saves you a Google search. An agent replaces a workflow that used to require a person. Klarna deployed an AI agent that handled 2.3 million customer service conversations, cutting resolution time from 11 minutes to under 2 minutes.
By end 2025, enterprise AI agent revenue hit $13 billion, up from $5 billion a year earlier. Over 68% of organizations plan to integrate agents by 2026.
Five Companies, Five Approaches
The agent platform war features five companies building competing visions: OpenAI (consumer-first), Anthropic (developer infrastructure), Google (integrated search), Microsoft (enterprise workflows), and Salesforce (CRM-native).
| Company | Agent Product | Approach | Key Advantage |
|---|---|---|---|
| OpenAI | ChatGPT Agent + Operator | Consumer-first autonomy | 900M+ weekly users |
| Anthropic | Claude Agent SDK + MCP | Developer infrastructure | MCP standard (10,000+ servers) |
| Gemini Deep Research | Integrated search + cloud | Existing enterprise base | |
| Microsoft | Copilot Studio | Enterprise workflow | Office 365 + Azure distribution |
| Salesforce | Agentforce 360 | CRM-native agents | Agentforce ~$800M ARR |
Anthropic built the plumbing via Model Context Protocol (MCP), now adopted by OpenAI, Google, and Microsoft as the industry standard. Microsoft owns distribution through Office 365. Google owns enterprise infrastructure. Salesforce owns the CRM layer, with Agentforce now past $800 million in ARR.
The Agent Stack: Four Layers of Value
The agent stack distributes value unevenly across four layers: compute, orchestration, application, and monitoring. Understanding company positioning tells you what needs to go right for the stock to work.
Who's Already Making Money
Several companies generate significant agent-related revenue. Here's where money flows currently:
Cursor hit $4 billion ARR automating code, then SpaceX agreed to buy it for $60 billion. Enterprise AI agent revenue reached $13 billion in 2025, up 160% YoY. Nearly half of global venture capital in 2025 went to AI. Top 55 AI startups each raised over $100 million.
The Picks-and-Shovels Play
Regardless of which agent platform wins, infrastructure companies benefit from all of them:
| Category | Public Companies | Why They Benefit |
|---|---|---|
| Cloud Infrastructure | AWS (AMZN), Azure (MSFT), GCP (GOOG) | Every agent runs on cloud |
| Semiconductors | NVIDIA, AMD, TSMC | Training + inference demand rising |
| Networking | Arista, Marvell, Astera Labs | Data movement is bottleneck |
| Power & Cooling | Vertiv, Eaton | Data centers at capacity |
| Testing & Validation | KLA, Teradyne | 48% growth from AI demand |
| Enterprise Automation | UiPath, ServiceNow | Agent + workflow integration |
As agents make thousands of API calls per task, networking becomes the performance bottleneck. Arista, Marvell, Astera Labs are positioned at this chokepoint. Testing equipment companies saw 48% growth in 2025 from AI chip verification.
The Readiness Gap
83% of organizations plan to deploy AI agents, but only 29% feel ready securely. This 54-point gap is the biggest constraint on 2026 adoption.
Multi-turn prompt injection attacks achieve 92% success rates. 90% of deployed agents are over-permissioned. Non-human identity attacks are the fastest-growing threat. This gap creates two implications: adoption will be slower, and companies solving security own the next bottleneck.
Regulatory tightening: U.S. issued RFI on agent security in January 2026. ISO 42001, NIST AI Risk Management, GDPR controls for autonomous systems all in development. Companies building compliance tools will have guaranteed customers.
Three Ways to Invest in the Agent Shift
Strong Bets
Cloud infrastructure spending will keep growing. Agents need compute. Networking layer underappreciated. Enterprise workflow companies capture agent revenue first. Security and compliance tools become mandatory.
Risky Bets
Single platform winning entire market. Consumer adoption matching enterprise. Startups sustaining growth post-IPO. Open-source frameworks not commoditizing. Valuations holding through correction.
Methodology
