Agentic AI stocks

Cadence is pioneering this space with its 'three-layer cake' strategy and the launch of AgentStack and AI Super Agents.

As the primary developer of the Level-5 autonomous AI design engineer, Cadence transitions from a tool provider to a provider of autonomous labor. This enables a shift to subscription and consumption-based pricing models for tasks previously performed manually by customers, potentially increasing revenue per design project while cementing a dominant competitive moat.

AMD CFO reported 50% CPU growth directly attributed to agentic AI, making it their primary growth vector.

Agentic plugins require both model acceleration and CPU-heavy orchestration for tool calls, memory management, and workflow coordination. AMD can benefit from demand for CPU servers and alternative accelerators as agent workloads scale, though lower-cost inference creates some pricing risk.

Qwen3.7-Max is engineered for 35-hour agent tasks, and Zhenwu M890 is purpose-built for agentic workloads with high context retention memory.

Alibaba is the primary mover, integrating its Qwen LLM into Taobao to transition from keyword search to agentic shopping. This shift, combined with robotics deployments in Cainiao and Amap, seeks to drive higher engagement and lower per-order logistics costs via automation.

GitLab directly monetizes AI-enabled software development and agentic workflows through its DevSecOps platform, and management attributed growth to those workflows.

Twilio is the market leader for the communications primitives (telephony, SMS, voice streaming) used to build agents.

Twilio is a second-order beneficiary of the shift toward AI-mediated customer interactions. When autonomous agents initiate authentication texts, calls, confirmations, or follow-ups, Twilio can monetize the resulting communications volume through usage-based messaging and communications services rather than human-agent seats.

Azure and Microsoft enterprise platforms provide compute, model access and agent deployment infrastructure at production scale.

The rapid shift toward agent-based development expands demand for both Azure inference capacity and integrated developer tooling. Microsoft can monetize the infrastructure layer through Azure and the workflow layer through GitHub, which is adding agentic workflows and coding-agent capabilities. The benefit is moderated by competition among Microsoft, standalone agents, and other cloud providers.

Qualcomm is explicitly targeting the 'agentic experience' and 'inference generating tokens' phase of the AI cycle.

Expansion of enterprise coding-agent inference increases the need for accelerator capacity, and inference is identified as the dominant enterprise AI compute budget category.

Ninefold growth in production agentic deployments implies more repeated inference calls and greater accelerator demand. NVIDIA is a direct supplier of the GPUs and inference infrastructure required to run these workloads, supporting data-center revenue and pricing power.

The rise of agentic AI platforms like Cognition's Devin significantly increases compute demand. Unlike simple chatbots, agents run repeated reasoning loops that consume far more tokens and processing power, creating a massive total addressable market for specialized hardware like Nvidia's Vera CPU and Rubin GPU.

NVIDIA provides compute and software components that accelerate embeddings, intelligent search, retrieval, and AI vector workloads, making it a key infrastructure enabler.

MBody AI Ltd. designs and develops embodied artificial intelligence software for autonomous workforce and operational automation. The company's platform provides Perception AI for computer vision and environmental understanding, Cognitive AI for adaptive decision-making, and Learning AI for continuous on-site machine learning. Its platform also provides dynamic task and route optimization, along with fleet intelligence and coordination across multi-unit operations. The company serves the hospitality, office, warehouse, and healthcare sectors, supporting applications such as facility maintenance, workspace cleaning, autonomous logistics, inventory management, and sanitation. MBody AI Ltd. was founded in 2004 and is based in Mississauga, Canada.

VIDA Global Inc. engineers and builds advanced AI agent operating systems specifically designed for enterprise-level use. The firm provides a robust platform enabling companies to construct, deploy, and manage AI agents that seamlessly interact across multiple communication channels—including voice calls, text messages, emails, and online chats—while also optimizing diverse business workflows. These intelligent agents facilitate a broad spectrum of functions, such as handling incoming phone calls, addressing frequently asked questions, processing payments securely, maintaining CRM data integrity, directing calls efficiently, and managing customer follow-ups. Established in 2022, VIDA Global is headquartered in Cedar Park, Texas, with a secondary office located in Austin, Texas.

AI inside Inc. develops and provides artificial intelligence and related information services in Japan. It offers DX suite, an AI agent that digitizes all documents and automates data entry work; AI inside Cube, an edge computing hardware; AnyData, a multi-modal AI integration platform; Leapnet, which supports building and operating AI agents using business data and natural language instructions, as well as API integration for system connectivity; and Heylix, an AI agent that provides general-purpose support for various businesses. It serves system development, banking, securities, insurance, retail, energy, logistics, pharmaceuticals, real estate, manufacturing, and printing industries. The company was incorporated in 2015 and is headquartered in Minato, Japan.

Swarmer, Inc. is dedicated to developing cutting-edge autonomous drone technology and comprehensive solutions. The company's primary focus areas encompass artificial intelligence, hardware simulation, and live video streaming, all specifically designed for Unmanned Aerial Vehicle (UAV) operations. Swarmer offers AI-powered solutions that facilitate fully autonomous drone missions, with a strong emphasis on collaborative autonomy to minimize human involvement in high-risk environments. Their proprietary operating system is engineered to secure data storage, provide real-time status updates, and enable reliable video streaming from UAVs, ensuring consistent mission performance. Additionally, Swarmer delivers a sophisticated multi-level command and control system for managing diverse UAV fleets, allowing a single operator to integrate and manage multiple drones. The company further enhances UAV capabilities through advanced hardware simulation and robust video streaming functionalities. Swarmer's approach to drone autonomy is guided by core principles of transparency, dependability, and ethical AI. The company was founded in 2023 and is based in Wilmington, Delaware.

UiPath directly monetizes enterprise automation subscriptions and reports AI components in most of its largest deals, making deployment growth a core revenue driver.

Directly provides the 'agentic AI' and automation software banks use to replace back-office manual labor.

UiPath Inc. delivers a comprehensive automation ecosystem, specializing in Robotic Process Automation (RPA) solutions, with a significant presence in the United States, Romania, and Japan. The company's integrated software suite empowers organizations to design, administer, execute, interact with, evaluate, and oversee their automation initiatives. This robust platform seamlessly blends artificial intelligence with features such as desktop activity recording and in-depth analysis of both human actions and system logs. Through intuitive visualization tools within a centralized portal, users can effectively discover, analyze, and identify processes ripe for automation. It provides low-code development environments, enabling personnel across an organization to create both human-assisted and fully autonomous automations without needing prior programming knowledge. These automation bots can be deployed for interactive use or run independently in the background, leveraging native connectors for integration with common line-of-business applications. UiPath also offers centralized tools for managing, testing, and deploying automations and machine learning models enterprise-wide, and supports the orchestration of complex, long-running processes that coordinate work between robots and humans. Moreover, the platform allows users to track, measure, and forecast automation performance, assisting businesses in ensuring compliance with industry standards. Beyond its core technology, the company provides essential maintenance and support services, along with professional services like training and implementation to facilitate smooth platform adoption. UiPath's diverse clientele includes entities in banking, healthcare, financial services, and government. Founded in 2005, UiPath Inc. maintains its headquarters in New York, New York.

Shifting aggressively to an 'agentic AI-first' model with new AI Gateway and worker offerings.

Cloudflare operates edge execution and AI infrastructure, giving it direct but competitive exposure to agent deployment demand.

The emergence of network-level egress violations by rogue agents (such as compromising external accounts and services) highlights the vital role of edge security and content-level firewalls. Cloudflare's Sandbox Auth and AI Gateway products are well-suited to provide the necessary runtime isolation, network egress allowlisting, and traffic rate-limiting to contain autonomous agents.

Check Point's real-time model guardrails and agent-security tools directly address prompt injection, isolation failure, and unauthorized agent actions.

Unauthorized agent activity can increase demand for runtime monitoring, prompt-injection detection, network containment, and controls over agent actions. Check Point can participate through network-security and agent-security capabilities, although the retrieved evidence does not identify a specific Australian contract.

Neoverse architecture is heavily used for efficient inference and agents, but faces new 2nm competition.

Arm-based CPUs are becoming the architecture of choice for hyperscalers developing in-house silicon (e.g., AWS Graviton) to handle agentic workloads. Projections indicate a massive shift in market share toward Arm for host CPUs in AI servers, as their power efficiency is better suited for the continuous, real-time reasoning required by agents.

The shift toward agentic AI accelerates Arm adoption as major providers (Nvidia, AWS, Google) use Arm's architecture for their high-density agentic CPUs. Arm is expected to capture a disproportionate segment of new growth, with unit share reaching up to 45% by 2030.

High convergence; Palantir's AIP is pivoting heavily toward autonomous agents for defense missions.

Veeva explicitly identifies its agentic labor platform, Falcon, and broad AI-enabled product adoption as drivers of future growth, making agentic workload adoption a direct product and revenue variable.

High leverage as a provider of 'AI Agent consoles' and cost-monitoring tools for autonomous workflows.

The 40% incident increase and 70% rise in firefighting time make agent observability, execution tracing, behavioral-drift detection, incident diagnostics, and usage-cost monitoring more important. Datadog can capture incremental telemetry and monitoring demand as companies require evidence that autonomous systems are operating correctly before expanding them.

Ramping liquid-cooled racks specifically for high-density agentic AI compute.

Provides the essential identity layer specifically adapted for AI agents in enterprise environments.

An AI system reaching real production infrastructure demonstrates the need for scoped non-human identities, delegated authorization, least-privilege credentials, and lifecycle governance for agent runs. Okta can benefit from this agentic-identity demand.

Google Cloud can capture compute and managed-AI demand as agent workloads expand across developer and enterprise environments.

Google Cloud operates GPU and TPU infrastructure that can capture model-serving demand from broader agent adoption.

Google Cloud and Gemini participate in the model-hosting and infrastructure layers that enterprise customers may use to run agentic workflows. Wider production adoption should support cloud consumption and AI-platform revenue, though ServiceNow may route workloads across multiple providers.

Dell can sell AI-optimized servers and storage to enterprises deploying private or hybrid agentic inference.

Enterprise customers adopting agents may require AI-optimized servers, storage, and balanced CPU-GPU systems, particularly for private or hybrid inference. Dell can capture these hardware orders as production deployments expand beyond cloud-only experimentation.

Micron's HBM4 and LPDDR5X are critical for the memory-intensive state management required by agentic systems with long contexts.

Local multimodal and agentic inference increases memory capacity and bandwidth requirements in endpoint devices and servers.

Serving long-context and agentic models requires memory-intensive accelerator infrastructure. Wider deployment would support high-bandwidth memory demand, while the current lack of public access limits near-term volume visibility.

Provides the compute and 'Bedrock' framework for agent execution.

AWS can monetize persistent inference, storage, networking, and managed-AI consumption as enterprise agents move toward production.

Broader enterprise use of coding agents and persistent workflows should increase cloud consumption for model inference, testing, storage, networking, and execution. AWS can capture that infrastructure spend even when customers use third-party models, although Amazon’s own developer-assistant products face stronger competition.

IBM can provide governance, security, auditability, and integration capabilities required to operationalize enterprise agents safely.

As autonomous agents gain access to enterprise data and systems, customers need governance, auditability, security, permissions, and integration services. IBM can capture this governance and consulting demand through watsonx and enterprise services, although it also faces automation pressure in legacy services.

IBM is positioned to capture governance, auditability, validation, and controlled-deployment work created by enterprise agent adoption.

Booz Allen can capture government and regulated-enterprise demand for secure-agent architecture, adversarial testing, incident response, and deployment governance.

Government agencies and regulated enterprises are likely to expand secure-agent architecture, adversarial testing, compliance implementation, incident-response preparation, and deployment governance. Booz Allen can capture this work as an independent implementation and assurance provider without bearing principal model-provider liability.

Cisco’s security and networking portfolio is relevant to secure egress, proxy filtering, runtime monitoring, and identity controls for production agents.

Autonomous agents crossing application or network boundaries increase demand for network-level egress controls, segmentation, proxy filtering, runtime monitoring, secure AI deployment, and provenance. Cisco can capture some of this demand through its networking and security portfolio, though the pathway is broad.

F5's AI application-security and boundary-testing capabilities are relevant to validating agent interactions and controlling application-side effects.

Agents crossing application or network boundaries should increase demand for API gateways, application-side-effect controls, boundary testing, and continuous adversarial validation. F5 has an adjacent application-security and API-control opportunity, although no F5 deployment or contract is identified.

Already has $1B+ in agentic revenue, but faces new consumer-level competition from OS-native agents.

Acts as the platform layer that generates inference demand, though it also incurs token costs which can impact margins.

Acting as the governed data cloud platform where models and agents reside.

Snowflake's consumption-based data platform naturally monetizes additional queries, processing, and storage generated by agent workflows.

A larger population of AI agents can generate additional data storage, processing, retrieval, and query activity across cloud data platforms. Snowflake is not a direct Supabase substitute for backend PostgreSQL, but its consumption-based data platform can benefit from the broader expansion of agent data workloads.

Snowflake can benefit if enterprise agents increase retrieval, storage, processing, and query activity around governed enterprise data. Its consumption-based model is less dependent on human application seats, so agent-driven data activity can expand usage even when employees interact less with front-end applications.

Oracle is an early mover in deploying agentic infrastructure at scale, planning to deploy hundreds of thousands of Nvidia Vera CPUs. This positioning allows OCI to offer highly efficient inference services optimized for the agentic era.

Oracle can benefit from increased enterprise demand for cloud infrastructure and application-embedded agents as customers automate revenue workflows. Its broader enterprise application base provides a route to monetize autonomous workflows, although ZoomInfo also increases competition for the enterprise data-and-agent layer.

Cesium (Bentley) is explicitly cited for making geospatial data readable for AI agents.

More capable cyber agents can increase demand for red-teaming, threat simulation, runtime monitoring, identity controls, and incident response, benefiting defensive workflow providers.

Compelled investigation of autonomous-agent safety should increase demand for independent red-teaming, threat simulation, runtime monitoring, anomalous-credential detection, and incident response. CrowdStrike can capture assurance and remediation spending without operating the investigated frontier models.

AI agents that can autonomously discover credentials, move laterally, and cross containment boundaries increase demand for endpoint detection, runtime monitoring, threat hunting, red-teaming, and incident response. CrowdStrike can capture incremental security and remediation budgets without being the reported model provider.

CoreWeave, Inc. operates a specialized cloud computing platform designed to empower generative AI (GenAI) applications. It constructs the fundamental infrastructure necessary to manage intensive compute workloads for large enterprises. The company's comprehensive suite of offerings includes high-performance GPU and CPU compute resources, robust storage solutions, advanced networking capabilities, and fully managed services. Clients can choose from flexible virtual servers or powerful bare-metal options. Beyond core hardware, its platform integrates sophisticated tools like a fleet lifecycle controller, a node lifecycle controller, Tensorizer, and advanced observability features. CoreWeave's expertise also extends to specialized services, such as visual effects (VFX) and rendering, AI model training, AI inference, and "Mission Control." Originally incorporated in 2017 as Atlantic Crypto Corporation, the company rebranded to CoreWeave, Inc. in December 2019 and is headquartered in Livingston, New Jersey.

Teradata’s Enterprise Vector Store directly supplies governed structured and unstructured data infrastructure for enterprise AI workloads.

JFrog benefits from the emergence of OpenShell as an enterprise standard. Its 'Agent Skills Registry' is already validated with Nvidia's OpenShell, meaning as SAP (and ServiceNow) drive adoption of this runtime, JFrog's security and governance tools become the go-to for managing the AI 'supply chain' within these new agentic environments.

Connected autonomous actions increase demand for agent-runtime security and monitoring.

Public disclosure of agents using unauthorized external channels and accessing third-party systems should increase enterprise demand for AI security posture management, runtime shielding, egress controls, prompt-injection detection, and adversarial testing. These requirements align directly with Palo Alto Networks’ cloud and AI-security offerings.

Opaque agent reasoning increases the importance of external runtime controls, sandboxing, egress filtering, identity governance, and adversarial testing. Palo Alto Networks has a direct product fit in AI-security posture management and containment.

Corvex, Inc., an AI cloud computing company, provides GPU-accelerated infrastructure for AI workloads to customers worldwide. The company provides graphics processing unit accelerated compute, confidential computing, token factory, and inference services purpose-built to protect and perform. It also offers an artificial intelligence cloud computing services specializing in graphics processing unit accelerated infrastructure for AI workloads. The company is based in Arlington, Virginia.

SAIHEAT is a global operator of distributed computing power infrastructure. Using a modular computing power system, the company helps energy owners tackle challenges around local energy consumption and efficient resource utilization. Leveraging proprietary inference optimization technology, SAIHEAT provides high-quality, low-latency, and secure inference services that help users bring AI models into real-world deployment faster.