Managed agents could become a demand amplifier for cloud infrastructure software
AI summary card
Managed agents could become a demand amplifier for cloud infrastructure software
Morgan Stanley believes that while managed agents take over some orchestration responsibilities, they still rely on external execution environments, data systems, network services, and observability, which could benefit infrastructure software companies such as Cloudflare, Akamai, Snowflake, MongoDB, Palantir, Datadog, and Dynatrace.
- The market has recently interpreted managed agents as AI-native vendors moving up into the cloud infrastructure software stack, causing several infrastructure software stocks to come under pressure within a week; for example, NET -21%, SNOW -20%, and DDOG -12%.
- The report's core view is that managed agents abstract the agent loop and orchestration complexity, but long-running tasks still require external tools, enterprise data, network services, execution environments, and monitoring systems.
- As agent deployment becomes easier, external tool calls, web search, data access, inference calls, and multi-location execution environments may increase significantly, lifting demand for CDN, edge computing, data platforms, and observability.
- The risk is that managed agents could replace some customer-built agentic applications, edge serverless platforms, or vendors' own agent products, but the report believes that latency, compliance, data governance, and enterprise context in different use cases will continue to support specialized infrastructure demand.
Report interpretation
Overview
This report discusses the impact of managed agent services on the North American cloud infrastructure software industry. Morgan Stanley believes the market is currently focused more on the potential disruption to infrastructure software companies from AI-native vendors launching managed agents, while overlooking the demand expansion that large-scale agent deployment could bring in external tool calls, data access, network traffic, edge inference, compute consumption, and end-to-end observability. The report covers CDN and edge computing, enterprise data platforms, and observability vendors.
Core views
The core view is that managed agents do not equal owning the full infrastructure stack. Their main value lies in taking over the orchestration layer, including calling models, deciding when to use tools, routing tool calls, and keeping the agent loop running; but execution environments, external data systems, MCP connection systems, network services, sandboxes, code execution, file editing, browsing, and external service calls still require separate infrastructure support. Therefore, if millions, billions, or even more AI agents emerge in the future, and each agent makes multiple tool calls, accesses governed data sources, and executes long-running tasks, this will expand demand for CDN, edge computing, data infrastructure, and observability.
Analysis framework
The report uses a value-chain decomposition method, splitting managed agents into an orchestration layer and an execution environment, and analyzing how each affects different infrastructure segments. For CDN and edge computing, it focuses on how long-running tasks, multi-step workflows, geography, latency, sovereignty, and policy control raise the value of edge nodes; for data platforms, it focuses on the compute consumption and data governance demands created when agents access enterprise data through MCP on platforms such as Snowflake, MongoDB, and Palantir; for observability, it focuses on tool failures, latency spikes, runaway token consumption, policy violations, model-quality drift, and cross-stack monitoring needs.
Methodology notes
Managed agents abstract the agent loop, but do not eliminate external execution, data, network, or monitoring needs.
The report separates the orchestration capability of managed agents from the actual execution environment, arguing that the former coordinates model and tool calls, while the latter still depends on CDN, edge computing, enterprise data platforms, MCP connection systems, and observability tools.
Asset mapping & comparison
Structured mapping from thesis to named assets (strengths, weaknesses, peers, risks).
- CLOUDFLARE INC / NET.USPotential beneficiary; a cloud infrastructure software company related to CDN, web security, and edge computing.
- Strengths
- Long-running agent tasks may increase request volume, internet and application traffic, API protection, low-latency delivery, caching, rate limiting, smart routing, and edge computing demand.
- Weaknesses
- Managed agent platforms may reduce some customers' need to build agentic applications on top of Cloudflare Workers and other serverless edge development platforms.
- Comparison
- Like Akamai, Cloudflare is positioned as a potential beneficiary in CDN and edge computing; compared with centralized managed agent platforms, edge platforms are more valuable in terms of latency, geography, sovereignty, and policy control.
- Risks
- If managed agent platforms absorb more application-building and execution capabilities, or if customers concentrate edge demand on centralized platforms, demand for some edge development platforms could be suppressed.
- Akamai TechnologiesA potential beneficiary in CDN and edge computing.
- Strengths
- Multi-step agent workflows increase the need for proximity, caching, request routing, and edge traffic control.
- Weaknesses
- Some customers may move to managed agent platforms instead of building their own edge agent applications.
- Comparison
- It belongs to the same CDN and edge computing beneficiary bucket as Cloudflare.
- Risks
- If managed agents primarily run in centralized data centers, customer usage of edge serverless platforms could change.
- Snowflake / MongoDB / PalantirPotential beneficiaries in the enterprise data platform segment.
- Strengths
- Agent access to enterprise data through MCP may drive more SQL queries, compute consumption, real-time analytics, and governed access demand.
- Weaknesses
- Snowflake's own agent products could be displaced by managed agents; Palantir could be weakened as an enterprise data storage layer.
- Comparison
- Snowflake and MongoDB mainly benefit from data querying and governed access; Palantir enables workflow execution and real-time data analysis through Ontology MCP.
- Risks
- Data may migrate from Snowflake to open-format data lakes such as Iceberg or Delta and be queried by other processing engines; enterprise customers could also use managed agents to build their own Ontology.
- Datadog / DynatracePotential beneficiaries in observability.
- Strengths
- Long-running tasks, autonomous agents, and multi-platform deployments expand the monitoring surface, including tool failures, latency spikes, token consumption, policy violations, model-quality drift, and real-time compliance controls.
- Weaknesses
- Customers may use agents and open-source technologies such as OpenTelemetry to build DIY observability solutions.
- Comparison
- Both are positioned as neutral observability beneficiaries across the stack, above the model layer.
- Risks
- Customers may build capabilities themselves that current observability vendors are monetizing or hope to monetize in the future, such as some cloud security or AI SRE features.
Key data
- Report date2026-04-13Morgan Stanley Research was published on April 13, 2026 at 08:19 GMT.
- Recent stock pressureNET -21%, SNOW -20%, AKAM -23%, DDOG -12%, DT -14%The report says the market has priced in the disruption risk from managed agents over the past week, putting pressure on several infrastructure software stocks.
- Latency exampleLondon-to-Virginia inference path about 28 milliseconds one-way latencyThe report cites an Akamai blog example to show that each additional model hop and tool call in a multi-step agent workflow magnifies the value of edge proximity, caching, routing, and traffic control.
- Rating horizon12-18 monthsThe disclosure notes that Morgan Stanley stock ratings and price targets typically use a 12- to 18-month horizon.
Impact & implications
For investment implications, the report tends to view managed agents as a potential catalyst for cloud infrastructure software demand rather than a pure substitution threat. CDN and edge computing vendors such as Cloudflare and Akamai may benefit from more requests, web traffic, application traffic, API protection, low-latency delivery, and distributed inference demand; data platforms such as Snowflake, MongoDB, and Palantir may benefit from the query compute, permission governance, and real-time analytics demand created when agents access enterprise data; observability vendors such as Datadog and Dynatrace may benefit from more complex agent architectures and cross-model, cross-tool, cross-platform monitoring needs.
Risks
- Managed agent platforms may replace some customer demand for building agentic applications on serverless edge platforms such as Cloudflare Workers and Akamai EdgeWorkers.
- Snowflake's own agent products, such as Snowflake Intelligence and Cortex, may face substitution risk from managed agents.
- Data may migrate from Snowflake to open-format data lakes such as Iceberg or Delta and be queried by non-Snowflake processing engines.
- Palantir may be downgraded to an enterprise data storage layer rather than an end-to-end AI or agentic platform.
- Customers may use managed agents and open-source technologies such as OpenTelemetry to build observability capabilities in-house, weakening the monetization potential of Datadog, Dynatrace, and similar vendors.
- Morgan Stanley discloses that it has, or seeks, investment banking and other service relationships with multiple covered companies, and investors should treat this research as only one factor in an investment decision.
What to watch
- Whether managed agent platforms materially increase external tool calls, web search, application traffic, and API request volume.
- Whether edge vendors such as Cloudflare and Akamai form partnerships with managed agent platforms to push compute and data processing closer to the data and end users.
- Enterprise adoption of connection methods such as Snowflake MCP, MongoDB MCP, and Palantir Ontology MCP, and the resulting compute consumption and data governance demand.
- Whether customers migrate data from proprietary data platforms to open-format data lakes such as Iceberg or Delta.
- Whether observability vendors such as Datadog and Dynatrace can use their data advantages to deliver AI operations and monitoring capabilities that are more effective and lower in total cost of ownership than general-purpose agents.