Founders Fund is a technology investor with AI holdings across foundation models, software agents, data and compute infrastructure. Its disclosed portfolio includes OpenAI, Cognition, Scale AI and Crusoe, connecting companies that develop AI systems with businesses supplying their data and computing capacity.
A coding agent needs more than a capable model: it also depends on evaluation, software tooling and reliable compute. The selected portfolio makes those layers visible without treating them as a single market.
Compute infrastructure faces capacity, power and utilization questions. An agent application faces workflow, quality and distribution questions. Shared exposure to AI demand does not make their business models comparable on a single metric.
The firm’s portfolio extends beyond AI. These selected relationships are useful for analyzing one part of its technology exposure, but should not be used to infer that every investment follows an identical AI thesis.
Selected disclosed relationships, reviewed November 4, 2026. Each source supports the investment relationship; this is not a complete portfolio or a statement of current holdings.
Listed in Founders Fund’s official portfolio.
Source
Cognition appears in the firm’s disclosed company portfolio.
Source
The official portfolio lists Scale; this record references the company identity.
Source
Crusoe is included in the official portfolio.
SourceCognition concerns agents doing software-engineering work, while Scale represents a data-focused part of the AI stack. Data preparation and evaluation can support automation without being the same product as an agent. The comparison helps distinguish enabling work from the application interface where a user delegates a task.
Cognition represents AI applied to software engineering, while Crusoe represents computing infrastructure. The connection is a dependency across the stack, not identical business economics. Comparing both helps distinguish demand for AI capabilities from the physical resources and operating requirements involved in delivering those capabilities at scale.
A coding assistant may suggest text or explain code while a person controls the workflow. An agent may execute several development steps and use tools toward a goal. Labels alone are insufficient: the meaningful questions are which actions it can take, how results are checked and when human approval is required.
No. An investment relationship does not determine data ownership or access rights. Those questions depend on customer agreements, applicable rules and the services involved. A portfolio entry should not be used to infer that an investor can access training data, annotations or confidential material belonging to the company’s customers.
Infrastructure economics depend on whether capacity is available when customers need it and used productively over time. Hardware, power and service commitments create costs that differ from those of a software application. Demand for AI does not remove the need to understand workload scheduling, operating costs and the terms of customer contracts.
A coding agent can assist with software tasks, but that does not automatically provide suitable data or sound evaluation criteria. Those inputs still need to match the problem and its risks. The portfolio connection is useful for examining how development tools depend on evidence about performance, rather than assuming automation eliminates that work.