Buy versus Build an LLM: A Decision Framework for Governments

TL;DR

The paper addresses how governments should decide between buying, building, or hybridizing LLM capabilities, balancing sovereignty, safety, cost, and sustainability. It offers a neutral, multi-dimensional decision framework, including acquisition pathways taxonomy, pre-decision factors, lifecycle planning, and evolving landscape considerations, enriched by practice insights from SEA-LION and Apertus. The contributions include a detailed taxonomy of pathways, lifecycle and cost analyses, and governance-focused guidance to support policy-makers in diverse national contexts. The work highlights that LLM deployment in public sectors should be treated as strategic infrastructure, with decisions tailored to national priorities, capabilities, and trust considerations, to maximize resilience and public value.

Abstract

Large Language Models (LLMs) represent a new frontier of digital infrastructure that can support a wide range of public-sector applications, from general purpose citizen services to specialized and sensitive state functions. When expanding AI access, governments face a set of strategic choices over whether to buy existing services, build domestic capabilities, or adopt hybrid approaches across different domains and use cases. These are critical decisions especially when leading model providers are often foreign corporations, and LLM outputs are increasingly treated as trusted inputs to public decision-making and public discourse. In practice, these decisions are not intended to mandate a single approach across all domains; instead, national AI strategies are typically pluralistic, with sovereign, commercial and open-source models coexisting to serve different purposes. Governments may rely on commercial models for non-sensitive or commodity tasks, while pursuing greater control for critical, high-risk or strategically important applications. This paper provides a strategic framework for making this decision by evaluating these options across dimensions including sovereignty, safety, cost, resource capability, cultural fit, and sustainability. Importantly, "building" does not imply that governments must act alone: domestic capabilities may be developed through public research institutions, universities, state-owned enterprises, joint ventures, or broader national ecosystems. By detailing the technical requirements and practical challenges of each pathway, this work aims to serve as a reference for policy-makers to determine whether a buy or build approach best aligns with their specific national needs and societal goals.

Figures (4)

• Figure 1: Acquisition pathways for language models, spanning buy, hybrid and build approaches, with progressively higher cost and capability requirements but increasing levels of control and strategic autonomy.
• Figure 2: Critical decision points across the sovereign LLM building lifecycle. This figure illustrates key strategic, technical and operational decision calls that governments face when building a sovereign LLM.
• Figure 3: Market concentration across application categories measured by the Herfindahl–Hirschman Index (HHI). HHI is a standard antitrust metric, ranging from 10,000 for a monopoly to 2,500 for a market with four equally sized firms. Categories on the y-axis represent use-case categories reported by OpenRouter openrouter, a marketplace and API gateway that routes developer traffic to selected LLM providers. Market shares are computed from total token usage on OpenRouter from January to December 2025. The figure reports concentration at the creator level (blue; the model developer/company, e.g., OpenAI, Anthropic) and the model level (orange; specific model variants). Adapted from Figure 19 of demirer2025emerging.
• Figure 4: Strategic evaluation framework for government LLM build-vs-buy decisions.