Cohere has established a Defence AI Accelerator in Singapore to help defence organizations and other sovereign customers move artificial intelligence projects from experimentation into secure operational deployments. The initiative will provide engineering support, technical advisory, secure development environments, and access to Cohere’s enterprise AI platform, North. Designed for organizations with strict requirements around data control, privacy, and governance, the accelerator will support on-premises, hybrid, and air-gapped deployments. The launch also advances Cohere’s broader Asia-Pacific expansion strategy, which includes plans to quadruple its regional presence by the end of 2027.
Cohere Targets Secure Defence AI Deployment in Asia-Pacific
Cohere’s new Defence AI Accelerator in Singapore is designed to address a growing challenge for governments and defence organizations: moving AI systems from controlled demonstrations into operational environments without compromising data sovereignty, security, or governance.
The accelerator will provide participating organizations with access to Cohere engineers, technical advisers, enterprise AI models, and its North agentic AI platform. Rather than operating primarily as a demonstration centre, the initiative is intended to support the complete deployment process, from proof-of-concept development and testing to integration with existing systems and production infrastructure.
This distinction is important in defence and other sensitive public-sector environments. An AI model that performs well in a demonstration may still be unsuitable for operational use if it cannot run within an organization’s security boundaries, connect to established workflows, or meet requirements for auditability and administrative control.
Cohere is positioning the accelerator as an engineering and deployment environment where sovereign customers can test applications under practical conditions before expanding them into operational systems.
On-Premises and Air-Gapped AI Address Sovereignty Requirements
A central part of the initiative is Cohere’s support for different deployment models, including on-premises, hybrid, and air-gapped infrastructure.
On-premises deployments allow organizations to run AI systems within infrastructure they control. Hybrid architectures can combine private environments with selected connected services, while air-gapped deployments isolate systems from external networks. These options are relevant to defence organizations and government agencies handling classified, sensitive, or mission-critical information.
The approach also addresses concerns about where data is processed, who can access models and infrastructure, and how organizations retain control over their AI systems. For sovereign customers, these questions extend beyond data residency to include operational control, infrastructure dependencies, and the ability to maintain services within national or organizational boundaries.
Cohere’s North platform is central to the accelerator. It is the company’s enterprise agentic AI platform, designed to connect AI capabilities with organizational information and business workflows. The accelerator will give participants access to North alongside Cohere’s models, with support for integration into existing operational systems.
Cohere says North has already been deployed with several regulated organizations in Singapore. It has not identified those customers or provided detailed deployment outcomes in the announcement, leaving the scale and nature of those implementations unclear.
From AI Pilots to Governed Operational Systems
The accelerator’s programme covers several stages of enterprise deployment. Participants will work with Cohere engineers to develop and validate applications, test them in secure environments, and connect them to existing workflows and infrastructure.
The company also plans to provide governance frameworks and administrative controls for managing AI systems. These capabilities are particularly important when organizations begin introducing agents that can retrieve information, use tools, and perform tasks rather than simply generate text.
In defence settings, the consequences of incorrect outputs or unauthorized actions can be significant. AI systems therefore need clear access permissions, human oversight, audit trails, and boundaries around the actions agents can perform. The announcement does not specify the defence use cases or operational permissions that the accelerator will support, so it remains unclear how these controls will be configured for different missions.
This deployment-first model reflects a broader enterprise AI shift. Organizations are increasingly evaluating AI providers on their ability to deliver secure infrastructure, integration expertise, and governance alongside model performance.
Sovereign AI Investment Shapes Cohere’s Regional Strategy
Cohere’s Singapore initiative also forms part of a wider Asia-Pacific expansion plan. The company says it intends to quadruple its regional presence by the end of 2027 through additional offices, local partnerships, and new investments. The accelerator is intended to strengthen its relationships with public-sector institutions and support customers seeking locally controlled AI infrastructure.
The timing reflects growing regional interest in sovereign AI. IDC’s January 2026 analysis predicts that by 2028, 80% of the region’s A2000 enterprises will prioritize AI sovereignty for mission-critical workloads through non-public hosting, open technologies, and regional partners. This forecast is specifically about A2000 enterprises across Asia/Pacific excluding Japan, rather than all APAC organizations.
The distinction matters because sovereignty requirements differ across markets and industries. Some organizations prioritize local data hosting, while others require control over model deployment, infrastructure, security architecture, and access to sensitive information. Defence organizations may have stricter requirements than commercial enterprises, particularly where systems must operate without external network connectivity.
Cohere’s strategy places it in a competitive market that includes cloud providers, infrastructure vendors, specialist AI companies, and national technology initiatives. Microsoft, Google, Amazon, and NVIDIA all offer technologies relevant to private or sovereign AI deployments, while regional providers are developing infrastructure and services tailored to local regulatory and operational requirements.
Cohere’s differentiation is its focus on enterprise models and agentic AI that can operate across customer-controlled environments. The Singapore accelerator adds hands-on engineering support to that proposition, potentially helping customers navigate the gap between acquiring AI technology and deploying it within complex institutions.
Deployment Evidence Will Determine the Accelerator’s Impact
The initiative is a strategic expansion, but its value will ultimately depend on measurable customer outcomes. Cohere has not disclosed participating defence organizations, named accelerator customers, specific mission applications, or a timetable for the first production deployments.
Those details will be important in assessing whether the accelerator can move beyond technical experimentation. Relevant measures could include the number of projects reaching production, integration time, model performance in isolated environments, governance effectiveness, and the ability to maintain systems without external connectivity.
For defence organizations, success will also depend on whether AI agents can be constrained to approved tasks and reliably supervised by personnel. Secure hosting is only one part of operational readiness; data quality, model evaluation, cybersecurity, human accountability, and resilience remain essential.
Cohere is betting that a dedicated engineering environment, combined with flexible deployment and governance capabilities, can help public-sector customers move more quickly while retaining control over sensitive systems. As governments across Asia-Pacific increase their focus on sovereign AI, the accelerator gives the company a more direct route into defence and other high-security markets. Its next challenge will be demonstrating that those deployments deliver dependable operational value.
Market Landscape
The sovereign AI market is shifting from infrastructure procurement toward complete, governable AI deployment environments. Defence organizations and government agencies increasingly need systems that can operate within strict security boundaries while integrating with existing institutional workflows.
Five trends are shaping the market:
- AI sovereignty: Governments want greater control over models, data, infrastructure, and deployment locations.
- Air-gapped AI: Isolated deployments are becoming important for sensitive environments that cannot depend on external cloud connectivity.
- Agentic AI governance: Organizations need permissions, auditability, human oversight, and administrative controls for agents that can execute tasks.
- AI industrialization: Enterprises are moving beyond pilots toward repeatable deployment processes and production-ready infrastructure.
- Regional AI ecosystems: Local partnerships, technical expertise, and infrastructure are becoming important to sovereign AI procurement.
IDC forecasts that by 2028, 80% of A2000 enterprises across Asia/Pacific excluding Japan will prioritize AI sovereignty for mission-critical workloads. This is a forecast for a defined enterprise group, not a measurement of all APAC organizations.
Cohere’s accelerator is positioned to address this demand by combining model access, deployment flexibility, engineering support, and governance capabilities in one programme.
Top Insights
- Cohere has established a Defence AI Accelerator in Singapore to help sovereign customers move AI projects from proof-of-concept to operational deployment.
- The accelerator supports on-premises, hybrid, and air-gapped infrastructure for organizations with strict data control and security requirements.
- Cohere’s North agentic AI platform will provide a foundation for developing and integrating AI applications into existing organizational workflows.
- IDC forecasts that 80% of A2000 enterprises in Asia/Pacific excluding Japan will prioritize AI sovereignty for mission-critical workloads by 2028.
- Cohere plans to quadruple its regional presence by the end of 2027 through investments, partnerships, and additional APAC offices.
Power Tomorrow’s Intelligence — Build It with TechEdgeAI
