Forum Markets is making its first major AI infrastructure investment through a joint venture with Edge Node AI to deploy inference computing across existing data centers and mobile communications towers in the United States. The venture starts with a 2 MW Dallas deployment, a 9 MW powered campus in High Point, North Carolina, and an initial Texas tower site, targeting low-latency AI workloads closer to end users.
AI infrastructure is entering a phase where simply adding more GPU capacity is no longer the only priority. As generative AI moves into production, inference workloads increasingly need to operate close to applications, users and data while maintaining predictable access to high-performance compute.
Forum Markets is positioning its latest investment around that shift. The company has formed Forum Edge AI LLC, a joint venture with AI infrastructure provider Edge Node AI, to own and operate distributed inference compute capacity at data centers and mobile communications tower sites where power and permits are already available.
Forum owns 51% of the joint venture and will manage and consolidate its financial results, while Edge Node AI owns the remaining 49% and will provide technical and operating capabilities. The venture represents Forum’s first major investment in AI infrastructure.
The initial portfolio spans three deployment types. In Dallas, the joint venture will own computing equipment and transformer capacity at an existing Edge Node AI facility, with approximately 2 MW already operating and capacity expected to increase to about 4 MW by early 2027. Forum says contracted revenue from the initial Dallas phase is expected to begin in late 2026 or early 2027.
The second major deployment is a six-building, five-acre campus in High Point, North Carolina. The site has 9 MW of live utility power available today, with an initial installation of approximately 6 MW of GPUs expected to become operational in the first quarter of 2027. Further capacity is planned as cooling upgrades are completed.
The third component is an initial mobile tower deployment in Texas near the Dallas facility. Forum says agreements with SBA Communications and Tillman Infrastructure provide a framework for additional tower deployments across multiple metropolitan areas, although future installations will depend on the performance of earlier sites.
The technical rationale is centered on inference rather than AI model training. Training large models generally requires concentrated clusters of accelerators and substantial infrastructure, while inference occurs repeatedly after a model has been trained. Applications such as real-time computer vision, interactive AI, autonomous systems and other latency-sensitive workloads can benefit from compute being located closer to where requests originate.
Edge Node AI’s distributed architecture is designed around that requirement. Instead of relying exclusively on large, remote hyperscale data-center campuses, the venture intends to place inference capacity inside metropolitan areas and existing edge locations.
That strategy also attempts to address one of the major constraints facing new AI infrastructure: the time and capital required to bring power, land, permits and connectivity online. Forum says its initial sites already have power and the necessary permits or operating structures, allowing the joint venture to concentrate on installing and financing compute equipment rather than developing entirely new greenfield facilities.
The model resembles a hub-and-spoke architecture. Owned data centers serve as hubs, while tower deployments operate as distributed edge nodes. Forum Edge AI personnel can service tower equipment from the larger facilities, according to the company.
The project structure is deliberately modular. Each site will be held in a separate special purpose vehicle, with projects financed and closed independently and without cross-collateralization. Where real estate is involved, Forum says land and compute equipment will also be held separately so equipment financing does not encumber the underlying property and vice versa.
That structure is primarily a financial and risk-management decision, but it also has implications for AI infrastructure expansion. A modular deployment model can allow additional compute sites to be added incrementally rather than requiring an entire network to be funded and built simultaneously.
Forum and Edge Node AI are also aligning their ownership interests. Subject to specified power and financing milestones, the companies will exchange 9.95% equity stakes in one another. Forum says Edge Node AI would receive approximately 1.46 million Forum shares based on 13.2 million shares outstanding as of October 2, 2026.
The commercial model is built around contracted compute capacity. Forum says the joint venture is allocating offtake to large counterparties under multi-year agreements covering the contemplated initial power capacity, with equipment financing matched to contract terms.
For Forum, the infrastructure strategy represents a move into a market increasingly dominated by demand for AI accelerators, power and data-center capacity. NVIDIA remains the central supplier in the GPU compute ecosystem, while hyperscalers such as Microsoft, Google and Amazon continue to build massive centralized AI infrastructure. The distributed model pursued by Forum and Edge Node AI targets a different part of the market: localized inference.
The distinction could become increasingly important as AI applications require lower latency and more continuous inference. Edge computing can reduce the physical distance between compute and users, although the economics depend heavily on utilization, power availability, network connectivity, cooling and the workloads ultimately deployed at each location.
Forum expects the initial portfolio to contribute between $125 million and $175 million in consolidated revenue during 2027, with the company saying that guidance reflects 100% of joint-venture revenue. It also expects to become cash-flow positive in early 2027. These are company projections rather than independently verified forecasts, and the timing depends on site construction, equipment deployment, financing and customer contracts.
The broader development is another sign that AI infrastructure is diversifying beyond hyperscale campuses. As inference becomes a larger component of enterprise and real-time AI workloads, distributed compute can complement centralized infrastructure by placing accelerators where applications actually need them.
Forum’s strategy therefore combines an infrastructure thesis with a deployment thesis: use existing powered locations, install financed GPU capacity, secure contracted workloads and expand site by site. Whether that model can deliver attractive economics at scale will depend on utilization and execution, but the initial portfolio provides a concrete example of how the next generation of AI compute could extend into existing data-center and telecommunications infrastructure.
Market Landscape
AI infrastructure is increasingly splitting into two complementary architectures: large centralized facilities optimized for model training and high-density workloads, and distributed infrastructure designed for inference closer to users and applications.
Hyperscalers and specialized data-center operators continue to build massive AI campuses, while edge infrastructure companies are targeting latency-sensitive inference. Forum Markets and Edge Node AI are entering the latter segment by using existing powered data centers and telecom tower locations.
The approach could reduce some development hurdles associated with greenfield AI facilities, particularly where power and permits are already available. However, distributed infrastructure introduces its own challenges around GPU utilization, networking, cooling, maintenance and customer demand.
Top Insights
- Forum Markets is investing in distributed AI inference through a 51%-owned joint venture with Edge Node AI.
- The initial portfolio includes 2 MW in Dallas, a 9 MW powered campus in North Carolina and a Texas tower deployment.
- The strategy uses existing power, permits and physical infrastructure rather than relying entirely on new greenfield data-center construction.
- Separate project-level SPVs are designed to isolate financing and operational risks between individual AI compute deployments.
- Forum projects $125 million to $175 million in consolidated 2027 revenue from the initial joint-venture portfolio.
Power Tomorrow’s Intelligence — Build It with TechEdgeAI
