KIDZ AI Targets $44.6M Canopy Wave Deal with Dallas Data Center MOU

For SaaS companies increasingly embedding AI features into their platforms, the availability of cost-efficient, dedicated inference compute is becoming a make-or-break infrastructure decision. KIDZ AI’s proposed deployment of NVIDIA Blackwell B300 GPUs at a Dallas data center, tied to a five-year, $44.6 million enterprise contract, signals a new model for AI compute provisioning — one that could offer SaaS developers predictable pricing and performance for open-weight model inference without the markup of hyperscale cloud services.

KIDZ AI, the EdTech company pivoting aggressively into GPU compute infrastructure, has taken its first concrete step toward physical deployment of a dedicated AI inference cluster. On August 3, 2026, the company announced the signing of a non-binding Memorandum of Understanding (MOU) with Texas-based Limestone Networks to host its proposed NVIDIA Blackwell B300 GPU infrastructure at the DFW3 data center in Dallas. The move directly advances KIDZ AI’s previously announced 60-month, $44.6 million definitive GPU compute services agreement with Canopy Wave, Inc., converting a contractual demand pipeline into tangible infrastructure planning. The deployment is designed to power high-throughput, cost-efficient inference workloads for open-weight AI models — a strategic niche that distinguishes KIDZ from general-purpose cloud GPU providers.

“

The move directly advances KIDZ AI’s previously announced 60-month, $44.6 million definitive GPU compute services agreement with Canopy Wave, Inc., converting a contractual demand pipeline into tangible infrastructure planning.

The planned infrastructure would be ordered and managed by Catalyst Compute LLC, KIDZ AI’s wholly owned subsidiary formed to execute this neocloud strategy. Under the MOU, Limestone and KIDZ are evaluating two physical configurations: a dedicated 1.125 MW data hall (Data Hall 150) or a dual-hall setup, pending technical confirmation that the distance and connectivity between the halls meet latency and bandwidth requirements. Initial deployment would begin at approximately 0.5 to 0.6 MW of power capacity and scale toward the full 1.125 MW as GPU servers are delivered, installed, tested, and commissioned. This staged approach mirrors typical hyperscale rollouts and reflects the uncertain but rapidly evolving supply environment for NVIDIA’s next-generation Blackwell B300 accelerators.

The announcement adds a physical dimension to what has been a largely financial story. KIDZ AI, originally a children’s educational technology company, diversified into AI infrastructure in 2025–2026, first by negotiating the $44.6 million Canopy Wave contract and now by securing a potential data center home. The $44.6 million figure represents committed revenue over five years, though actual deployments and payments are contingent on infrastructure readiness. The MOU with Limestone — a regional cloud and colocation provider with a strong Dallas presence — signals that KIDZ is moving beyond paper agreements to operationalizing GPU capacity, a transition that often makes or breaks neocloud startups. The choice of Blackwell B300 GPUs, NVIDIA’s latest inference-optimized architecture, underscores a focus on serving the expanding market for open-weight large language models (LLMs) such as Llama 3, Mistral, and their enterprise-tuned variants, which can be hosted cost-effectively on dedicated clusters rather than rented from hyperscale clouds at premium rates.

From a market perspective, the deal highlights the intensifying competition for specialized AI inference capacity. Inference — the process of generating predictions from trained models — is rapidly becoming the dominant AI workload as enterprises move from training foundational models to deploying them at scale. KIDZ’s inference-only, open-weight model focus positions it to capture demand from enterprises that require low-latency, high-throughput AI services without vendor lock-in. However, the non-binding nature of the MOU means that final terms, including power pricing, connectivity costs, and lease duration, are still to be negotiated. The company’s ability to secure NVIDIA Blackwell B300 units in a supply-constrained market also remains a key risk. With major cloud providers and on-premise deployments vying for the same silicon, KIDZ’s relatively small scale could pose challenges in procurement.