Investors & Strategic Partners
Sky Fusion is building distributed residential AI inference infrastructure for AI labs, hyperscalers, enterprise compute buyers, project-finance partners, and strategic investors.
We combine residential solar, battery storage, secure liquid-cooled GPU nodes, heat recovery, and SkyFusion OS orchestration into a behind-the-meter inference network. Our goal is to create a faster, more resilient, and more community-aligned layer of AI infrastructure that complements hyperscale data centers.
Structural Constraints
- Multi-year permitting timelines and complex approvals.
- Grid interconnection queues and transmission bottlenecks.
- Increasing land-use friction and local regulatory hurdles.
The AI infrastructure bottleneck
AI inference demand is compounding faster than traditional infrastructure can respond. Centralized data centers remain essential, but they face long permitting timelines, grid interconnection delays, transmission constraints, land-use friction, local opposition, and rising public skepticism.
The Sky Fusion Model
Sky Fusion offers a complementary model that deploys more quickly by using existing residential rooftops, batteries, broadband, electrical infrastructure, and thermal loads.
VANGUARD ALPHA
Vanguard Alpha and preliminary unit economics
Vanguard Alpha is Sky Fusion’s initial proposed pilot: 1,000 homes and approximately 2,200 GPUs, or roughly 2.2 GPUs per home on average. The pilot will be deployed in gated tranches, beginning with a smaller beta deployment before expanding. Each tranche must validate uptime, telemetry, security, homeowner acquisition, installation performance, and SLA metrics before additional capital and hardware are deployed.
Sky Fusion expects a tiered deployment model rather than a one-size-fits-all design. Smaller urban homes may support 1 GPU, core suburban homes may support around 2 GPUs, and larger estate homes may support 3–4 GPUs. The current blended target is approximately 2.2 GPUs per participating home and roughly $120,000 of installed capex per home, including solar, battery storage, compute hardware, liquid cooling, heat-recovery integration, secure networking, monitoring, controls, and installation. This implies approximately $55,000–$65,000 of installed capex per deployed GPU.
Expected wholesale CPPA pricing is $2.25 –$2.75 per H100 GPU-hour, depending on contract duration, utilization guarantees, take-or-pay provisions, workload profile, security requirements, and project-finance terms. At 2.2 GPUs per home and 70% utilization, this implies roughly $30,000–$37,000 of gross compute revenue per home per year before debt service, homeowner benefits, operations, maintenance, insurance, downtime, and GPU refresh reserves.
Business model
Execution platform
Long-term vision
Sky Fusion is designed to be a software, orchestration, and operating platform, not a balance-sheet-heavy owner of all physical infrastructure.
The target model is for Sky Fusion to earn a platform and operating fee equal to approximately 20% of contracted CPPA revenue. This fee compensates Sky Fusion for SkyFusion OS, secure workload orchestration, homeowner acquisition, installer coordination, fleet monitoring, billing, settlement, maintenance coordination, customer reporting, and SPV operations support. Physical assets can be owned and financed through dedicated infrastructure SPVs backed by contracted compute revenue.
Sky Fusion is being built with operational support from the team and assets of SBBnet, Inc., a 27-year operating business with more than $200 million in cumulative revenue.
SBBnet brings a 20-person team across software development, IT, sales, business development, marketing analytics, and accounting. Sky Fusion also intends to leverage CleanEnergyAuthority.com, an SBBnet asset that has helped solar installation companies acquire residential solar customers for 18 years. This operating base is directly relevant to homeowner acquisition, solar installer relationships, residential energy-market experience, software development, analytics, customer operations, and financial discipline.
Vanguard Alpha is intentionally modest relative to the scale of the opportunity. If the model works, Sky Fusion could scale into a national behind-the-meter inference layer. At 100,000 participating homes, using the same average of 2.2 GPUs per residence, Sky Fusion would represent roughly 220,000 distributed GPUs — comparable in GPU count to some of the largest centralized AI supercomputing deployments. This is designed to complement, not replace, hyperscale data centers.
Partnership opportunities
Sky Fusion is seeking AI infrastructure partners, venture and strategic investors, project-finance partners, and solar and installer partners who see value in a distributed, community-aligned inference layer.
We are focused on structuring long-term, contracted wholesale capacity that can integrate cleanly into existing API, enterprise, and batch channels while aligning incentives with participating households and local communities.