Why SBX GRID
AI needs more compute than the old model can build.
Demand is compounding, power is scarce and ownership is concentrated. A community-funded, energy-first network is a different way to close the gap.
81 GW
Global data centre capacity in 2024
222 GW
Expected by 2030, roughly 18% growth a year
$3 to 4T
AI infrastructure spend projected by 2030 by NVIDIA's CEO
Industry projections cited in the SBX GRID whitepaper.
Three pressures, arriving at once.
Compute demand has exploded
Frontier AI models need tens of thousands of GPUs. In 2024, startups reported months-long waits for cloud GPUs unless they paid steep premiums.
Power is the new bottleneck
Data centres already use 1 to 2% of global electricity. Grid upgrades take years, and some hubs have paused new builds entirely.
Control sits with a few companies
A handful of hyperscalers own most capacity. Even many Web3 apps run on their servers, exposed to outages, lock-in and deplatforming.
Landscape
Where SBX GRID fits.
Traditional clouds have trust but not openness. Web3 networks have openness but struggle with trust. SBX GRID is designed to offer both.
Hyperscale cloudsAWS, Microsoft Azure, Google Cloud, Oracle
Strength
Unmatched scale, mature services and trusted SLAs.
The gap SBX GRID targets
Centralised, premium GPU pricing, egress fees and lock-in. Users never share in the value they create.
AI mega-projectsMulti-gigawatt private builds and national AI supercomputers
Strength
Power-first thinking at the scale of power plants.
The gap SBX GRID targets
Closed to outside investors, concentrated in a few locations and built for one company's needs.
Web3 compute marketsAkash, Render, Flux, Golem, Aethir and others
Strength
Permissionless participation and access to idle hardware.
The gap SBX GRID targets
Inconsistent reliability, few owned facilities, limited service stacks and often no legal wrapper for enterprises.
Side by side.
Who owns the infrastructure
- Hyperscale cloudsCorporate shareholders
- Web3 compute marketsToken holders, few physical assets
- SBX GRIDCommunity, backed by real sites
How decisions are made
- Hyperscale cloudsClosed boards
- Web3 compute marketsDAO, often no legal wrapper
- SBX GRIDBicameral DAO with a legal foundation
Reliability
- Hyperscale cloudsEnterprise SLAs
- Web3 compute marketsVaries by volunteer node
- SBX GRIDOperated hubs plus distributed edge
Large AI training runs
- Hyperscale cloudsIn-region GPU clusters
- Web3 compute marketsLimited by node latency
- SBX GRIDHigh-speed interconnect inside hubs
Energy
- Hyperscale cloudsMixed grids, growing PPAs
- Web3 compute marketsNot addressed
- SBX GRIDRenewable-first, on-site power
Users share in revenue
- Hyperscale cloudsNo
- Web3 compute marketsToken emissions
- SBX GRIDRevenue-linked distributions
| Criteria | Hyperscale clouds | Web3 compute markets | SBX GRID |
|---|---|---|---|
| Who owns the infrastructure | Corporate shareholders | Token holders, few physical assets | Community, backed by real sites |
| How decisions are made | Closed boards | DAO, often no legal wrapper | Bicameral DAO with a legal foundation |
| Reliability | Enterprise SLAs | Varies by volunteer node | Operated hubs plus distributed edge |
| Large AI training runs | In-region GPU clusters | Limited by node latency | High-speed interconnect inside hubs |
| Energy | Mixed grids, growing PPAs | Not addressed | Renewable-first, on-site power |
| Users share in revenue | No | Token emissions | Revenue-linked distributions |