OpenAI announced on Friday that it and SoftBank had invested $1 billion in SB Energy to construct and operate a data center of 1.2 gigawatts in Milam County in Texas.
It looks like another massive deal in the rapidly growing AI infrastructure arms races.
It’s a sign of something even more serious: the biggest bottleneck in AI production is electricity.
Capital, chips or code are all worthless if you cannot solve the power issue.
Economical arguments are compelling. One gigawatt continuous power can supply roughly 750,000 American households.
Data centres now concentrate these needs in concentrated geographical zones, straining the grids designed decades ago for predictable, steady industrial loads.
Data centres’ electricity consumption is expected to double between 2017 and 2023. This demand will be driven by AI-accelerated servers.
Lawrence Berkeley National Laboratory (US Department of Energy) estimates data center consumption to reach 325-580 Terawatt Hours in 2028. This is up from the 176 TWh consumed in 2023.
The International Energy Agency estimates that by 2026, the United States will have a demand for electricity exceeding 250 TWh.
The hidden chokepoint of AI is power
The growth curve reveals a harsh truth: the majority of American electric grids are unable to handle this demand. In some areas, grid interconnection lines now extend seven years.
Utilities usually project demand over years and not months. AI data centres are announcing gigawatt scale builds in quarterly timeframes.
Gridlock is the result, not a shortage but a misalignment of infrastructure development cycles with AI deployment speeds.
OpenAI’s investment with SoftBank avoids the bottleneck of a lack of dedicated generation.
SB Energy, SoftBank’s subsidiary, will build “powered infrastructure” at the Milam County site. This means it will develop or secure a supply of power in advance.
It isn’t a new strategy. Major cloud operators have pursued on-site production and contracts dedicated to renewables for years. But the speed and scale are unprecedented.
This $1 billion figure reflects capital intensity. To provide reliable AI-grade electricity, upfront investments are required in transmission interconnects and battery storage, which utility capex can’t keep up with.
The implications of the agreement for policy and markets
The partnership has three key advantages from a tactical perspective: stable and long-term electricity pricing, independent of the volatile wholesale market; quicker site commissioning through pre-securing access to grid; and reduced risk due to private coordination, rather than utility coordination.
SB Energy is now both a developer and an infrastructure provider. This reduces the time required for construction and permitting by several months.
Market-shaping is the broader impact. Hyperscalers signal that grid constraints will dictate AI infrastructure deployment, and not scarcity of capital.
The investment rationale for renewable energy, storage and transmission is reshaped. Developers of wind and solar near clusters of data centres gain immediate demand.
Data centre interconnections are a priority for regional transmission operators over industrial and residential projects.
The local regulators are already overloaded by the volume of proposals, and now face demands for concentrated power from tech companies with White House support.
The post OpenAI’s $1B bet with SoftBank reveals the AI essentials may change as new information is released.
This site is for entertainment only. Click here to read more
