Pike 园区电力容量，OpenAI 将入驻
Core Highlights
NVIDIA laid out the logic of this partnership in plain terms: lock down the electricity first, then talk about compute. Together with SB Energy, it has secured LPS power capacity at the PORTS-Pike technology park in Ohio, and the usage of that capacity is restricted to the exclusive deployment of NVIDIA compute capability, with OpenAI signing on as the tenant.
The decision is notable because it treats electricity not as a background utility but as the central asset being bought and sold. In effect, NVIDIA is purchasing optionality on the grid itself, converting a shared public resource into a private competitive advantage before rivals even start negotiating, and signaling that the company now thinks like a utility as much as a chipmaker. It is a reminder that in the age of AI, the most valuable real estate is increasingly measured in megawatts rather than square feet, a reversal few in the industry anticipated even two years ago.
What It Does or What Happened
The significance of this step is that scarce power resources become NVIDIA's moat. Once the park's power capacity is locked in, it is reserved exclusively for NVIDIA compute, and outside competitors cannot claim that portion of the load. OpenAI, as the tenant, effectively pre-books stable power supply and the accompanying compute for the coming years.
This reverses the usual order of operations. Traditionally a customer finds land and power, then orders chips. Here the power is secured first, and the chips are assumed to follow, bundled into the same delivery commitment so the tenant never has to haggle with utilities over usage quotas item by item, a process that routinely stalls competing builds for months and injects uncertainty into every revenue forecast. For OpenAI, the arrangement removes a major source of schedule risk, since power is the input least likely to be forgiving when demand spikes and the grid is already constrained.
Technical Details
LPS, or large power delivery capacity, refers to the total power ceiling a park can draw from the grid. "Locking" it means reserving long-term capacity on the grid side, avoiding later expansion bottlenecks caused by transmission constraints. For an AI factory, supply stability directly determines whether the cluster can run at full load.
The reservation also de-risks the construction timeline. Knowing the wattage is guaranteed lets engineers size substations, cooling, and cabling without waiting on uncertain utility approvals that often stretch for years. Any curtailment or permitting delay would otherwise push back the entire go-live schedule, so the lock-in is as much about certainty as about raw megawatts, and certainty is what tenants will pay a premium to obtain. The guarantee also lets financing close faster, because lenders can underwrite a known quantity of energy rather than a hope that the grid will cooperate on an unknown timeline.
Versus Competitors
Most cloud providers are still negotiating usage quotas for individual data centers with utilities case by case, a slow and contested process. NVIDIA instead chose to lock the power for an entire park at the source, then lease it to large customers as a finished delivery, moving noticeably faster than peers who must assemble the same pieces reactively.
The contrast is strategic as much as operational. Competitors compete for power reactively, after demand appears; NVIDIA now competes for it proactively, treating grid capacity as a product feature it can promise to tenants and thereby converting power scarcity into its own bargaining chip against both utilities and rival clouds that may have the chips but not the watts. Rivals that treat power as someone else's problem will find the cheapest watts already spoken for by the time they try to scale their own clusters in earnest.
Industry Impact or Use Cases
Put simply, the bottleneck for AI infrastructure is shifting from chips to the grid. Whoever secures long-term, stable power capacity first gains control of the next generation of compute centers, and PORTS-Pike is merely the first piece in this scramble for electricity, with more park-scale lock-ins likely to follow as the race for watts intensifies across the industry and energy contracts themselves become strategic assets on corporate balance sheets. The deal may also redraw how those balance sheets are read, since power purchase agreements start to look as strategically important as chip inventories to investors evaluating who can actually deliver intelligence at scale.