The Gulf of America can be an off-grid power source for US-based AI

To produce the off-grid power needed for AI, we can look to the resources and infrastructure already available in the Gulf of America.

For many years, the Gulf has played host to well-developed, well-regulated, and efficient oil exploration and extraction. The United States has vast technological, industrial, and human capital resources dedicated to safe, well-managed oil and gas operations.

In addition to these oil reserves, the Gulf also contains vast, undeveloped natural gas resources. These reservoirs have not been produced due to low natural gas pricing at the time of drilling, combined with a lack of pipeline and platform infrastructure. Essentially, the economics of these reservoirs were not advantageous at the time of initial exploration.

However, when this natural gas is developed for onsite use to power AI data centers, a very positive economic (not to mention environmental) outcome emerges. To that end, our focus is to develop “stranded” natural gas to power AI.

Gulf-based natural gas can power AI data centers for many years

GGAI will therefore tap into the immense reserves of natural gas in the shallow-water outer continental shelf of the Gulf of America. Here is how it works.

Electricity Generation

First, mobile jack-up drilling rigs, well-established and regulated industrial equipment used to drill for oil and gas in the Gulf, are repurposed as mobile production units (MOPUs) and used as mobile data centers. We will call these repurposed jack-up rigs GenPods. A GenPod is towed to an undeveloped natural gas reserve. The reserves are relatively shallow below the ocean floor and are identified by (a) wells that have been drilled but not produced, and (b) a specific seismic signature that indicates the presence of hydrocarbons (natural gas). Natural gas occurs in sandstone reservoirs with high permeability and porosity, which, in geologic terms, means it is easy and safe to extract the proven natural gas reserves from these locations.

Next, in place of all of the equipment required for oil and gas drilling, the GenPod will carry a series of conexes, military-grade weatherized containers that contain AI servers and all of the requisite tech to facilitate AI operations.

The GenPod will be located adjacent to an already-drilled, completed well that will produce the natural gas reserve. The gas supply well will be tied back to the GenPod by a riser or flowline. The hydrocarbons will flow from the well, through the GenPod’s production facilities, and to the turbine generator to produce electricity. The turbines will clean-burn the hydrocarbons to generate electricity, and the electricity will flow through transformers to power the compute on the GenPods. Each GenPod is equipped with a StarLink Base Station to receive and transmit processing requests from the client.

The MOPU will be positioned on the natural gas reserve until it is depleted. When a well is depleted, we can simply move and connect the MOPU to a new well. The decommissioning of the depleted well is a small cost.

A smaller natural gas reservoir will provide enough power for 2-3 years of AI production. A larger reservoir will power AI GenPods for 3-7 years. This is considerable power output per reservoir and there are hundreds of these untapped resources throughout the Gulf.

Cooling

Land-based data centers use air conditioning and scarce freshwater resources for cooling, which is expensive on multiple fronts. As a result, running a cooling operation in the Gulf of America presents two key environmental advantages.

First, we can use an unlimited amount of seawater from the Gulf to cool the GenPods. The cooling system pumps ocean water into a large holding tank. The water from the holding tank flows over radiators filled with water heated by the AI operations. The hot water in the radiator is cooled by ocean water circulating through the cooling tank. This will happen continuously to cool AI servers, at no cost to consumers or the environment.

Second, this means that these data centers will not drain any domestic freshwater resources. AI data centers increasingly use so much power and run so hot that they need abundant water resources to keep them cool. A Gulf-based solution offers significant water-use savings.

NIMBY Issues

A big benefit of Gulf-based AI development is that it uses infrastructure (drilling rigs, work boats, shipyards, etc.) that already exists because of oil and gas exploration and extraction. By using these existing resources, GGAI avoids a host of social issues that are emerging from AI data centers. In this case, no municipalities or neighbors will be negatively affected by these operations. Nor will tourism or sightseeing be negatively affected by these operations.

And because natural gas is the primary energy source, pollution and emissions are far lower and are less of a concern.

Regulatory Issues

Everything related to this operation (drilling, extraction, gas, power generation, air quality, water usage) is already regulated by the US government. The permit required to operate a GenPod is already regulated by BSEE. The rig certification is through the American Bureau of Shipping (ABS), the Bureau of Safety and Environmental Enforcement (BSEE), and the Coast Guard.

Other Considerations

MOPUs and GenPods are designed to operate during extreme weather events, such as hurricanes. Normal operating procedure will be to have the GenPod jacked up above hurricane wave height. In this case, we can jack the MOPU above water level, keeping the GenPods out of the most damaging storms.

The logical choice for more sustainable AI

The biggest issue is that this solution can be ready within months, not years. The need and desire for more AI solutions are only growing. Every firm in the world seems to be investing in it, and today, we already see that AI is part of much of the technology we consume. To keep the US at the forefront of AI, we must address energy use more efficiently and effectively.

Tapping into energy in the Gulf offers a wide array of benefits that can be operational within a year, making this solution the most logical choice.