Natural gas generation is likely to expand as a proliferation of data centers are built across North America, offering a fortuitous advantage for midstream giant Kinder Morgan Inc., according to Chief Commercial Officer Will Brown.
The top executive discussed the Houston-based company’s gas strategy during a keynote address at the LDC Gas Forums Rockies & West conference in San Diego.
“Is it real? Or is it hype?” Brown asked the audience about what data centers could mean for gas demand. It’s both, he answered. “It’s hyper…scale.”
The company foresees solid gas growth across North America for its vast pipeline system by 2030 – under any forecast.
For one thing, more LNG is going to be flowing overseas in 2025. Additional liquefied natural gas capacity is ramping up on the Gulf Coast. Western Canada next year should begin exporting gas – for the first time – to Asian markets. And Mexico continues to draw copious imports from Texas.
Nobody has a handle – yet – on exactly how much natural gas will be used by the data centers used to house artificial intelligence (AI).
Mega-tech sponsors, including Amazon, Meta and Nvidia, are eyeing underutilized capacity already on the grid. Some projects also plan to use power purchase agreements that would lean on renewables and alternative energies.
Natural gas, though, is predicted to be a big player. How big? That’s TBD.
3-6 Bcf/d? That’s ‘Reasonable’
Forecasts are all over the map. Consumption estimates by 2030 range from 1.3 Bcf/d to 10 Bcf/d, with “upside” cases topping 16 Bcf/d, Brown noted.
The company’s forecast is a bit more “reasonable.” It is pegging 3-6 Bcf/d of incremental demand directed to the data centers by 2030.
Based on forecasts, the Northeast could see gas consumption by 2030 increase by nearly one-third because of the data center build out. Gulf Coast demand could climb more than 14%, with a proliferation of projects proposed in Texas.
In the Southeast, incremental gas demand may rise almost 12% by 2030, while the Rockies and the Midwest each may see consumption up by close to 11%. In the Pacific Northwest, demand from the AI centers is forecast to climb 8%, about the same as in the Great Lakes region. And in California, data centers may lead to a 6% increase in generation capacity over the period.
Virginia, already the No. 1 data center location, is likely to continue to house the most newbuilds, based on research. Other states likely to capture more data centers are, in order, Texas, California, Illinois, Oregon, Arizona, Iowa, Georgia, Washington and Pennsylvania.
Kinder is set to be in the mix, with commercial discussions now underway to grab around 5 Bcf/d-plus of overall U.S. incremental power demand capacity. That number includes 1.6 Bcf/d specifically powering data centers.
[Overcoming obstacles in the LNG market: NGI sits down with Baker Botts' Jason Bennett, department chair of global projects, to discuss U.S. LNG projects and the hurdles they currently face. Despite those obstacles cropping up, global demand for LNG is going nowhere, and the U.S. remains a prime supplier. Find out more from NGI’s Hub & Flow podcast.]
‘Hyperscalers Are Coming’
Where the data centers are being built would be in a sweet spot for gas delivery, Brown noted.
The domestic natural gas transmission system, which carries about 40% of U.S. supply, traverses 66,000 miles. It also has stakes in 702 Bcf of working storage capacity – or 15% of the nation’s total.
AI is becoming an indispensable tool, both in everyday life and at the office, Brown noted.
“The hyperscalers are coming. They're here. They've been here for a while…There's going to be a lot more.”
Regardless of the data center opportunities, company research estimates that between 2023 and 2033, the compound annual growth rate for North American gas production will be 2.2%, reaching 153 Bcf/d from 124 Bcf/d.
“We feel very confident in that number because of the fundamentals,” Brown said. Natural gas exports from the Gulf Coast, Western Canada – and potentially Mexico – are forecast to take a big share of the new output.
Even as more alternative generation comes online, and as the hyperscalers seek to use it, “renewables have an inherent intermittency,” Brown told the audience. “They can’t produce 24/7…”
Natural gas is “going to be a lot easier to permit” for the new centers too. “It’s going to be a lot easier” and “much, much quicker…” The hyperscalers want “clean power, but they want their deals at once – way more than they want their clean power…So really it comes down to that we have pipelines” that can supply the power.
The data center movement began to gain steam last year after the fourth iteration of the Generative Pre-trained Transformer, or GPT, was released. GPT, the language model created by U.S.-based OpenAI, is publicly available via ChatGPT.
A ChatGPT query consumes about four times more energy than an email and 10 times more than a Google search, according to data by energy researcher Thunder Said Energy.
The growth in AI “is driving increasing demand for data processing, which is very energy intensive,” Brown said. “As AI models grow in sophistication, so too does their power consumption.”
A “typical” data center was using around 30 MW, he noted. Today, “the hyperscalers are talking about gigawatts and campuses, which means a tremendous amount of power generation will be needed.”
Brown showcased some Kinder research that delved into integrated resource plans, or IRPs, to see what utilities were using in determining their future generation requirements.
Since 2023, “there is 3.9 Bcf/d of incremental data center demand” that the researchers found via utility IRPs and company press releases. That included 1.5 Bcf/d in the western part of the country and 2.4 Bcf/d in the East. Because of “irregular” release dates, some utilities had not yet updated their IRPs, but “additional demand is likely to be announced.”
NGI senior energy analyst Josten Mavez, who also attended the LDC conference, added that large-scale critical operations like the Port of Los Angeles, with strict reliability standards and ambitious renewable mandates, have long been struggling to keep adequate and reliable supply. An outage earlier this month at the Port of Los Angeles, while brief in nature, resulted in multiple disruptions at the terminal, he noted.
“At this point, end users and even municipal utilities” such as the Los Angeles Department of Water and Power, “just want to be able to buy any power they could get,” Mavez said.
“Adding another 8,000 data centers by 2030, especially hyperscalers, whose downtime amounts to about 26 minutes out of an entire year and whose business model depends on operating near 100% uptime, have certainly been put front and center here at LDC and across the industry as a whole.”
