AI and Affordability Dominate First Day of BloombergNEF Energy Summit

Article content

(Bloomberg) — After two decades of flat electric demand, artificial intelligence has emerged as the US power industry’s great hope. More recently, it’s become clear AI is also its biggest threat.

Article content

It’s no wonder, then, that AI, data centers and their impact on the utility bills of homes and businesses dominated conversation Monday at BloombergNEF’s annual summit in San Francisco.

Article content

Article content

Already, the data center boom has shaped the US economy. In November, it swayed elections in New Jersey, Virginia and Georgia. It’s now poised to play a key role in the congressional midterms this fall.

Article content

Article content

Here are five takeaways from the first day of the BNEF summit.

Article content

Article content

Affordability is an 800-pound gorilla

Article content

With concerns about energy affordability mounting, data centers are under increasing pressure to limit the burden on individuals. Otherwise, “consumers may end up holding the bag,” said RMI co-founder and chairman emeritus Amory Lovins.

Article content

Data centers are the main driver of utility cost growth, so “they also need to be the ones to cover that,” said Ryan Wiser, senior scientist, Lawrence Berkeley National Laboratory. 

Article content

Efforts are being made to address this. Just this month, President Donald Trump and a group of governors in the Northeast and Mid-Atlantic agreed to push for an emergency wholesale electricity auction that would effectively compel tech companies to fund new power plants. The plan is intended to address two challenges: getting data centers the power they need while attempting to tame utility bills. 

Article content

Nuclear may miss the initial AI moment

Article content

Tech companies have been unequivocal: They want nuclear energy, which is clean and can provide round-the-clock power. Meta Platforms Inc. struck major deals with startups and Microsoft Corp. has led the charge to restart a shuttered power plant, to name two recent moves to kickstart the nuclear industry.

Article content

Article content

Yet not a single new small modular reactor, or SMR, has been built in the US, and only one design has been approved by the US Nuclear Regulatory Commission. And a traditional nuclear power plant takes about a decade to stand up, far longer than the timeframe required to meet burgeoning AI demand.

Article content

The longstanding question for nuclear power is whether or not reactors can “be built on time, on budget and actually be competitive with natural gas,” said BNEF analyst Musfika Mishi. 

Article content

Today, nuclear energy in the US is three times more expensive than natural gas-fired power, Mishi said. While SMR companies have promised lower costs, she said, time will tell if they can succeed.

Article content

Is AI demand overhyped?

Article content

There’s an even more fundamental question than what energy technologies are best-positioned to meet demand: Just how much demand will there be? PJM Interconnection, operator of the biggest US grid, recently dialed back its electricity use forecast for the summer of 2027 after taking a closer look at data center connection requests. But on the whole, US data center demand is set to grow.