Did someone forward you this? Subscribe here free!
Good morning, Roca Nation. Here are today’s four need-to-know stories:
A federal judge said he would approve a settlement requiring Purdue Pharma to pay at least $7.4B to resolve lawsuits over the company's role in the opioid crisis (free)
President Trump removed tariffs on over 200 food products on Friday
President Trump withdrew his support for Marjorie Taylor Greene over her push to release files related to Jeffrey Epstein
A poll found 40% of American women between the ages of 15 and 44 said they would like to permanently leave the US if they could (free)
By Max Towey
OpenAI’s new video tool Sora 2 is remarkable – not merely in its capabilities but in its costs.
With just a few keystrokes, one user made a mini-South Park episode about pickleball; another generated surveillance footage of Sam Altman shoplifting at Target.
While some marvel at Sora 2’s technology, others wonder about its societal costs: Will it blur the line between truth and reality on social media? Will bad actors use it to blackmail people? Will it plunge kids into the metaverse? Will its benefits actually outweigh its risks?
Debates like these are sure to unfold over the months and years to come, but there is one Sora 2 cost that’s both pressing and undeniable: Its energy costs.
According to MIT, a five-second video on Sora 2 requires roughly 3.4M joules worth of energy – the equivalent of running a microwave for over an hour or biking 38 miles on an e-bike. The math starts to get scary with longer videos: The final rendition of the above mini-South Park episode, for example, required the energy equivalent of running a microwave for 36 hours. For what? A user-generated pickleball spoof that gets 2,000 likes on X?
Sora and ChatGPT’s current usage is but a small sample of what is to come with AI’s energy demands: In recent weeks, OpenAI announced a strategic partnership with Nvidia to deploy 10 gigawatts of Nvidia systems to power OpenAI computing. 10 gigawatts is equivalent to the output of about 10 large nuclear reactors and enough energy to power both New York City and San Diego. That scale might be hard to fathom, but it’s only just one strain AI companies are putting on the power grid and only a small fraction of what is to come.
Tech behemoths like Google, Microsoft, and Amazon are erecting data centers at a breakneck pace, often clustering in “data center alleys” like Northern Virginia, where over 300 facilities already operate.
But what exactly is a data center? How much electricity are they really using? And what’s the cost to consumers? That is the subject of today’s deep-dive.
Let’s start by turning the clock back to the very start of the data center occupation of our power grids: The 2010s.
This full report is for paid subscribers, who fund our journalism. If you start a two-week free trial today, you’ll be automatically entered to win a free year. Once you sign up, you can access all of our articles here!
Editor’s Note
We want to hear from you: Have energy prices risen in your area? Are you concerned about the increased energy demand from data centers? Let us know what you think by replying here.
And if you’re catching up from the weekend, find our most recent articles below:
Thanks for reading. Stay tuned for part two tomorrow.
—Max and Max
