Dollars per Watt

TBPN，科技每日秀 Substack 专栏。关注

Happy Wednesday.

We’re live on YouTube and 𝕏.

The current thing in tech and business is everything OpenAI announced at DevDay.

Today’s Lineup

NVIDIA Senior Director of AI Software Ali Golshan at 12:00 PM
CoreWeave CEO Mike Intrator at 12:20 PM
Robinhood Co-Founder & CEO Vlad Tenev at 12:35 PM
Flow Engineering Founder & CEO Pari Singh at 12:55 PM
Elise AI Co-Founder & CEO Minna Song at 1:05 PM
MongoDB Interim President & CEO Dev Ittycheria at 1:15 PM
Imprint Founder & CEO Daragh Murphy at 1:30 PM
OuterSignal Co-Founder Noah Friedman at 1:40 PM

Run of Show

Dollars per Watt

Great discussion featuring Elon Musk, Jensen Huang and Gavin Baker yesterday at the America.gov summit.

Here’s the key section of the transcript:

Elon Musk: Just to frame this: the average power consumption in the United States is about 500 gigawatts.So every 5 gigawatts of incremental steady-state production is a 1% increase in the power used.And I think that will actually materialize as approximately a 1% increase in GDP, because the intelligence per watt keeps increasing.If you plot how useful intelligence per watt is increasing, it’s happening at both levels: at the hardware level, Jensen’s GPUs keep getting better in terms of compute per watt, and the algorithms also keep getting better at producing more intelligence per watt.So I think that’s an interesting point. I would bet anyone that a 1% increase in power usage corresponds to roughly a 1% increase in GDP.And therefore 10 gigawatts would be a 2% increase in GDP.Gavin Baker: Let’s do the simple math. Each gigawatt can probably generate, call it, $40 or $50 billion of economic output each year.Elon Musk: Well, 1% of the United States—Gavin Baker: About $300 billion.Elon Musk: Yeah. So 5 gigawatts would be $300 billion.Gavin Baker: Exactly. So about $60 billion per gigawatt.Elon Musk: Yeah, that’s right.Gavin Baker: The math maths.

So there are a bunch of interesting stats flowing downstream of this math. It’s actually convenient that “giga” is a billion, so just as a gigawatt is a billion watts, a gigadollar would be a billion dollars. The average electrical load in the United States is indeed around 500GW, it was actually closer to 479 GW average in 2025, but ~500GW is a very reasonable round number. US GDP for 2025 was roughly $30.8 trillion. So the simple math is America currently produces roughly $60–65 billion of GDP per continuously consumed gigawatt.

What’s interesting is that the frontier labs are tracking almost perfectly in line with this, both Anthropic and OpenAI are reportedly sitting between $60b and $70b in terms of ARR, and both of them have around 1 GW live (although way more has been contracted).

Now, two important notes. First, this is obviously just a moment in time snapshot. The labs have contracted way more compute, and if revenue doesn’t increase proportionally, the ratio falls dramatically. On the flip slide, Elon’s point about “useful intelligence per watt increasing” could increase this ratio significantly to the upside.

The second factor is that GDP ≠ revenue. GDP measures value added. OpenAI/Anthropic revenue includes payments that ultimately pay Nvidia, AWS, Oracle, electricity generators, employees, etc. If you calculated the labs’ actual value added, the number would be substantially lower.

But it’s a fascinating coincidence that, right at the moment it feels like we are entering the AGI era, one continuously powered watt in the American economy supports about $65/year of GDP and one watt of frontier AI compute is supporting roughly $65/year of AI revenue.

My last question was how has this tracked over time, so here’s the economic productivity of American energy over the last hundred years.

Clip Spotlight: Sam Altman says the advice to pick one thing to build and pour all your focus into it no longer applies.

"The canonical advice that was correct two years ago was, 'Pick one, and put all your effort into it, because you can't build all 30 things.'"

"You can now build all 30 things. And you can get people to use them all. And you can work more on the ones that work."

"So I think this mindset of lots of ideas, quick feedback, and being open to a lot of things not working is great for this new era for how people build."