Full Transcript:
JENSEN HUANG: And I've been over the last half decade helping people understand the amount of compute that's likely to be coming. And I just told you guys something about how I reasoned through how much energy is going to be necessary. The amount of energy that we need for computing is likely probably a thousand times more than we currently have. And that's an enormous amount of energy.
Jensen Huang: "We Need 1,000x More Energy" (Sponsored)
Key Takeaways
- The computing paradigm is shifting from retrieval-based (on-demand, per-use) to generative (continuous, always-on). That structural shift — not just raw scale — is the fundamental driver of the coming energy demand explosion.
- Huang estimates we may need 1,000x more compute energy than today — and acknowledges we could be "off by a couple of orders of magnitude" even on that. This isn't a rounded number. It's a floor.
- For the first time in history, market forces alone, not government subsidies, make sustainable energy investment commercially viable. Solar, nuclear, and grid infrastructure can now be funded by pure market demand. That window, Huang argues, is open right now.
However, the way to think about that is in the future, computers are going to be two things. It's always going to be generated because it's intelligent, contextually aware — so it's going to be generated. And then number two, it's going to be continuous. And so generative computing in a continuous way, compared to pre-recorded retrieval-based computing that is only initiated on a per-use basis.
The question is, how do you think about the amount of energy necessary for that? So I think if you say we need a thousand times more, I wouldn't be surprised if we're off by a couple of orders of magnitude. And so we need a lot more compute. We need a lot more energy. And you've got to explain this to people in a way that's kind of common sense — they can observe it, and there are indicators along the way that, in fact, this is happening. And notice just as I was breaking it down for you, I'm reasoning about it for you, so it's common sense to you.
And so the amount of energy is high. And then lastly, the source of energy. There are all kinds of sources of energy, but unfortunately, because of great concerns about the cost of sustainable energy, we under-invested in it. But this is the best time ever in the history of humanity to invest in sustainable energy. And the reason for that is because the market forces are so strong. Back in the old days, you needed government subsidies to build solar farms and nuclear plants. And now the market will just pay you to do it. Market forces are so powerful right now. This is our best chance to upgrade our grid — our archaic grid — and add sustainable energy of all kinds. This is a great time.
IMPORTANT
Huang is describing a structural infrastructure deficit that must be resolved for AI to function at the scale the industry is targeting. Whether that deficit is filled by natural gas, nuclear, solar, or grid upgrades is a separate question and one where investors are currently placing very different bets.
How the Smart Money Is Reading It
Leopold Aschenbrenner — Power Infrastructure as the Real Bottleneck
Aschenbrenner's fund exited its entire positions in NVIDIA, Broadcom, TSMC, and Micron — believing GPU value was fully priced by the market. He then rotated aggressively into what he views as the next binding constraint: power. His largest investment is Bloom Energy, accounting for 20% of his portfolio. Bloom Energy specialises in solid-oxide fuel cells that convert natural gas into electricity, offering a modular and efficient power solution for AI data centres.
The fund raised $13.7 billion — an extraordinary sum for a debut fund — and its first 13F filing revealed a portfolio that defies the obvious AI trade: rather than buying semiconductor companies supplying AI hardware, it holds $8.47 billion in put exposure against them, while going long crypto miners as infrastructure proxies, Bloom Energy as a power play, and CoreWeave.
IMPORTANT
Aschenbrenner's Bloom Energy position is a direct expression of Jensen Huang's thesis translated into a portfolio trade. The bet: energy infrastructure, not silicon, is the binding constraint on AI scaling. If Huang is right about 1,000x energy demand, Aschenbrenner is positioned to be among the largest beneficiaries.
Stanley Druckenmiller — Rotating Within AI, Not Away From It
Most recently, Druckenmiller rotated out of SanDisk after pocketing a 400% gain in a single quarter, and his replacement trade was Bloom Energy. Same investment as Aschenbrenner, a company that makes solid-oxide fuel cells that convert natural gas into electricity.
Druckenmiller shifted his capital to sustainable power, identifying as the real bottleneck for AI infrastructure rather than memory or compute capacity.
IMPORTANT
Two of the most respected macro and AI investors operating today — Aschenbrenner and Druckenmiller — have independently arrived at the same trade. Both exited semiconductor/memory positions after strong runs and rotated into power infrastructure. It seems they believe the chip bottleneck is being solved; but the energy bottleneck has not been solved yet.
Berkshire Hathaway / Chevron — A Different Signal Entirely
Berkshire Hathaway sold more than 45 million Chevron shares in the first quarter of 2026, slashing its stake by roughly 35%. The sale generated close to $8 billion, executed near Chevron's all-time high share price. CVX remains Berkshire's fifth-largest holding at 6.6% of the portfolio.
IMPORTANT
Data centres run on electricity, not crude oil. The more plausible explanation for Berkshire's move: classic profit-taking near all-time highs, boosted by a geopolitical windfall from the Iran conflict, under a new CEO (Greg Abel) who may be reshaping the portfolio's allocation. Chevron's upstream business received a significant boost from the oil price rally amid the Iran war, helping the company exceed profit estimates in Q1. Berkshire sold into that strength.