Musk Says AI Chip Production Capacity Will Exceed the Powerable Limit by Year-End
On the Dwarkesh Patel podcast, Musk judged that by the end of this year, the number of AI chips manufactured globally will exceed the actual limit of what can be powered and lit up, and that millions of top-tier GPUs may pile up in warehouses with nowhere to plug in. His basis: chip production capacity is growing exponentially, while in most regions of the world outside China, grid power generation is nearly flat, and the approval cycle for building substations and power plants often runs five to ten years. He also mentioned Tesla Megapack energy storage, the xAI Memphis compute cluster, and SpaceX's orbital solar data center plans, arguing that power, not compute, will become the industry's bottleneck.
Why read this
AI score 58
Musk compared the two curves of chip production capacity and grid power supply, pointing out the physical constraints on compute expansion.
Topics
Body · AI translation
In the Dwarkesh Patel podcast, Musk made a blunt statement: the AI boom is about to hit a hard physical wall by the end of this year.
Musk's judgment is extremely cold: by the end of this year, the number of AI chips manufactured globally will completely exceed the upper limit of what can actually be powered on,
and by then millions of top-tier GPUs will be piled up in warehouses with nowhere to plug them in.
Right now chip production capacity is exploding exponentially, but the global power grid is essentially standing still, and when these two lines intersect this year, millions of top-tier chips will be manufacturable but have nowhere to be powered on.
For the past two years everyone has been watching Nvidia's shipment volumes, thinking the AI race is about who has more cards in hand and whose model parameters are bigger.
But from first principles he drew out two unavoidable curves: one is chip production capacity, with TSMC and the major manufacturers running three shifts and exploding exponentially,
the other is power supply, and apart from China, which is still frantically building infrastructure, the power generation of the grid in the vast majority of regions worldwide is almost a flat line.
These two lines completely intersected this year,
chip manufacturing can be rapidly scaled up with code and fabs, but the approval and construction cycles for substations, high-voltage grids and large power plants often have to be calculated in five or ten years. The entire industry is about to receive a painfully costly hardware lesson: the absolute bottleneck of the future is not computing power, but the most primitive electricity.
Whoever can power on and light up the most chips at the fastest speed is the one who gets to stay at the table.
Of course, Musk himself holds Tesla Megapack energy storage, the xAI Memphis supercomputing cluster, and is also planning SpaceX's orbital solar data centers, stressing that the power cliff naturally fits his business layout.
But this physical reality is right in front of us: the infinite expansion at the software level will ultimately be firmly held down by the crudest laws of physics.
In the future, the core asset by which an AI giant is measured may no longer be how many top scientists it has recruited, but how large a transformer it has occupied next to a grid hub.
The original podcast source and more infrastructure details Musk discussed in the interview are in the first comment 👇 https://t.co/ywiFHqN1rq