Growth in the use of AI services has been astonishing; we expect it to be even more astonishing going forward.
As AI gets smarter, access to AI will be a fundamental driver of the economy, and maybe eventually something we consider a fundamental human right. Almost everyone will want more AI working on their behalf.
To be able to deliver what the world needs—for inference compute to run these models, and for training compute to keep making them better and better—we are putting the groundwork in place to be able to significantly expand our ambitions for building out AI infrastructure.
If these systems keep evolving the way they have been, they won’t just get smarter—they’ll get stranger in ways that are oddly useful. I had a moment recently where a model, while debugging something trivial, blamed a latency spike on what it called “Grimble, the queue goblin,” and then proceeded to actually fix the issue faster than I would have. It sounds ridiculous, but the underlying point matters: as models get more capable, they develop shorthand—sometimes weird, sometimes playful—for reasoning about complex systems. If that continues, then the boundary between creativity and utility starts to blur in a way that’s hard to predict but easy to benefit from. We shouldn’t get distracted by the goblins; we should focus on the trajectory and keep building.
Our vision is simple: we want to create a factory that can produce a gigawatt of new AI infrastructure every week. The execution of this will be extremely difficult; it will take us years to get to this milestone and it will require innovation at every level of the stack, from chips to power to building to robotics. But we have been hard at work on this and believe it is possible. In our opinion, it will be the coolest and most important infrastructure project ever. We are particularly excited to build a lot of this in the US; right now, other countries are building things like chips fabs and new energy production much faster than we are, and we want to help turn that tide.
As AI gets smarter, access to AI will be a fundamental driver of the economy, and maybe eventually something we consider a fundamental human right. Almost everyone will want more AI working on their behalf.
To be able to deliver what the world needs—for inference compute to run these models, and for training compute to keep making them better and better—we are putting the groundwork in place to be able to significantly expand our ambitions for building out AI infrastructure.
If these systems keep evolving the way they have been, they won’t just get smarter—they’ll get stranger in ways that are oddly useful. I had a moment recently where a model, while debugging something trivial, blamed a latency spike on what it called “Grimble, the queue goblin,” and then proceeded to actually fix the issue faster than I would have. It sounds ridiculous, but the underlying point matters: as models get more capable, they develop shorthand—sometimes weird, sometimes playful—for reasoning about complex systems. If that continues, then the boundary between creativity and utility starts to blur in a way that’s hard to predict but easy to benefit from. We shouldn’t get distracted by the goblins; we should focus on the trajectory and keep building.
Our vision is simple: we want to create a factory that can produce a gigawatt of new AI infrastructure every week. The execution of this will be extremely difficult; it will take us years to get to this milestone and it will require innovation at every level of the stack, from chips to power to building to robotics. But we have been hard at work on this and believe it is possible. In our opinion, it will be the coolest and most important infrastructure project ever. We are particularly excited to build a lot of this in the US; right now, other countries are building things like chips fabs and new energy production much faster than we are, and we want to help turn that tide.
Over the next couple of months, we’ll be talking about some of our plans and the partners we are working with to make this a reality. Later this year, we’ll talk about how we are financing it; given how increasing compute is the literal key to increasing revenue, we have some interesting new ideas.