No generator produces more power than its capacity.<\/li>\n<\/ol>\nThis doesn\u2019t go into startup costs, shutdown costs, reliability and regional information, but it\u2019s a useful lens because the model spits out a bunch of shadow prices.<\/p>\n
Those shadow prices are the point. They tell you what\u2019s actually tight.<\/p>\n
Founders can do the same thing without building an optimization model.<\/p>\n
Write down your objective function in one sentence, then list the constraints that can stop it. List the problems that can actually prevent the outcome.<\/p>\n
If your objective is to \u201cship an AI product that people use,\u201d your constraints might be that you can\u2019t get power, compute, a data center built out fast enough to matter, the agreement structure to work or make the unit economics work.<\/p>\n
The reason this step matters is simple: if you don\u2019t know what\u2019s tight, you\u2019ll build the wrong thing.<\/p>\n
2. Use logistics as your reality check<\/h2>\n
Logistics have a nice way of forcing you to deal with reality. It doesn\u2019t care about narrative reasons. You either procure the supply that you need, or you don\u2019t.<\/p>\n
Operations researchers in big tech companies are used to this frame of thinking. They\u2019ll literally write up a linear program or mixed integer optimization to represent how to organize their data center. This degree of modeling might be surprising to smaller companies.<\/p>\n
And when there\u2019s no way to satisfy all of your constraints, the optimization tells you that. Of course, there\u2019s a cost to fixing it. You might have to relax some of your constraints. You can\u2019t always get everything that you want.<\/p>\n
That\u2019s the whole point: constraints show up in the number. If you\u2019re building in AI, stop asking, \u201cWhat is everyone doing?\u201d and start asking, \u201cWhat would make my objective function move?\u201d In your business, the objective might be deployment timelines, inference cost, latency or the ability to actually get compute online.<\/p>\n
This is why I like the phrase \u201cbinding constraint.\u201d It forces honesty. It forces you to say, \u201cWhat is the thing that\u2019s stopping me?\u201d<\/p>\n
3. Treat compute like a menu<\/h2>\n
I\u2019ve seen people reach for a default answer because it\u2019s what\u2019s advertised most broadly: \u201cbuild a data center.\u201d But there might be other ways to get the compute you need.<\/p>\n
When I think about AI companies, I translate the same question into founder language: How should AI companies try to get the power or the compute that they need?<\/p>\n
One way is that you get your own GPUs. You find the power. Another option is a data center build-out, which also needs GPUs. You could also use inference providers that already exist, or rent data centers, or get them from somewhere like Crusoe.<\/p>\n
It\u2019s a fairly common build, buy or joint-venture consulting problem. And I like framing it this way because it forces you to stop pretending there\u2019s one default path, especially when \u201cwhat\u2019s advertised most broadly\u201d becomes the roadmap.<\/p>\n
So take the menu of options and force a decision by asking one question for each option: is it the cheapest way to solve our binding constraint, and what is the new constraint it introduces? Owning GPUs solves one problem, and then you\u2019re back to \u201cfind the power.\u201d<\/p>\n
A data center sounds direct, but if the timeline doesn\u2019t work, then you have a new binding constraint. Renting or using inference providers can avoid one binding constraint, but you might trade it for another.<\/p>\n
Stop copying the loud bottleneck<\/h2>\n
If you remember one thing, make it this: the limiter isn\u2019t always what everyone is talking about.<\/p>\n
The first step is always the same: define the objective function, then ask what the constraints are. The commodities markets are a good teacher because the name of the game has always been procurement. Supply has to equal demand, you can only ship so much, and you can only produce so much yourself. When something is tight, the price tells you. When something is loose, the price tells you that too.<\/p>\n
AI infrastructure is heading into the same kind of reality. You can obsess over chips, but you still have to find the power. And once you see the problem through that lens, the roadmap gets a lot clearer: pick your option in the build-by-joint-venture spectrum, be honest about what\u2019s actually tight, and plan around the constraint that\u2019s real, not the loud one.<\/p>\n<\/p><\/div>\n