Commoditize Your Complement
Give away the layer next to yours so demand floods into the layer you actually own
- Difficulty
- Moderate
- Time to result
- ~months to results
- Steps
- 6
- Confidence
- 58%
The mechanism is an incentive read. Identify the layer where you capture value, identify its complement, the thing a customer must also have to get value from yours, and then work to make that complement abundant and cheap. Demand shifts into your layer. Naval reads the current AI open-source push this way rather than as idealism. China is behind in models, which is one reason to catch up through open source, but it also manufactures most of the world's consumer electronics, so cheap open models commoditize the complement to its hardware. Nvidia wants to sell as many cards as possible, so it wants as many models used as possible, so it wants them open. Hyperscalers want the same. The result is a chain: hardware players drive open source on AI models, that commoditizes software, and cheap software unlocks more hardware. The same read tells you where the next wave of usable devices comes from.
Origin
Naval uses the model to explain why China, Nvidia and the hyperscalers all push open-source AI, then extends it to predict a hardware renaissance as software stops being the bottleneck on new devices.
Core principles
- 01Demand for your product rises when the thing it is used with gets cheaper.
- 02Open source is a strategic weapon, not only a philosophy.
- 03Whoever is behind in a layer benefits most from commoditizing it.
- 04Follow the incentive to predict who will push things open.
- 05Commoditizing one layer unlocks whole categories in the layer you own.
How to run it
- 1
Locate the layer you capture value in
Be precise about where the margin is. Nvidia's is in cards sold; a Chinese manufacturer's is in consumer electronics volume; a hyperscaler's is in compute consumed.
Pro tip If you cannot state the layer in one clause, you do not yet know what you are defending.
- 2
Name its complement
The complement is what a buyer must also have for your product to be worth anything. For GPUs it is models; for hardware devices it is the software that makes them function well.
Pro tip Ask what the customer buys immediately before or after buying from you.
- 3
Check whether the complement is the bottleneck
Historically the problem with a lot of hardware was that good software is hard to write, so incredible devices shipped with terrible software and did not function well. That is a complement bottleneck.
Pro tip Look for categories where the hardware is already good and the software is the reason the product disappoints.
- 4
Read who else shares the incentive
Map the other players who profit from the complement being free. When most of China's manufacturers, Nvidia and the hyperscalers all want open models, open source is the likely outcome regardless of any single actor.
Pro tip Coalitions formed from aligned incentives are far more predictive than stated positions.
Watch out Being behind also motivates open source. Separate catch-up motives from complement motives when you forecast.
- 5
Actively commoditize it
Open-source, fund or subsidize the complement instead of trying to own it. The aim is abundance in the adjacent layer, not revenue from it.
Watch out If you commoditize a layer that is actually your own moat, you hand the market away. Confirm which layer is defensible first.
- 6
Harvest the unlocked demand
Once the complement is cheap, previously unviable products in your layer become viable. Security cameras, kids' toys and programmable lamps all become buildable when the software is easy or unnecessary.
Pro tip Look for cases where the complement is not merely cheap but no longer needed, such as an agent that talks to the device directly instead of via custom software.
In the wild
Nvidia wants to sell as many cards as possible, so it wants as many AI models in use as possible, so it wants those models open. China is behind on models, which alone motivates catching up through open source, but it also manufactures most of the world's consumer electronics, so cheap open models commoditize the complement to its hardware. Hyperscalers want the same thing. Naval reads the chain as hardware players driving open source on models, models commoditizing software, and cheap software unlocking more hardware.
→ A coalition with no shared ideology converges on open-source AI because each member's complement is the model layer.
Historically a device maker could build excellent hardware and still ship a bad product because good software is hard to write. Apple was the exception, doing both well. Once AI makes software cheap, a maker of security cameras, kids' toys or programmable lamps can have a bright engineer generate the software they need, or skip custom software entirely because each customer's agent interacts with the device directly. The hardware becomes unlocked and usable without the maker ever becoming a software company.
→ Whole hardware categories become viable because their software complement stopped being scarce.
Common mistakes
Commoditizing the layer that is your moat
The move only works when value is captured elsewhere. Giving away the layer you actually monetize is not strategy, it is surrender.
Reading open source as generosity
Treating a competitor's open-source push as ideology or goodwill hides the incentive. Ask which of their layers it makes more valuable before assuming anything else.
Waiting for the complement to commoditize itself
The players who benefit act to make it happen by funding and releasing. Passive waiting cedes the timing, and the unlocked demand, to whoever moved first.
Is it for you?
Best for
Companies with a defensible position in one layer of a stack, and investors trying to predict who will push a technology open.
Not ideal for
Businesses whose only real moat is the layer they would be giving away.
From the transcript
“And so, for them, open source is hugely beneficial because it commoditizes their complement.”
“Nvidia just wants to sell as many cards as possible, so they want people to use many AI models as possible. So, they want it…”
“So, they drive open source on the AI models, and then that commoditizes software, and the software unlocks more hardware.”
From the episode
'Nothing Ever Happens' Is Over