VCs zoom in in 2 years of fast growth and apply a 30x ARR. Pipelines, wires and factories are not sexy to the average investor. The sleek UX still gets all the attention.
One surprise for me in the last 12 months is how much value capture is happening at the model layer. I would have thought that the models would behave like near-perfect substitutes and that competition and open source (Meta) would push inference prices to near zero. But it looks increasingly like OpenAI will end up being a valuable toll booth.
Hm, the models are all converging on the same set of capabilities. So they look like commodities to me, not value accumulators. Most of the value is accumulating in the infrastructure: data centers, power generation, land, permits, water cooling, GPUs. There's a reason OpenAI is spending $500 bln on its Stargate project!
Totally right. But even if functionality converges to commodity parity it doesn’t necessarily mean that all economic profits at that level get competed away, because other things like distribution matter too. Exxon and Shell sell the exact same product but they each have distinct contractually owned distribution channels that they can lean on with some (nonzero but not large) level of pricing power. So some models look likely to introduce switching fiction in such a way that they can hold onto distribution even with totally commoditized products (e.g. better memory). The fact that OpenAI can race to an early lead and has the opportunity to cement some of those customer relationships over time (through various means) feels significant.
All of that just to say that I was expecting the model business to play out like a commodity market where suppliers are price takers, but the reality might be more like the Standard Oil era where business tactics and brinksmanship (and path dependence) really matter for establishing who owns market share when the dust settles.
Agree with your analysis, although I feel the P&L you have shown in the article might be more generous, cogs are under estimated for AI. In my estimation, it will be higher for AI and hence net margin is likely more negative.
VCs zoom in in 2 years of fast growth and apply a 30x ARR. Pipelines, wires and factories are not sexy to the average investor. The sleek UX still gets all the attention.
One surprise for me in the last 12 months is how much value capture is happening at the model layer. I would have thought that the models would behave like near-perfect substitutes and that competition and open source (Meta) would push inference prices to near zero. But it looks increasingly like OpenAI will end up being a valuable toll booth.
Hm, the models are all converging on the same set of capabilities. So they look like commodities to me, not value accumulators. Most of the value is accumulating in the infrastructure: data centers, power generation, land, permits, water cooling, GPUs. There's a reason OpenAI is spending $500 bln on its Stargate project!
Totally right. But even if functionality converges to commodity parity it doesn’t necessarily mean that all economic profits at that level get competed away, because other things like distribution matter too. Exxon and Shell sell the exact same product but they each have distinct contractually owned distribution channels that they can lean on with some (nonzero but not large) level of pricing power. So some models look likely to introduce switching fiction in such a way that they can hold onto distribution even with totally commoditized products (e.g. better memory). The fact that OpenAI can race to an early lead and has the opportunity to cement some of those customer relationships over time (through various means) feels significant.
All of that just to say that I was expecting the model business to play out like a commodity market where suppliers are price takers, but the reality might be more like the Standard Oil era where business tactics and brinksmanship (and path dependence) really matter for establishing who owns market share when the dust settles.
Corn syrup is the output of a huge industrial process. But Coca-Cola still is more profitable than corn farming.
Agree with your analysis, although I feel the P&L you have shown in the article might be more generous, cogs are under estimated for AI. In my estimation, it will be higher for AI and hence net margin is likely more negative.
My two articles related to this
Cloud vs AI: Same Stack Different Game
https://open.substack.com/pub/pramodhmallipatna/p/cloud-vs-ai-same-stack-different
Tokenomics. Analysis on the cost, pricing and economics
https://open.substack.com/pub/pramodhmallipatna/p/the-token-economy
How long until savvy VC’s figure this out?
Some VCs understand this. Not all VCs are dumping cash into ai startups.