very interesting article, Bob, thank you. It’s so remarkable that you and I and everyone gets these powerful tools for free, in gemini or grok or wherever, you can type in complicated questions and get immediate thoughtful answers, even if not perfect. But to your point, someone’s got to pay for the enormous energy burn that goes into these calculations. consumers are not going to pay much if anything for AI-powered websearch. advertisers will have to pay for that. which implies continued disruption to traditional forms of marketing as we’re already seeing. the more interesting question is whether companies are getting AI tools from providers for less than full cost, and what are the true ROIs. we can all see that LLMs have enormous power, but as people point out there is a “jagged edge” to AI capabilities, meaning that the reliability and accuracy isn’t there for many applications with low tolerance for risk. the shakeout you are referencing is probably only getting started — but that’s a wild guess from someone who doesn’t know
Thank you Ken. This is exactly the tension I’m wrestling with.
We’re watching a power-consuming business without a clear business model.
The tools feel free. The consumer certainly isn’t going to pay much for AI-powered search. That likely means advertising becomes even more important, which implies further disruption to traditional marketing channels. But that only explains part of the ecosystem.
The more interesting question is whether AI providers are currently delivering enterprise value below cost — subsidized by venture capital and equity markets — in anticipation of being able to price that value later. Enterprise software and SaaS companies eventually figured out how to quantify productivity gains and charge accordingly. Perhaps AI firms will as well. That would explain the extraordinary capital they’ve been able to raise.
But it’s difficult to see the retail user as the long-term revenue engine.
In that sense, this feels closer to Amazon in the late 1990s than to a conventional software rollout. Amazon consumed enormous capital for years before its durable cash-flow engine became obvious. Investors were underwriting a structural shift before the monetization was visible.
AI may be similar. The productivity gains are real. The cost structure is real. The business model is still forming.
The shakeout, as you suggest, is likely just beginning.
very interesting article, Bob, thank you. It’s so remarkable that you and I and everyone gets these powerful tools for free, in gemini or grok or wherever, you can type in complicated questions and get immediate thoughtful answers, even if not perfect. But to your point, someone’s got to pay for the enormous energy burn that goes into these calculations. consumers are not going to pay much if anything for AI-powered websearch. advertisers will have to pay for that. which implies continued disruption to traditional forms of marketing as we’re already seeing. the more interesting question is whether companies are getting AI tools from providers for less than full cost, and what are the true ROIs. we can all see that LLMs have enormous power, but as people point out there is a “jagged edge” to AI capabilities, meaning that the reliability and accuracy isn’t there for many applications with low tolerance for risk. the shakeout you are referencing is probably only getting started — but that’s a wild guess from someone who doesn’t know
Thank you Ken. This is exactly the tension I’m wrestling with.
We’re watching a power-consuming business without a clear business model.
The tools feel free. The consumer certainly isn’t going to pay much for AI-powered search. That likely means advertising becomes even more important, which implies further disruption to traditional marketing channels. But that only explains part of the ecosystem.
The more interesting question is whether AI providers are currently delivering enterprise value below cost — subsidized by venture capital and equity markets — in anticipation of being able to price that value later. Enterprise software and SaaS companies eventually figured out how to quantify productivity gains and charge accordingly. Perhaps AI firms will as well. That would explain the extraordinary capital they’ve been able to raise.
But it’s difficult to see the retail user as the long-term revenue engine.
In that sense, this feels closer to Amazon in the late 1990s than to a conventional software rollout. Amazon consumed enormous capital for years before its durable cash-flow engine became obvious. Investors were underwriting a structural shift before the monetization was visible.
AI may be similar. The productivity gains are real. The cost structure is real. The business model is still forming.
The shakeout, as you suggest, is likely just beginning.