Generative Artificial Intelligence (GenAI) development consists of three stages. All must fire for the technology to meet its promises.
The first stage is infrastructure, and we already know the answer to that. Nvidia got to a $2 trillion market cap, and ARM Holdings is worth just as much (on a price to sales basis) because the Cloud Czars are buying infrastructure as fast as they can. That means upgrading hundreds of huge data centers to handle AI workloads, which are many times larger than those of previous Internet applications.
The second stage is development, and here too the early results are promising, just not convincing. GenAI is an outgrowth of database computing, and companies that were born in that field as zooming. Palantir, Salesforce, and ServiceNow are on fire, even before all the new revenue kicks in. Salesforce sales, for instance, were up just 11% year over year in the most recent quarter. But the stock price has risen 50% since early in November.
The third stage is application. That’s where I have questions. ChatGPT, the leading GenAI app, had just $30 million in revenue as of December. That’s after 110 million downloads of its mobile app. Adobe Systems, which is a much more mature GenAI application, saw 2023 revenue rise just 10%.
This is where the rubber meets the road, and this is why you can expect GenAI to have a “dot bomb” moment soon. Just as in the late 1990s, expectations are way ahead of results. That doesn’t mean the technology is underhyped. The Internet wasn’t. But the “trough of disillusionment” is ahead, and we’re going there with all deliberate speed.
