In this week’s Perspectives Plus subscriber-only issue 💎, I want to reflect on the state of AI from an adoption perspective. Specifically, adoption that turns into sales of software licenses and services.

Unsurprisingly, we heard Microsoft talk about everything related to Copilot and the agentic future of work at Ignite 2024 (see earlier weekly perspectives for a quick recap). There certainly is a lot of technological investment made, yet how are we on track to meet the expectations from a business perspective? Since it’s not just about future products anymore but something sold already today, we the customers/partners should take a critical look at signals of whether others are buying into the vendor AI pitch or not.

Copilot doesn’t exist in a silo, which is why I also try to keep track of the news and sentiment around other key players. Salesforce is obviously close to Microsoft in the business applications market, yet their entire business segment is just one slice of revenue that Microsoft amasses, so it’s not a 1:1 comparison. OpenAI, on the other hand, is the poster child of the brand new “AI first” business opportunity. Zero legacy on software, everything to gain (and lose) on their march towards AGI promised land.

They all sell software subscriptions from their clouds. How’s that business going? Let’s list a few observations and try to draw some conclusions based on this analysis. Starting from the new kid on the block.

OpenAI

There probably hasn’t ever been a company quite like OpenAI before. This isn’t merely due to the way they shocked the world with the ChatGPT launch two years ago. It’s also about the unimaginable amount of money they are throwing at both running the services (including APIs) and training the next generation of foundational models.

The initial funding for handling the GenAI resource demand shockwave came from Microsoft, in how they agreed to provide the server capacity in exchange for stocks. Regardless of how the company valuation keeps climbing up, there doesn’t appear to be any real-world path for them to stop bleeding money and reach the enterprise software margins of MSFT and SFDC. By being more successful, they just lose more money.

To dive into the mysterious business logic (or lack of it) behind OpenAI’s figures, read this:

The big bet on one day reaching AGI is one route that the company might turn a profit - via a business model that human intelligence alone presumably cannot grasp… The original contract clause that OpenAI drafted for capping Microsoft’s unlimited access to their IPR at “the AGI moment” is now reported to be reconsidered. Confirming that the company’s valuation relies massively on these three letters, however they may eventually be defined.

In the meantime, how does OpenAI make money? Today, it is still mainly through ChatGPT:

Born almost accidentally as a publicly available demo of GPT-3+, there surely wasn’t a lot of market research behind the original price level of ChatGPT Plus at $20/month. Given that OpenAI is nowhere near turning a profit with their current operating expenses vs. revenue, perhaps the monthly price tag should have been higher? The topic of gradually raising the Plus subscription price has been discovered from internal OpenAI documents.

That’s old news by now, though. Last week’s introduction of ChatGPT Pro showed another way for covering the true cost of GenAI services. Going 10x on the previous price point of ChatGPT Plus subscription, this sure doesn’t sound like an obvious upgrade for the 10 million existing subscribers. It’s almost as if the Pro subscription was launched as a way for the AI fanatics to signal that they’re “true believers” in the business benefits that ChatGPT can deliver them.

When you put things into more traditional tech business perspective, though, it’s really all about positioning your product. The same underlying core technology and services can be sold at wildly different price points. Comparing the “from free to a hundred quid” models between OpenAI with its ChatGPT and Microsoft with its Power Platform reveals strong similarities:

Subscribe to Plus to read the rest.