As I learn more about what is happening with the AI wave, and as I talk with folks in the industry, I am struck by the incredible scarcity orientation. Companies that were sitting on top of the world just two years ago are absolutely scrambling. They are rethinking, re-strategizing and reinvesting, changing almost everything to ride the wave, rather than get crushed by it.
It is almost amusing to see Google worried about access to computing power. It is almost heartbreaking seeing Apple desperately trying to balance its promise as “the privacy company” with the need to give Siri, MacOS and iOS AI-based brain transplants.
The new reality is that you need abundant training data, of super high quality, freely usable, and\ ample compute power on the latest arrays of AI tuned GPUs in order to constantly retrain and try out different models and types of models, and you need surfaces with lots of users to deliver intelligence into and derive data from.
Companies that lack access to any of these are in a world of downward-spiraling hurt. If you don’t have the data, your product / service can’t benefit from the exponentiating intelligence wave. If you have the data but not the crunch, you can’t leverage that data using the latest models. Your product gets relatively stupider, much more quickly. If you don't have the users, you can't experiment, evolve and learn.
As we’ve seen from the ChatGPT- 4o announcement, those few companies that have all three are releasing more and more frequent generations of AI capability, leapfrogging everyone else.
Let’s look at the current big players:
OpenAI
Has built a distinct advantage, being ahead of the curve and training on the widest possible data sets. Fueled by investment from Microsoft, and with direct subscriptions, OpenAI is further gathering prompt information from its many users, and feeding that back into better and better models. As the mindshare leader, OpenAI has been able to acquire premium crunch and maintain their edge. OpenAI is already struggling with how to maintain this position of centrality, as it doesn’t have a traditional source of customer/user data.
Microsoft
On top of their stake in OpenAI and strong partnership, Microsoft has essentially acquired Inflection AI and is building on its own internal AI knowledge. In terms of crunch, Microsoft has the entire Azure platform, which was already growing faster than the other platform competitors. In terms of data, Microsoft already houses, or at least handles, a significant majority of enterprise data. By building copilot into every part of Microsoft’s portfolio, they are, like OpenAI, harvesting incredible amounts of prompt and usage data, enabling the creation of smarter and smarter software.
Meta
Although I hate to admit it, Meta is sitting on a treasure trove of personal data. Social Networks are addictive, and Meta has always done a good job of managing its T&Cs such that it can make liberal use of personal data to personalize content, features, and ads. Though Meta doesn’t offer a public facing cloud, it does have world class infrastructure, and a cash cow in its ads business. Meta will continue to deploy AI to do a better job of engaging users, though this alone won’t offer the kind of revenue growth that Microsoft is poised to enjoy.
Amazon
Amazon is still the biggest cloud platform. They handle or house tremendous amounts of enterprise data. However, they are less of a player at the corporate productivity layer than Microsoft. That is, AWS hosts storage and compute, but has much less involvement in the “meaning layer” than Microsoft. Though they may be able to access the documents, they can’t access the signals associated with use of those docs, at least not as easily. In terms of consumers, though, Amazon is in a very good position, basically knowing what consumers are buying, when, and why. This will create a virtuous cycle for them in delivering ever more dominant shopping experiences and targeted ads.
Sad to say, Google is in a hard place. They were already third and lagging in Cloud (I have lots of theories as to why, but this article isn’t the place for that), and are now under siege for Search (which also means Ads). As people start to use chatbot interfaces instead of Search, Google’s position of primacy is not only in question, it could decline exponentially. Google has scrambled to build Gemini into Search, seeking to intermediate itself, but this legitimizes the idea of using something “other than Google Search”, pushing experimentation with Copilot in Bing, ChatGPT and others. In terms of enterprise data, Google is nowhere near the player that Microsoft is, though they do have strong surfaces for those that have adopted Workspace.
Apple
Apple, IMHO, has painted itself into a corner. As evidenced by the latest product announcements, they are positioning their role in AI as being about edge AI. While it is true that smart edge devices will enable faster response in some use cases, the big story in AI will continue to be personalized results of well-trained global models. That is, without access to both private and federated data, your AI won’t be competitive. Apple is rumored to be making agreements with OpenAI to provide the models that will power iOS18. This seems, to me, like an admission that they can’t or won’t compete in the base model space. Managing the relationship with a close Microsoft partner will be … interesting. Apple does have something only Google shares - insight into actual use of the most pervasive computing platform - the mobile OS. I predict a huge backpedaling from Apple to get permission to use mobile computing data in order to make Siri smart again.
So, to summarize, the companies in the best position have access to ample high value training data, direct signals from a wide user base, and massive computing infrastructure at least for training. There are other players than those listed above, but all will be trying to put together this winning combination.
Put simply, if you want to serve the user/customer better, you have to understand them better, and you have to deliver products that build on that understanding. That means data, feedback and crunch. If you don’t have all three, in some kind of protectable form, it will be increasingly painful.
Comments