With every new technology, there is a learning curve. With AI, that curve is far steeper than people are capable of absorbing. This becomes a problem for anyone building products using the new technology. They’re taking random shots because customers don’t know what they want or need yet because they haven’t deployed the products. Even if deployed, they don’t fully understand them yet.
So not only are product OEMs in the dark as to customer needs, but ingredient companies who depend on the OEMs for product guidance are even more out of the loop because the information they need from the sources they typically use, end users, isn’t mature or even available.
I had a similar problem when I was in Market and Business analysis at IBM. I solved it by going to installers who were, at the time, the only people who had enough knowledge about what was going on at a scale that I could use to determine what IBM needed to do. I got an award for that, and AMD is taking the same path with a far larger budget by buying Silo AI earlier and ZT Systems recently.
These are unusually strategic moves. I say unusual because, typically, due to an overabundance of financial types on boards, companies are over focused on tactical decisions like doing layoffs and other cost-cutting efforts instead of building for the future. And AI is and will be incredibly disruptive, so companies that aren’t operating strategically are unlikely to survive this AI wave.
AMD, with both of these acquisitions, clearly intends to not only be a survivor but to eventually challenge for leadership.
ZT Systems and Silo AI
Both ZT Systems and Silo AI provide AMD with critical information on where they need to focus. Silo AI does AI implementations and has one of the largest available pools of AI experts who know and understand what AI can and can’t do. Its AI efforts have been unusually successful, leading it to become one of the most sought-after AI deployment partners in the world.
ZT Systems builds AI hardware and stands out as being unusually successful with unique hardware solutions that larger vendors haven’t yet been able to successfully match. This would seem to create a competitive problem with AMD’s OEMs, but AMD plans to eventually divest the manufacturing group while retaining the R&D and intellectual property that this company has created, and then use this to help its OEMs become more successful rather than become a competitor to them.
AMD could still retain manufacturing and use it as NVIDIA sometimes does internally to build products the OEMs don’t want to build but that AMD thinks are critical to the advancement of AI development. NVIDIA has done this a number of times, often passing hardware ownership to OEMs later on when the need or market size becomes more obvious to the OEM. Regardless, AMD plans to retain enough of this capability to create concepts that can be used to convince the OEMs to build better, more performant products similar to, though potentially at a larger scale than, NVIDIA now does.
So, Silo AI will give AMD insight into solutions, and ZT Systems insight into AI-focused hardware that is better optimized as a system. This should put AMD in the lead in terms of understanding what the market should want in an AI solution, and once this information makes it into AMD product design, likely in the next few months, roadmaps will change to better address these newly discovered unmet needs.
This isn’t a path to AI competitiveness. Instead, it’s a path to AI leadership. While nothing is ever certain in a play like this, given it follows a process I personally found to be very successful, I believe the result will be the emergence of a far more competitive AI-focused AMD by the end of 2025.
Wrapping Up:
Companies often complain in a new market that they are just making guesses about what to build. Initially, short of more advanced simulations than we currently have available, there’s no way to close this information gap. But knowledge does get built quickly in firms like Silo AI, which is successfully deploying solutions, and in ZT Systems, which has discovered what companies want in hardware. AMD’s acquisition of both should put it in one of the strongest AI positions once these mergers are final and the information makes it through the design process.
The only thing these companies will need to watch out for is that the acquiring company sometimes fails to listen to the acquired company. But given Lisa Su’s IBM training, which was similar to my own, I’m betting she won’t make that mistake.
This was incredibly impressive work!
