A few days ago David Vellante posted a summary in LinkedIn about a Breaking Analysis from theCUBE Research about the state of GenAI in the Enterprise.
It is a very interesting reading with lots of data points and reflections about the market behavior.
🎯 Data Points I found interesting
- Enterprise use cases are ChatGPT-like (not surprised).
- AI-monetization doesn't follow the hype (not surprised).
- Complex Use Cases, Longer ROIs and Paybacks, but higher upsides (not surprised)
- Investments in GPUs <10M will wait >1 year to be fulfilled.
- 18% of Respondents are “GenAI Laggards” (are not doing “anything” with GenAI)
- A picture of an “Expanding Market”: the Net Score chart of adoption by Vendor (how many will remain here in 3-5 years? This is an Enterprise Architecture challenge)
- The Cloud vs. On-prem debate: a 3 player game (Hyperscalers, GPU Clouds, On-prem Vendors) with an “old but recurring narrative” about “cost reduction” 🤦🏻♂️
- Beyond Model Development, more is yet to come: superchips, bigger memories, faster interconnects … (and more).
- NVIDIA monopoly could last a decade (I agree), but ARM will win AI on the Edge non-monopolistically (I agree too)
👎🏻 Questions which I disagree with (at least partially)
- Focus on Process and People, not on Technology Architecture…
My take: You can (and should) do the later by focusing on Agnostic Capabilities, Patterns and Building Blocks. This way, you could map Constraints and Business Capabilities and plan accordingly. Additionally, in my humble opinion, you can't address Process and People these days without Technology.
🤓 My own considerations
- Monopolies = “Shoulders of Giants”. This is both a Threat and an Opportunity.
- GPU Investment Size is a potential “qualification criterion” for Solution Design.
- “Why” Enterprises don't feel transformational value here? … (Work in progress) 🤷🏻♂️
- Fixation in “cost-reduction” conversations might mean a mindset not focused on Innovation (redefining existing problems under new terms, exploring areas that were not addressable before, etc.)
- There is more to come beyond superchips: Analog Computing? Quantum Computing? …
- Monopolies that last decades produce “legacy” and “market gravity” for the next one (Journey To “The Next Thing”).
- What about the expected Productivity improvements? What about the “Future of Work”? Maybe on a future study?