The Scramble for Enterprise AI Spend

AI has been uniquely successful in "Crossing the Chasm" of the Technology Adoption Life Cycle (from the "Early Adopters" to the "Early Majority"); it has done so by convincing the Enterprise decision makers of its immediate value.

What's needed to tap into this opportunity?

First, why now?

Even though IBM and Watson have been around for a while, it's the Natural Language Processing (NLP) based applications such as ChatGPT (and the likes), placed in the hands of the enterprise decision makers, that has convinced them of its huge productivity potential. In Product Management speak, it has already demostrated its proof of value, how AI can boost productivity.

ChatGPT and Prompt-Engineering has gamified AI for them. Even the sporadic negative publicity of Hallucinations has been neutral if not helpful (i.e. bad things happen to others, the monkey weilding a sword story).

So what?

Enterprises understand the value of data and now have a way to extract it without they themselves being proficient in advanced data engineering. This creates fertile ground for the Edge ecosystem to evolve quickly. Because AI is that elusive Killer-App for Edge Computing.

Saying once again for re-emphasis:

AI is the Killer-App for Edge Computing.         

So what's the hurdle then?

It's all happening too quickly for the ecosystem to keep pace with and come up with the best architecture. The largest business segment that's looking for Edge AI is Industrial IoT and Operational Tech (OT)+Edge infrastructure. Other segments such as retail, cities and transport, health and education, also have common needs.

1. Problem number 1: How can Application developers show a proof of value to its User Buyers without becoming an expert in system and infrastructure programming. Incidentally, for the coming 4 years, this market is almost twice the size of the next two segments in this space combined.
2. Problem number 2: How can hardware manufacturers (ODMs) and segment specific system integrators (SIs) map commercial off the shelf (COTS) devices into AI appliances. Why it is important, because until an Edge Cloud emerges, it looks like AI for Enterprises is following an Appliance model - it's a CTO vs CIO vs Operations/COO conversation (let's keep it for another day). This Appliance model however needs to be moduar and mass customizable - to keep prices in tight control.
3. Problem number 3: There are at least 4 types of computes feeding into this Edge ecosystem, which one to prioritize.

• NVDIA HW: This is the most mature platform but it can be too costly for the Appliance model.
• The x86 compute: A combined CPU/GPU/NPU with Linux as a common platform is a good fit for a broad range of use-cases. This will likely be a dominant ecosystem in the near future.
• The ARM based hardware: This is an interesting emerging space with Qualcomm and also ARM themselves jumping in. It is particularly attractive for certain market segments where ARM already plays significant role. And, once Apple sees (if not already seen) the User Buyer market segment I am convinced they will throw all their weight behind this, purely because of it's strong fit to their DNA.
• RISC-V Hardware: This is the dark horse in this race. There are several startups in this area and they are prioritizing the right end of the business segment but it'd take some time for the ecosystem to emerge from the hardware to platform to product journey. It's a space to keep an eye on, similarly for the FPGA and other ASICs ecosystems.

What's the takeaway?

The Edge Ecosystem needs an Open Platform to match the pace of AI - this short article gives an outline of the product market fit of such an Open Platform:

• Prioritize User Buyer persona
• Free-to-try, Easy-to-buy
• One platform, multiple Use-cases
• One platform, multiple system binaries and libraries,
• One platform, multiple Hardware options

The Open Platform for Data Science and Analytics at the Edge must be built using:
- Abstraction,
- Encapsulation, and 
- Modularity        

I'll stop here, keep the key use cases emerging in this space from IoT, Industrial and Utilities, Retail, Smart Cities and Transport, and Education Health and Highstreet Banking for some other day.

Please let me know if I missed something - or should develop certain aspects further in a subsequent article.

Fir milenge .. Till we meet again