Building Sustainable AI For Supply Chain

Nvidia is investing $100 billion in OpenAI to build data centers, and this is just a small portion of the capital being poured into AI infrastructure. But will it be a well-balanced investment that takes both companies into the lead position for a foundational LLM?

AI companies are investing trillions of dollars to fund their computing power and technology, but projected revenue is likely to fall below what is needed to fund this level of growth over the next three to five years. Add a forecasted power requirement of 200GW (over double the annual power used by France), and we have an unstable recipe for a sustainable business model.

Supply chain leaders must consider these factors as CEOs push them harder than ever to adopt all types of AI solutions, from an ever-growing number of technology vendors.

CSCOs will likely see an increase in their cost allocations from IT. Basic access to an LLM such as ChatGPT for enterprise runs about $40 to $60 per month per person. Once considered a nice-to-have, LLM access is now a standard productivity tool for most professionals. A small supply chain organization could spend $150,000 to $200,000 annually on this tool alone. This is just the tip of a broader deployment of AI-based tools that will move well beyond personal time savings and quality improvements.

Potential revenue shortfalls for AI providers will cause roll-out delays, industry consolidation and price increases. Choosing the most sustainable AI platforms at this stage is being overlooked in a rush to adopt the tech. We have been through this before with other tech platforms. Let's avoid the same mistakes this time.

Supply chain problems are growing more complex, and use cases are almost limitless.

The better defined a problem is, the easier it is to solve. We should shift from finding a place to use the tech to applying the right level of tech to solve well-defined problems.

Return on investment will look much the same in an AI-driven supply chain as it does today. But there is a twist. The return period is under pressure, with many expecting significant returns within 12 to 18 months.

AI and machine learning have been used in supply chain for many years. Applications have been standalone or baked into larger platforms straight from the vendors. This trend looks to continue, but it won’t be a one-size-fits-all AI model.

Large enterprise applications are adding AI functionality to their existing platforms. This includes agentic AI and GenAI, in addition to machine learning and algorithms already embedded. These capabilities will impact large processes in supply chain, but they won’t address all the needs.

For example, GenAI is being used to help code and design robotic process automation (RPA), clean data, build data warehouses, mine and optimize processes. These won’t all come from the same vendors or at the same price tag. Be selective and frugal.

Companies state AI will reduce employment. This elephant in the room must be addressed.

A recent MIT study authored by Pierre Bouquet and Yossi Sheffi finds that AI could impact 1.1 million full-time transportation employees and augment or automate $65 billion worth of tasks.

However, employment numbers at the macro level do not show this yet. In the Eurozone and the United States, employment levels are stable and either at, or approaching, full employment. These numbers do not reveal the demographic cracks that researchers are seeing.

Historic norms of bringing talent in from universities and letting them work up through a career path will need to be reengineered as AI will assume many entry-level roles. Companies must avoid the risk of having stagnant and insufficient talent pools as a result.

At this stage of the journey, adoption should be the highest priority. This will be one of the most difficult challenges facing CIOs and CSCOs. We are all looking over our shoulder for the agent coming for our job, which doesn’t encourage trust.

Companies making quick headway are developing robust governance structures and being very transparent about their plans to use AI. A new trust framework must be built that includes what roles and activities could and will be replaced by AI.
 

Wade L. McDaniel
VP Distinguished Advisor
Gartner Supply Chain
Wade.Mcdaniel@gartner.com