Until now, the AI race has mostly been about scale. Companies and governments have poured billions into larger models, more computing power, and increasingly sophisticated AI systems. But as AI moves from experimentation into enterprise operations, the competitive equation is changing.

For enterprises, that question is getting more practical: where should a workload run, what should it cost, and how much computing power does it really need?

MIT Sloan Management Review Middle East put those questions to Intel’s Ahmad Alabdul-jabbar, General Manager of Intel Saudi Arabia; Islam Kandiel, Telco Key Account Director; Andrew Mironov, Technical Lead for AI Datacenter and Edge, and Kareem Sherif, Director of AI and Datacenter Business Development in a discussion about the infrastructure choices facing enterprises as AI moves into operations.

Inside the conversation, you’ll discover why:

• The most powerful model or hardware may not always be the right fit for a specific AI workload.
• Enterprises are weighing cloud, data center, edge, and client environments based on different business and technical requirements.
• Cost, latency, security, and data sovereignty are becoming central to decisions about where AI runs.
• Building flexibility into the AI infrastructure stack could become increasingly important as technologies and workloads evolve.

Watch the full conversation for an in-depth discussion on workload placement, infrastructure costs, and the trade-offs enterprises face when scaling AI.