#

Key takeaways

• Large AI data centers depend on stable power as much as on chips and software.

• Grid shortages can force operators to plan for curtailment and compensation.

• For Moroccan readers, energy sourcing and backup power should be part of AI planning from the start.

• Data availability, procurement, language mix, and compliance can shape how AI systems are deployed.

• Governance matters because outages, privacy, and cybersecurity risks can affect service quality.

What happened

TechCrunch reported that PJM Interconnection plans to temporarily cut electricity supply to data centers and other large users during shortages. The report says the measure would start in June 2027 and would apply to data centers of 50 megawatts or larger. It also says affected customers would be compensated.

This is a useful signal for anyone planning AI infrastructure. A modern AI stack is not only a software project. It is also an energy project, a procurement project, and a resilience project. For Moroccan readers, that matters because the same logic applies wherever large compute loads depend on reliable power.

Why this matters for AI infrastructure

AI systems need steady electricity for servers, cooling, networking, and storage. When power becomes constrained, operators may need to reduce load, shift workloads, or rely on backup systems. That can affect uptime, costs, and service quality.

The PJM case shows that grid operators may treat large data centers as flexible demand during shortages. That does not mean every market will do the same. But it does mean planners should assume power constraints can become part of the operating model.

For Morocco, the lesson is practical. Any organization building AI services should ask how much power the system needs, what happens during shortages, and how long backup systems can carry the load. Those questions are basic, but they are often postponed until late in the project.

Morocco context: planning for power, not just compute

Moroccan companies and public institutions that want to use AI may focus first on models, vendors, and use cases. That is understandable. But infrastructure decisions can be just as important as the software choice.

A realistic AI plan in Morocco would need to consider energy sourcing, backup power tradeoffs, and the cost of keeping systems available. It would also need to consider procurement timelines. Large infrastructure purchases often move slowly, and that can make it harder to react when demand grows faster than expected.

Language mix is another issue. Many Moroccan deployments need to handle Arabic, French, and sometimes other languages. That increases data preparation work and can raise storage and processing needs. If the infrastructure is already tight, multilingual workloads can make planning more difficult.

Use cases in Morocco

The PJM report is about grid management, but the broader lesson applies to many Moroccan use cases.

Public services

Government teams using AI for document processing, citizen support, or internal search would need reliable uptime. If systems go down during peak demand, service quality drops quickly. Backup power and failover planning would be part of responsible deployment.

Banks and insurers

Financial institutions often need strong availability, logging, and security controls. AI tools used for fraud detection, customer support, or document review may require continuous access to data and compute. Power interruptions can create operational risk and compliance pressure.

Industry and logistics

Factories, ports, and logistics operators may use AI for forecasting, maintenance, or routing. These systems can be sensitive to delays in data flow. If compute is interrupted, decisions may become stale. That can affect efficiency and planning.

Startups and service providers

Smaller Moroccan startups may not run their own large data centers. They may rely on cloud services or colocated infrastructure. Even then, they still face energy and resilience questions. They need to know what service levels they are buying and what happens if capacity is constrained.

Risks and governance

The biggest risk in this story is not only a blackout. It is poor planning. If an AI system depends on uninterrupted power, then the organization must design for failure from the start.

That means thinking about data availability, cybersecurity, and privacy together. A backup system that restores service but exposes data is not a good solution. A resilient system should protect access, logs, and sensitive information at the same time.

Compliance also matters. Moroccan organizations would need to align AI operations with their own internal policies and any applicable legal or contractual obligations. The source article does not discuss Morocco-specific rules, so this should be treated as a general governance point, not a legal claim.

There is also a procurement risk. Teams sometimes buy compute first and ask about resilience later. That can lead to expensive retrofits. It is better to include power continuity, cooling, and failover requirements in the initial request for proposals.

What Moroccan policymakers and operators should do next

The PJM report suggests a simple checklist for Moroccan decision-makers. It is not a prediction about Morocco. It is a planning framework that could reduce surprises.

1. *Map power needs early.

• Estimate the load for training, inference, storage, and cooling.

2. *Plan for shortages.

• Define what happens if power is limited for hours or days.

3. *Test backup options.

• Check whether generators, batteries, or cloud failover can keep critical services running.

4. *Review data dependencies.

• Identify which datasets must stay available and which can be delayed.

5. *Build for multilingual workloads.

• Make sure Arabic and French processing are included in capacity planning.

6. *Align security and compliance.

• Treat resilience, privacy, and cybersecurity as one design problem.

For Moroccan readers, the main lesson is that AI infrastructure is becoming more like utility infrastructure. It needs the same discipline. That includes budgeting for energy, planning for outages, and setting clear service priorities.

Bottom line

PJM's planned curtailment policy is a reminder that large AI systems depend on more than software performance. They depend on the grid, on backup power, and on operational discipline.

For Morocco, the practical takeaway is to plan AI projects with energy resilience in mind. Teams that do this early may avoid costly disruptions later. Those that do not may discover that the hardest part of AI is not the model. It is keeping the system on.