Uploaded February 2025 | Updated September 2026, 1 week ago
One could say that Artificial Intelligence is the biggest thing to happen in our networks since the advent of the World Wide Web. In this tutorial I would like to explain the various jargon that goes with this topic, leading to the reason why so much processing power and electrical power is required for today's AI.
With hyperscalers and new AI entrants scrambling to create Gigawatt scale facilities - where will the power come from? And what are they planning to do to ensure that reliable, clean energy will feed the AI monster?
In this session I will try to address the following...
- The taxonomy of AI, ML and neural networks.
- Where is AI in the classic Gartner Hype Cycle?
- How does AI work? Predicting the next word with LLMs
- Why are GPUs the preferred approach?
- Why does Nvidia dominate the market?
- Who else is developing AI chips, and why do they feel the need?
- What’s the different between Training and Inference?
- How do we make AI smarter?
- New proposals for AI
o Non tensor models
o Spiking neural networks
o Neuromorphic processors
o Agentic AI
- But in the meantime…Why does Training use so much power?
- Countries who are starting to struggle…
o The Singapore Experience
o Ireland coming unstuck
o The Memphis Bell End
- What are the hyperscalers doing to get enough power?
o Amazon
o Google
o Meta
o Microsoft
o Oracle
o xAI
- The challenge with renewables
- The New Dr Strangelove – or “How I stopped worrying and learned to love nuclear power”
Speaker Geoff Bennett
One could say that Artificial Intelligence is the biggest thing to happen in our networks since the advent of the World Wide Web. In this tutorial I would like to explain the various jargon that goes with this topic, leading to the reason why so much processing power and electrical power is required for today's AI.
With hyperscalers and new AI entrants scrambling to create Gigawatt scale facilities - where will the power come from? And what are they planning to do to ensure that reliable, clean energy will feed the AI monster?
In this session I will try to address the following...
- The taxonomy of AI, ML and neural networks.
- Where is AI in the classic Gartner Hype Cycle?
- How does AI work? Predicting the next word with LLMs
- Why are GPUs the preferred approach?
- Why does Nvidia dominate the market?
- Who else is developing AI chips, and why do they feel the need?
- What’s the different between Training and Inference?
- How do we make AI smarter?
- New proposals for AI
o Non tensor models
o Spiking neural networks
o Neuromorphic processors
o Agentic AI
- But in the meantime…Why does Training use so much power?
- Countries who are starting to struggle…
o The Singapore Experience
o Ireland coming unstuck
o The Memphis Bell End
- What are the hyperscalers doing to get enough power?
o Amazon
o Google
o Meta
o Microsoft
o Oracle
o xAI
- The challenge with renewables
- The New Dr Strangelove – or “How I stopped worrying and learned to love nuclear power”
Speaker Geoff Bennett
