The Energy Industry’s Role in the Future of Intelligence

11 June 2024

As AI Advances Rapidly, the Demand for Energy and High-Quality Data Rises

In 2023, the greatest challenge to artificial intelligence (AI) advancement was the availability of specialised chips. The semiconductor industry faced complex supply and demand imbalances, particularly affecting AI accelerator chips. While revenue declined, companies invested in new facilities and technologies to meet future AI demand.[1][2]

Now, in 2024, the focus has shifted to the availability of voltage transformers as AI’s computing capacity continues to expand rapidly. Massive transformer-based language models like GPT-4 are driving the need for more electricity and voltage step-down transformers to power chips – “transformers for transformers.”[3][4] Taiwan Semiconductor Manufacturing Company (TSMC) is expanding advanced chip packaging capacity to meet the growing demand for AI accelerators.[5]

Looking ahead to 2025, experts predict that energy capacity and high-quality training data will become the primary constraints on AI progress. Global electronic compute capacity is increasing by an estimated 10x every 12 months, driven largely by new AI supercomputer complexes. This exponential growth in computing will inevitably drive up energy demand and costs until new sources like nuclear fission can be implemented at scale.[6]

Many AI researchers believe artificial general intelligence (AGI) – AI with human-level abilities across all domains – could be developed by the end of 2025.[7] AGI would be a profound technological breakthrough with immense implications for society. However, the path to AGI is still unclear and will likely require fundamental innovations in algorithms and computing architectures.[8]

As AI marches forward, its increasing compute demands are on a collision course with real-world energy and infrastructure constraints. The energy industry must rapidly evolve and expand to power the future of AI or risk becoming a bottleneck to progress. Simultaneously, the AI field needs to prioritise the efficient use of compute through algorithmic innovation and high-quality data curation.[9]

In the coming years, AI and energy will become inextricably linked – the future of intelligence depends on sustainable and abundant energy. Policymakers, energy providers, and technology companies must collaborate to strategically direct resources to enable the responsible development of ever more powerful AI systems. The transformative potential of advanced AI is immense, but reaching it requires solving significant challenges at the intersection of energy and technology.

References:

[1] The Semiconductor Industry’s Biggest Challenges in 2023 (https://omdia.tech.informa.com/blogs/2023/mar/the-semiconductor-industrys-biggest-challenges-in-2023)

[2] 2023 in Review: AI Takes Center Stage in the Eternal Quest for Innovation (https://www.synopsys.com/blogs/chip-design/semiconductor-industry-trends-2023.html)

[3] CoWoS Capacity Shortage Challenges AI Chip Demand, while Taiwanese Manufacturers Expand to Seize Opportunities (https://www.trendforce.com/news/2024/02/19/insights-cowos-capacity-shortage-challenges-ai-chip-demand-while-taiwanese-manufacturers-expand-to-seize-opportunities/)

[4] Chip hunting: The semiconductor procurement solution when other options fail (https://www.mckinsey.com/industries/semiconductors/our-insights/chip-hunting-the-semiconductor-procurement-solution-when-other-options-fail)

[5] Will superintelligent AI sneak up on us? New study offers reassurance (https://www.nature.com/articles/d41586-023-04094-z)

[6] AI and Compute (https://openai.com/blog/ai-and-compute/)

[7] Five Experts Explain Whether AI Could Ever Become as Intelligent as Humans (https://www.sciencealert.com/five-experts-explain-whether-ai-could-ever-become-as-intelligent-as-humans)

[8] How close are we to AI that surpasses human intelligence? (https://www.brookings.edu/articles/how-close-are-we-to-ai-that-surpasses-human-intelligence/)

[9] Data Governance: The Key to High-Quality Training Data for AI Models (https://innodata.com/data-governance-the-key-to-high-quality-training-data-for-ai-models/)

– Jason Jackson, Product Owner (R&D)

Share this article

View All Articles

Get in touch and see what we can do for you

Request further information to see how our solutions can support your business.

Contact us today

Join the conversation.

Sign up to receive our latest news and product updates.

This website is hosted Green - checked by thegreenwebfoundation.org

© 2024 Flui Technologies is a trading name of AMT-SYBEX Ltd

Gladstone House, Hithercroft Road, Wallingford, England, OX10 9BT

Website by Yello