This is an unedited manuscript accepted for publication and provided as an Article in Press for early access at the author’s request. The article will undergo copyediting, typesetting, and galley proof review before final publication. Please be aware that errors may be identified during production that could affect the content. All legal disclaimers of the journal apply.
Naved Ahmed,
Mona Quaisar,
MD. Maqsood Alam,
Sudhisth Kumar,
- Student, Electronics and Communication Engineering Department, Al-Kabir Institute of Management & Technology, Jharkhand, India
- Professor, Department of Electronics and Communication Engineering (ECE),Al-Kabir Institute of Management & Technology, Jharkhand, India
- Professor, Department of Electronics and Communication Engineering (ECE),Al-Kabir Institute of Management & Technology, Jharkhand, India
- Professor, Department of Electronics and Communication Engineering (ECE),Al-Kabir Institute of Management & Technology, Jharkhand, India
Abstract
Artificial Intelligence (AI) is not only about algorithms. AI works like a full “stack” of layers, from electricity to real-world user applications. In this paper, we explain a simple and student-friendly Five- Layer Architecture of AI: (1) Energy, (2) Chips, (3) Infrastructure, (4) Models, and (5) Applications. Each layer supports the next layer, like a cake with multiple layers. If any layer is weak, AI systems become slow, costly, or unreliable. We describe why energy availability and carbon impact matter for training and running AI models, how specialized chips like GPUs/TPUs improve performance, and how cloud data centers, networking, and storage form the infrastructure backbone. We also summarize modern model development ideas such as transformer architectures, scaling laws, efficient training, and safe deployment. Finally, we connect these technical layers to practical applications in healthcare, education, agriculture, finance, and public services. We propose a methodology that students and researchers can use to study the AI stack in any country: define indicators for each layer, collect datasets from trusted sources, normalize and compare across countries, and interpret gaps and opportunities. A small cross-country comparison is presented using common infrastructure indicators to show how differences in energy, connectivity, and compute readiness can affect AI growth.
Keywords: Artificial Intelligence, AI stack, compute, energy efficiency, GPUs, data centers, infrastructure, transformers, scaling laws, AI applications
References
[1] E. Strubell, A. Ganesh, and A. McCallum, “Energy and policy considerations for deep learning in
NLP,” in Proc. ACL, 2019.
[2] D. Patterson et al., “Carbon emissions and large neural network training,” arXiv:2104.10350, 2021.
1
[3] A. Vaswani et al., “Attention is all you need,” in Proc. NeurIPS, 2017.
[4] J. Devlin, M.-W. Chang, K. Lee, and K. Toutanova, “BERT: Pre-training of deep bidirectional
transformers for language understanding,” in Proc. NAACL-HLT, 2019.
[5] T. B. Brown et al., “Language models are few-shot learners,” in Proc. NeurIPS, 2020.
[6] J. Kaplan et al., “Scaling laws for neural language models,” arXiv:2001.08361, 2020.
[7] J. Hoffmann et al., “Training compute-optimal large language models,” arXiv:2203.15556, 2022.
[8] P. Mattson et al., “MLPerf training benchmark,” in Proc. MLSys, 2020.
[9] A. Jobin, M. Ienca, and E. Vayena, “The global landscape of AI ethics guidelines,” Nature Machine
Intelligence, vol. 1, 2019.
[10] National Institute of Standards and Technology (NIST), “AI Risk Management Framework (AI RMF
1.0),” 2023.
[11] Oxford Insights, “Government AI Readiness Index,” 2023.
[12] Tortoise Media, “Global AI Index,” 2023.
[13] International Energy Agency (IEA), “Data centres and data transmission networks,” 2024.
[14] World Bank, “Access to electricity (% of population),” World Development Indicators, accessed
2026.
[15] International Telecommunication Union (ITU), “Fixed-broadband subscriptions,” ICT statistics,
accessed 2026.
[16] J. Wei et al., “Chain-of-thought prompting elicits reasoning in large language models,”
arXiv:2201.11903, 2022.
[17] R. Bommasani et al., “On the opportunities and risks of foundation models,” arXiv:2108.07258,
2021.
[18] OpenAI, “GPT-4 technical report,” arXiv:2303.08774, 2023.
[19] A. Chowdhery et al., “PaLM: Scaling language modeling with pathways,” arXiv:2204.02311, 2022.
[20] A. Radford et al., “Language models are unsupervised multitask learners,” OpenAI Technical Report,
2019.
[21] D. Amodei et al., “Concrete problems in AI safety,” arXiv:1606.06565, 2016.
[22] M. Mitchell et al., “Model cards for model reporting,” in Proc. FAT*, 2019.
[23] T. Gebru et al., “Datasheets for datasets,” Communications of the ACM, vol. 64, no. 12, 2021.
[24] A. Dosovitskiy et al., “An image is worth 16×16 words: Transformers for image recognition at scale,”
in Proc. ICLR, 2021.
Journal of Artificial Intelligence Research & Advances
| Volume | 13 | |
| 03 | ||
| Received | 10/06/2026 | |
| Accepted | 19/08/2026 | |
| Published | 30/09/2026 | |
| Publication Time | 112 Days |