OpenAI has signed a multi-year deal worth over $10 billion with AI chipmaker Cerebras to secure 750 megawatts of computing power through 2028 (Gupta & Ranjan, 2024). This landmark agreement highlights a critical change in the role of AI compute: it is no longer a technical support, but a strategic resource that shapes the future of AI development. As AI models continue to scale and are increasingly demanded for real-time performance, global competition in AI is extensively intense and is rapidly shifting away from simple algorithms toward access to compute infrastructure and advances in chip innovation.
What is Compute? Why Compute Matters in AI Generation?
Compute, or computational power, is the foundation in the development of artificial intelligence (AI)-based solutions at scale, enabling AI models to process vast amounts of data, perform complex calculations, and make intelligent decisions by continuously training their models and learning underlying patterns and logic (Gupta & Ranjan, 2024). Therefore, compute becomes particularly important when it comes to the current generation of AI, where real-time responses are essential for AI applications such as ChatGPT and Capilot, directly impacting user experience and the scalability of AI products. In this sense, compute functions not only as essential infrastructure but also as a strategic competitive barrier in the AI landscape.
From GPUs to Specialized Chips: The Next Step in AI Compute
According to Flinders and Susnjara (2024), Graphics Processing Units (GPUs), is an electronic circuit developed by NVIDIA in the 1990s and designed to accelerate computer image processing on various devices such as video cards, system boards, mobile phones and personal computers. It has long dominated AI compute due to their high parallel processing capabilities. However, the launch of ChatGPT in 2022 and the following era of the AI boom created unprecedented demand for faster and more efficient systems. This gap allows companies like Cerebras, which design specialized AI chips that are faster than GPU-based systems in task optimization such as real-time inference, to rise rapidly (Gupta & Ranjan, 2024)
Looking Ahead
Change is constant. As AI continues to contribute to socio-economic progress, the global semiconductor industry is expected to expand in response to rising compute demand. Looking ahead, AI infrastructure investment is likely to increase, accompanied by a diversification of compute solutions beyond traditional GPUs. According to Humphrey (2025), future chip design should place more emphasis on improving density, performance, and production efficiency should be placed, while new materials and automation further extend AI capabilities. Ultimately, the future of AI will depend not only on how intelligent models become, but on the strength and innovation of the chips and compute powering them. As compute and advanced chip capabilities become increasingly concentrated, competing at the infrastructure or foundational model level is no longer a capital-efficient path for most startups. Venture-scale opportunities are more likely to arise in specific applications and industries, where success depends on unique data, specialized knowledge, and effective business strategies instead of just computing power
Citations
Chakraborty, S., Jha, G., & Paliwal, S. (2025, September 8). Global AI Race. BCG
Global. https://www.bcg.com/publications/2025/india-global-ai-race
Dorge, R. (2025, December 16). Global Semiconductor Chip Market 2024–2033. Custom
Market Insights. https://www.custommarketinsights.com/report/semiconductor-chip-market/
Flinders, M., & Susnjara, S. (2024, February 26). What is a graphics processing unit
(GPU)? (I. Smalley, Ed.). Ibm.com. https://www.ibm.com/think/topics/gpu
Gupta, A., & Ranjan, A. (2024, April 30). A Primer on Compute. Carnegie Endowment
for International Peace. https://carnegieendowment.org/posts/2024/04/a-primer-on-compute?lang=en
Humphrey, S. (2025, October 24). Semiconductor Industry Trends and the Future of
Manufacturing. PTC. https://www.ptc.com/en/blogs/electronics-high-tech/semiconductor-industry-trends-and-challenges
Ropek, L. (2026, January 14). OpenAI signs deal, worth $10B, for compute from Cerebras | TechCrunch. TechCrunch. https://techcrunch.com/2026/01/14/openai-signs-deal-reportedly-worth-10-billion-for-compute-from-cerebras
