Technology Roadmap - Where are You Going with an Industry Led by the AI?

Summary High-performance computing has historically been central to the geophysical industry, with companies like Viridien (formerly CGG) once influencing the global supercomputing landscape. In the 1990s, geophysics represented a significant portion of the HPC market, attracting major vendors such as Cray and Fujitsu. However, since the 2010s, the rise of cloud technologies, social media, and hyperscalers has reshaped the industry, reducing the influence of oil and gas on HPC investments. By 2022, hyperscalers held 37% of global data center capacity, surpassing traditional enterprise infrastructure. The emergence of artificial intelligence has further shifted focus away from geophysics in the eyes of hardware vendors. This presentation outlines the computational landscape of the geophysical industry, beginning with an overview of its specific hardware needs. It continues with a review of CPU and GPU development, emphasizing energy consumption as a growing constraint. Finally, it explores the use of specialized AI hardware—like Nvidia’s Tensor Cores—in seismic applications, examining trade-offs in numerical precision. While AI often favors low-precision formats for speed, geophysical computations continue to rely on FP32 to ensure accuracy in tasks like seismic migration and full-waveform modeling.