Global AI Trends

Artificial intelligence (AI) continues to evolve at a rapid pace, influencing a vast range of sectors. The State of AI Report 2024, compiled by analysts at Air Street Capital, consolidates the key trends in research, business, policy, and security. Let’s take a closer look at the current AI landscape and the changes we can expect in the near future.

Until recently, OpenAI reigned unchallenged in the field of large language models (LLMs). However, in 2024, companies like Meta, Google DeepMind, and Anthropic managed to narrow the gap significantly. For instance, Meta’s new Llama 3 series has, for the first time, reached a performance level comparable to closed-source models such as GPT-4.
Models now do more than just generate text—they can learn programming, solve mathematical problems, analyze images, and even process biological data. Leading enterprises are betting on multimodal models that can handle multiple types of information simultaneously: text, images, video, and even molecules.
A central focus is optimization and cost reduction in AI. Organizations are implementing quantization (reducing model size), distillation (transferring knowledge from larger models to smaller ones), and decreasing parameter counts without sacrificing accuracy. These approaches make it possible to run powerful models directly on smartphones.

NVIDIA continues to dominate the AI hardware market. In 2024, its market capitalization reached $3 trillion, and the new Blackwell B200 and GB200 Superchip GPUs promise to reduce model training costs by a factor of 25.
Competitors such as AMD, Intel, and Google are striving to capture market share, but thus far, without significant success. Even giants like OpenAI and Meta still rely on NVIDIA for their hardware needs.
Notably, Chinese AI companies have ramped up efforts to develop their own large language models, despite U.S. sanctions. Chinese labs are successfully creating models that rival their Western counterparts, using fewer computing resources.