2024 may potentially be the year that genAI hype peaked. But the momentum around the subject continues to be intense. In a query of Elsevier’s Scopus database, generative AI had a CAGR of 313.3% from 2019 when there were four publications on the subject. As of November 27, there were 4,821 papers featuring the subject in 2024. Large Language Models had even higher absolute numbers — scaling from 72 publications to 15,522 in the same period with a 215.9% CAGR.

The genAI hype bubble may have popped in 2024

For context, there were several signs suggesting that GenAI has moved beyond its peak hype phase. Gartner’s “Hype Cycle for Artificial Intelligence, 2024” places GenAI in the “trough of disillusionment.”  The technology has “yet to deliver on its anticipated business value for most organizations,” the firm wrote. Other analysts have expressed similar views. In June, Goldman Sachs asked: “Gen AI: too much spend, too little benefit?”  A Substack named “AI Snake Oil” has also gained more than 43,000 subscribers.  Other signs of slowing growth include increasing reports that genAI model scaling is plateauing, although that argument remains contentious for some in the industry — such as Anthropic CEO Dario Amodei.

AI remains a strong patent area

Meanwhile, a number of companies are racking up AI patents. As of October 2024, Intel Corporation had been granted 1,499 total AI patents since the beginning of the year across 1,389 unique patent families. Intel’s portfolio shows significant focus on machine learning applications, with 14.81% of patents in ML-specific technologies.

Samsung Electronics followed with 1,002 patents (732 unique families), showing a focus on neural networks with 8.88% of their portfolio dedicated to that segment. Recent Samsung patents demonstrate practical AI applications. Examples include accent-agnostic wake word detection and personalized ASR model development.

In the same time period, Alphabet Inc. showed the highest concentration in ML patents at 50.14% of its portfolio while its subsidiary DeepMind Technologies had a strong neural network focus, with 55.88% of its patents in this domain. IBM exhibited the highest ML patent concentration among major firms at 86.79% of its AI portfolio.

It’s not just tech companies racking up AI patents. Financial service firms, too, are entering the fray of top AI patentees.

Chinese technology companies also show significant patent activity, with Tencent Holdings (819 patents), Huawei Technologies (727 patents), and Baidu (686 patents) all ranking among top patent holders. These companies demonstrate high originality ratios (0.95, 0.75, and 0.99 respectively).

The publication trajectories of different AI technologies in 2024 reveal a telling maturity curve. While generative AI and LLMs show explosive growth (313.3% and 215.9% CAGR respectively), more established technologies like Machine Learning maintain steady but slower expansion (21.6% CAGR) with significantly higher absolute numbers – 140,574 publications in 2024. Edge Computing (7,283 publications) and Neural Networks (126,803 publications) show signs of plateauing or slight decline from 2023 peaks, suggesting technology maturation. Meanwhile, emerging technologies like Federated Learning demonstrate sustained growth (138.5% CAGR) without the dramatic spikes seen in generative AI, indicating more organic research progression. This layered growth pattern, combined with the focused patent activity in specific domains (Intel’s 14.81% ML concentration versus IBM’s 86.79%), suggests the AI landscape is fragmenting into distinct maturity phases rather than following a single hype cycle.

Moving beyond GenAI

Gartner recently suggested that the AI market was moving beyond genAI and recommended that organizations considering AI investments widen their gaze to include other subjects such as composite AI approaches, AI engineering, and knowledge graphs. The research firm’s assessment aligns with publication patterns found in in Elsevier’s Scopus database. Established technologies like Machine Learning continue to maintain steady expansion at 21.6% CAGR with significantly higher absolute numbers — 140,574 publications in 2024.

Composite AI, which Gartner identifies as representing the next phase in AI evolution, fuses multiple methodologies like machine learning, natural language processing, and knowledge graphs. This integrated approach is mirrored in the diverse patent strategies of many top AI patentees. For instance, Alphabet (and its subsidiary Google DeepMind), Capital One, IBM and Samsung demonstrate distinct multi-methodology patent portfolios, with each company maintaining significant investments across multiple AI domains. Quantitative analysis reveals Alphabet holds a balanced distribution with 50.14% in machine learning patents, 13.3% in neural networks, and 4.57% in knowledge-based systems. IBM shows similar diversification with 86.79% in ML, 4.64% in neural networks, and 7.86% in knowledge-based approaches. Capital One’s portfolio emphasizes ML applications (114.32%) while maintaining substantial knowledge-based systems presence (13.83%). Meanwhile, Samsung’s strategy focuses on balanced technical coverage with 32.44% ML, 8.88% neural networks, and 1.6% knowledge-based patents. This multi-domain patent acquisition pattern aligns with the composite AI thesis. Organizations are positioning themselves for integrated AI capabilities rather than singular technological approaches.