3D printing is predicted to be one of the fastest-growing technologies this year. Credit: Stenko Vladislav Vitalievich/iStock via Getty
Machine learning, blockchain databases and 3D printing will be some of the fastest-growing technologies this year. That’s the prediction of the artificial-intelligence-powered 2026 Momentum 100 list of rapidly emerging technology, released by Australian researchers in December 2025. Soft robotics, augmented reality and ’omics — large‑scale, data‑driven studies of biological molecules such as DNA, proteins and metabolites — were also among the top-ranked technologies.
The inaugural Momentum 100 list was derived from an open-access data set called Cosmos 1.0, published in the journal Scientific Data1. To produce the data set, the study authors used a large language model (LLM) to extract information from thousands of Wikipedia pages about science and technology (see ‘Momentum leaders in the Nature Index’). The LLM then clustered the content on the basis of written and hyperlinked connections to create a map of the emerging-technology landscape.
The data set, which is available on the repositories Figshare and Github, can be filtered in various ways, adjusting indices such as a technology’s age and its Wikipedia pageview trends over time. (Figshare is owned by Digital Science, a firm operated by the Holtzbrinck Publishing Group, which has a share in Nature Index’s publisher, Springer Nature.)
“The Momentum 100 concept was to use the indices to filter emerging technologies and provide an interesting lens that we could repeat on an annual basis,” says Paul McCarthy, co-founder of the data-science consultancy League of Scholars in Sydney, Australia, which led the analysis.
The top emerging technology was a type of machine learning called reinforcement learning, in which a system learns by trial and error (see ‘Leaders in the Momentum 100’). This is a versatile mode of AI because it enables systems to make sequential decisions in complex, changing environments. It has so far been used in areas such as drug design2, drone racing3 and game-playing4, with AI models based on reinforcement learning beating top human players in chess, the strategy game Go, and shogi, or Japanese chess.
Blockchain technology came a close second in the Momentum 100 list. The topic has captured broad research interest beyond its cryptocurrency origins, as highlighted by several highly cited publications in the Nature Index between 2015 and 2025. The papers include one on swarm learning — a method that lets hospitals and laboratories collaboratively train AI on medical data without sharing patient information — which has been cited more than 800 times5. Blockchain technology has been applied to ensure the integrity of the data in food supply chains, clinical-trial records and renewable-energy generation.
Leveraging language models
To build the Cosmos 1.0 data set, McCarthy and his team used a pretrained language model called Wikipedia2Vec, which can turn Wikipedia articles into multidimensional number strings called embeddings, enabling researchers to do a geometric analysis of how closely connected certain Wikipedia articles are. “Embeddings encapsulate the contextual meaning of the article based on the words and sentences, but importantly, also the links they contain,” McCarthy says.
From a single Wikipedia ‘seed’ article titled ‘List of emerging technologies’, the team collated a list of almost 55,000 Wikipedia pages that were most similar to the seed article, then filtered these by article type to produce a list of more than 23,000 technologies and technology-adjacent concepts. “We then developed a range of indices around those technologies to characterize them in different ways,” McCarthy says.
The output was cross-validated by comparing it with other data sets, including academic publication and patent data.
Annual lists of emerging technologies are more than just fun talking points, says Daniele Rotolo, who studies emerging technologies at the University of Sussex near Brighton, UK. Such technologies have the potential to destroy existing industries and create new ones, he says, which makes them an important area for policymakers and industry leaders to keep an eye on.
“Worldwide, there is a great interest from governments and companies in these technologies because there is a strong expectation that they could change the status quo,” says Rotolo. “Especially with current geopolitical tensions, the interest in these technologies has become even more important because governments want to invest at the right time and develop a competitive advantage.”
Emerging technologies can also create ripples through the research sector, representing areas in which governments might increase research support through public funding, Rotolo adds.
Despite the imperative of identifying these technologies, forecasting their future effects is difficult, says Rotolo. “The problem is that at this early stage, we have few real traces in the traditional data” that can be used to determine whether they are poised for real-world impact, he says.
AI versus the experts
Because of this challenge, most annual emerging-technologies lists — including those created by the World Economic Forum (WEF), Stanford University and MIT Technology Review — are based heavily on expert opinion rather than pure data analysis.
McCarthy says that this is the point of difference for the Momentum 100 list — it’s based on trends in data, and does not incorporate expert opinion. “Our work was motivated by the idea of mapping technology from a more granular, bottom-up approach, using AI’s ability to reveal latent knowledge in large, complex systems,” he says. “We think there are some inherent benefits to this idea, compared with asking a small group of experts.”
According to Catherine Aiken, who leads the data-science team at Georgetown University’s Center for Security and Emerging Technology in Washington DC, the approach of McCarthy and his team streamlines what can be a slow and laborious process.
“In the six years that I’ve been dialled into this work, the approaches researchers use have been a bit stagnant,” she says. “How we pick an emerging technology and understand its context and potential implications is very expert-driven, very manual, very case-by-case.”
But in the past year or two, there’s been growing interest in using LLMs to extract more insights more quickly, Aiken says. “The Cosmos 1.0 approach is a useful addition to the field. Anything that tries something different to the expert-led, manual process is a useful contribution, because we’re trying to figure out where LLMs can plug in here.”
