Estimated reading time: 5 minutes

Technology shapes our world. Constant innovation has made life today very different from just a generation ago. Now, as we enter the second quarter of the 21st century, which new ideas can we expect to see? Many leading industry analysts, such as Gartner, have already published their forecasts. Here, we summarise their most popular predictions for 2025...
 

Multi-modal artificial intelligence 

AI is the transformational innovation of the age, threatening to impact every vertical. But what advances can we expect to see in the next 12 months? One is multimodal AI, which processes text, images and videos simultaneously. 

The interest in multimodal AI is understandable since the tech mimics human behaviour. People process information via sight, sound, and touch. So, it’s logical that humans should interact with machines using this multimodal approach too. 

Google Gemini and Chat GPT-4 started the ball rolling in 2023. OpenAI even launched its multimodal product with the headline: “ChatGPT can now see, hear, and speak.” But the forthcoming ChatGPT-5 is expected to offer vastly improved multimodal functionality.

Multimodal AI could drive major change across diverse sectors—enhancing education, revolutionising retail with personalised shopping, and enabling advanced technologies like voice and facial recognition in financial services.

Energy efficient computing

Every year, demand for digital services grows. Since 2010, global internet traffic has expanded 20-fold as consumers switch to cloud-based offerings. These services are extremely energy hungry. The data centres that underpin them account for about 1.5 percent of global electricity use. 

So, the energy efficiency race is on. One obvious target is over-heating. Data centres generate a lot of heat, and use elaborate cooling systems. This is a waste of resources. Architects are trying various remedies. For example, new designs isolate heat-generating machines in one area, rather than across the whole site, making them easier to cool. They are also learning how to capture and repurpose heat to warm nearby homes and businesses. 

Chip companies are working on the problem too. In 2024, Nvidia unveiled its next-generation Blackwell graphics processing units (GPUs), which have 25 times better energy consumption and lower costs for AI processing.
 

Post-quantum cryptography 

Quantum computing is mind-boggling tech. It uses the quantum world’s counterintuitive properties — which enable a bit of data to act as both a 0 and 1 at the same time — to make calculations that are impossible on a conventional computer.

On the plus side, this could lead to extraordinary breakthroughs in drug design, weather forecasting, industrial efficiency and more. However, quantum computing does pose a serious security challenge.  ‘Classical’ computing uses complex mathematical problems to keep out attackers. Quantum processors can, in theory, solve these problems and therefore breach security systems. 

To resist this threat, researchers are working on post-quantum cryptography technology. They are developing new algorithms to generate mathematical challenges even quantum computers can’t crack. The techniques include lattice-based cryptography, code-based cryptography and multivariate cryptography. 

Industry organisations are actively exploring these solutions. One is the US Department of Commerce’s National Institute of Standards and Technology (NIST). In August 2024, it finalised its encryption algorithms. This eight-year effort assessed 82 algorithms before settling on 15, which it is now encouraging developers to incorporate into their systems.

New AI-based cybersecurity threats

AI tools are frequently described as ‘co-pilots’ that help people to complete tasks. Regrettably, AI makes fraudsters’ lives easier too. They are now using the tech to devise devious new scams.

One recent study reported an alarming 28 percent increase in phishing emails from Q1 2024 to Q2 2024. It’s clear that AI is fueling much of this activity. The new AI-enhanced threats take many forms. They include: 

•    Phishing emails with perfect grammar and personal details
•    Large scale campaigns that perform thousands of individually targeted phishing attacks simultaneously.
•    Attacks that access information fed by employees into AI platforms like ChatGPT or Google Gemini.
•    AI-enhanced phishing that mimics the behaviour, appearance, and voice of individuals. 
•    Campaigns that exploit social media platforms to steal personal information and manipulate users.
•    Phishing as a service (PhaaS) platforms. They enable non-technical attackers to create templates for fake login pages, scripts to automate email distribution and tools for managing stolen credentials. 

Thermal batteries

Renewable energy’s share of global energy consumption was 13 percent in 2023, and is expected to hit nearly 20 percent by 2030. But there's an obvious problem: unpredictability. What happens when the wind doesn't blow and sun doesn't shine?

The solution is energy storage. If we can retain the power generated by renewables using batteries, we can make up for the shortfall. Regrettably, most battery tech can only store energy for short periods – and the batteries themselves tend to be very large. 

Which is why thermal battery tech is attracting attention. As the name implies, the tech stores electricity as heat, rather than chemical changes. And it uses inexpensive material such as bricks, glass, salt, sand and metal.

Thermal is just one of many emerging battery techs. But its use of cheap materials makes it compelling. In 2024, the University of Michigan researchers revealed they had developed a thermophotovoltaic cell with record efficiency of 44 percent. By comparison, the average steam turbine manages about 35 percent.

Chiplets

Is Moore's Law coming to an end? In 1965, Fairchild Semiconductor executive Gordon Moore forecast that the number of transistors stored on a computer chip would double every two years. His prediction came true. Today's most advanced chips have nearly 50 billion transistors.

However, the limit is near. As the industry closes in on transistors measuring just a few nanometers, it has to deal with unpredictable quantum effects. Chip specialists are looking for solutions to this problem. One of the most promising is the chiplet. 

Traditionally, makers would combine chip designs for multiple purposes – CPU, graphics, memory etc. – on a single 'monolithic' die. With chiplets, they create single-purpose designs and stitch them together into a package.

The tech is not new. But it has been limited by a lack of standardisation across different chip makers. A new consortium called Universal Chiplet Interconnect Express (UCIe) is now working on an open-source approach that any industry player can use. More than 100 semiconductor firms have signed up.