Battery energy storage is essential to improving the reliability of renewable energy. It can collect extra energy from solar and wind power to provide electricity when needed. However, artificial intelligence (AI) is taking battery management to the next level.
Experts say AI software is now key to managing large battery systems. Companies are applying AI not only to basic tasks but also to energy trading, safety monitoring and predictive maintenance.
Advanced AI technologies enhance battery energy storageBattery systems use intelligent tools such as machine learning, deep learning, predictive analytics and reinforcement learning. They are becoming a key tool for managing large battery systems.
By combining these technologies, AI ensures that:
Batteries store and release energy efficiently based on demand.
Optimize performance by processing real-time data to reduce waste and improve efficiency.
These upgrades provide a stable and reliable power source, making battery energy storage more viable and cost-effective.
S&P Global says the market demand for battery energy storage systems is growing. However, the integration of AI is just getting started. Lithium-ion battery energy storage developers are well positioned to meet this demand.
Henrique Ribeiro, lead analyst for batteries and energy storage at S&P Global Commodity Insights, said:
“As market competition intensifies and more capacity is deployed, maximizing returns may become increasingly difficult. Therefore, such tools may become an advantage.”
Improving battery energy storage safety
Scientists, researchers and experts agree that manufacturing high-quality batteries is technically complex and challenging. As lithium-ion battery production grows, especially in China and the United States, AI analysis will become increasingly important.
A major challenge is the rapid pace of innovation. If manufacturing errors go unnoticed, they can cause serious problems. Thermal runaway is one of them, which can cause dangerous fires. However, AI can help detect problems early and avoid costly failures.
Like other energy storage systems, batteries also have safety risks. This is a cause for concern. However, this challenge also provides an opportunity for the industry to improve safety measures.
Organizations such as the Industrial Electrotechnical Committee (IEC) and UL Solutions are raising safety standards. Therefore, effectively managing these risks is essential to maintain industry momentum.
As energy storage technology develops, AI will help optimize operations and ensure grid reliability and sustainability.
Category: news
TDK announces optical breakthrough to solve biggest bottleneck of generative AI
TDK Corp. has announced a breakthrough in optical technology that can process data 10 times faster than existing electronics and solve a key bottleneck that has held back the development of generative AI.
The Apple supplier said it has demonstrated the world’s first “spin photodetector,” which combines optical, electronic and magnetic components and has a response time of 20 picoseconds, or 20 trillionths of a second, potentially replacing existing semiconductor-based photodetectors that transfer data between chips.
The speed at which AI processors can transfer data is severely limited by current electronics, said Hideaki Fukuzawa, senior manager at TDK’s Next Generation Product Development Center.
“This data transfer is the biggest bottleneck for AI, not the performance of semiconductor GPUs,” he said. “Because we were able to break through many of the current bottlenecks, we think this technology will revolutionize the AI and data center industries.”
Arata Tsukamoto, a professor of electrical engineering at Nihon University in Tokyo, who tested the new device for TDK as a research partner, said he believes “spin photodetectors hold extraordinary promise from both a scientific and technological perspective.”
Currently, data is transferred between processors via electrical signals, but the large amounts of data in AI need to turn to optical technology because light travels faster.
TDK plans to conduct further tests to confirm continuous light at ultra-high speeds before providing samples to customers by the end of March 2026 and moving into mass production in the next three to five years.
Although the technology is immature and there are huge challenges in building an ecosystem for the technology with integrated circuit designers, TDK believes its device can have a cost advantage over other solutions by reducing the number of wafer processes.
TDK supplies batteries for iPhones, but it has applied its head technology to hard drives to achieve breakthroughs in photonics.
Its new device also consumes less power — another important issue for the expansion of AI data centers. Augmented reality and virtual reality smart glasses and high-speed image sensors are also potential markets for the technology in the future.
The device is part of a market for photonic integrated circuits that tech researcher IDTechEx predicts will expand more than tenfold over the next decade due to demand for AI generation
Major AI companies have also been working to develop transceivers that integrate optical technology into chip packages, and TDK’s technology will be a challenger to this next-generation silicon photonics.
Taiwan Semiconductor Manufacturing Co., the world’s largest chipmaker, is working on it, with the goal of achieving production within five years.
Global education must incorporate artificial intelligence and be people-centered
In his speech, he highlighted the dual nature of technological advances such as AI which offer both great potential and considerable risks.
“Education is an essential component for everyone to realize their full potential, and for socioeconomic development and prosperity,” said Mr. Guterres.
The promise and risks of AI
The UN Secretary-General stressed that AI and other technological innovations can greatly help students and teachers by providing access to broader information and advanced learning tools.
“But great rewards come with great risks. As AI-driven systems become increasingly powerful, it is easy for human intentions and machine-driven influences to go astray,” he added.
This year, UNESCO, the United Nations educational and cultural agency, dedicated a day to explore the opportunities and challenges of AI
Director-General Audrey Azoulay called for greater investment in teacher and student training so that funds can be used responsibly.
“AI offers major opportunities, provided that its deployment in schools is guided by clear ethical principles. To realize its full potential, this technology must complement the human and social dimensions of learning, not replace them,” said Mr. Azoulay.
Human rights at the heart of AI
This year’s AI Day reminds us that unlocking the potential of AI “depends on putting human agency and human rights at the heart of this rapidly evolving technology,” the UN Secretary-General stressed, calling for guarantees that all users have the right tools and knowledge to “use this technology wisely, safely and ethically.”
UNESCO’s competency framework aims to support learners and teachers in integrating AI into their learning, while the recently adopted Global Digital Compact is expected to help ensure that humans retain control over AI development and governance.
In his closing remarks, Mr. Guterres called for a commitment to “put the humanities at the heart of education systems around the world.”
Divided views onAI
As AI is increasingly integrated into education, countries remain divided over its use. According to the latest data from UNESCO, more than two-thirds of secondary school students in high-income countries are already using generative AI tools to support their learning.
However, education professionals still face significant challenges and a lack of clear guidelines.
In May 2023, a UNESCO survey of 450 educational institutions showed that only 10% of schools and universities had an official framework for the use of AI
Meanwhile, more countries are imposing restrictions on new technologies in the classroom. The latest data from UNESCO shows that nearly 40% of countries now have laws or policies prohibiting the use of mobile phones in schools, a significant increase from 24% in July 2023.
For more insights, listen to Farida Shaheed, the UN Special Rapporteur on the right to education, discuss the role of AI in schools and the challenges it presents in a recent interview with UN News.
T-Mobile 5G Home Internet Review
T-Mobile 5G Home Internet uses signals from nearby 5G base stations, rather than physical cables or fiber, to provide an Internet connection to your home. If you live in an area where your local Internet service provider has limited cable network buildout, T-Mobile 5G Home Internet is a great alternative to cable or fiber. Plus, T-Mobile 5G Home Internet doesn’t require a long-term contract, has simple equipment, and is fast enough for light to moderate use. We like its unlimited data and variety of pricing plans, but if you’re a heavy gamer or regularly watch a lot of HD video, a cable provider (if you have network) may be a better choice. Availability, Plans, and RatesT-Mobile 5G Home Internet isn’t available in all areas, but it has nationwide coverage with more than 6 million subscribers, according to the company’s most recent earnings report. Just enter your street address on the T-Mobile website to see if the service is available in the area you need.
Generally, it’s available in many of the markets where T-Mobile offers 5G service. T-Mobile’s most expensive All-In Home plan costs $75 per month ($70 with autopay and $55 with voice) and offers some nice perks. The plan includes the same 5G gateway and speeds as the Amplified plan, plus an additional Mesh Wi-Fi access point to help with coverage. AT&T and Verizon both offer 5G home internet service, and while coverage isn’t as extensive as T-Mobile’s, their plans and pricing are similar. Verizon ranks second with 4.2 million home internet subscribers. Like T-Mobile, you’ll need to enter your address on the Verizon website to see if the service is available in your area.
Verizon generally doesn’t offer5G home internet service in areas where it offers wired Fios service. Its plans include a modem and router combo, unlimited data, and a two-year price lock guarantee. If you live in an area without wired or wireless coverage, Starlink may be your only option. Starlink relies on satellites and covers most of the U.S., but costs more and offers lower quality service.
Microsoft beefs up Copilot AI with personalized podcasts, real-time vision, automated actions, and chat memories
Microsoft has enhanced the capabilities of its Copilot AI chatbot to make it more helpful and autonomous for users. With increased memory, the AI is now more personalized, can analyze its surroundings in real time to answer complex prompts, and can take actions on behalf of users. Microsoft has made improvements to its Copilot AI chatbot with new features that improve its ability to answer questions, entertain users, and remember all the information spoken.
When users want to be entertained or learn through audio, Copilot can now create customized podcasts based on personal interests and topics. The AI has gained the ability to conduct deep research, which means it will handle complex prompts like a human by solving problems step by step, using information from the web, and combining answers to create useful reports.
For mobile users, the AIcan view your surroundings in real time to assist in answering questions. For Windows users, the AI can view your desktop and work with you to change settings, manipulate files, search for information, and interact with content to assist users with tasks and projects.
Copilot now remembers every interaction and chat (with authorization) and all the information in it. This allows the AIto create pages summarizing a person’s thoughts and notes on various conversations and projects. Users can customize their avatars for more personalization.
The chatbot can automatically provide reminders and suggestions based on what it knows about your life. This also includes finding deals on items the user wants to buy. The new Actions feature allows the AIto complete tasks such as booking flights and dinner reservations on the user’s behalf.
After Siri crisis, Apple joins AI data center race
While other big tech companies invested heavily in AI data centers, Apple (AAPL) chose to sit on the sidelines and avoid a surge in capital spending. But that seems to have changed, and Apple realized it needed to get in on theAI data center game.
Apple is ordering about $1 billion worth of Nvidia (NVDA) GB300 NVL72 systems, Loop Capital analyst Ananda Baruah said late Monday. That equates to about 250 servers, each worth $3.7 million to $4 million, he said in a client note. Apple is working with server makers Dell Technologies (DELL) and Super Micro Computer (SMCI) to develop large server clusters to support generative AI applications.
“AAPL is officially in the large server cluster GenAI game… and SMCI and DELL are the primary server partners,” he said. “While we are still gathering more complete information, it seems likely that this is a Gen AI LLM (large language model) cluster.”
Barua believes that Apple’s change of strategy is due to the trouble it has encountered in bringing its AI-powered Siri digital assistant to market. Apple has delayed the release of the new Siri indefinitely. The company had hoped to launch the AI features earlier this year, after previewing them at the Worldwide Developers Conference last June. Apple has reportedly reorganized its executive team to deal with the company’s difficulties in releasing AI features. According to Bloomberg, an executive called the delays and missteps “unpleasant” and “embarrassing” because the company has been promoting the AI features in TV ads.
DeepSeek launches V3 AI model update to compete with OpenAI. What’s new?
DeepSeek’s AI has sparked a debate about whether cutting-edge platforms can be built for far less than the billions of dollars invested by US companies to build data centers. Chinese AI startup DeepSeek has released an update to its V3 model that promises better programming capabilities.
The V3-0324 update, which was initially announced on Hugging Face this week but has not yet been officially released, claims to address real-world challenges while setting benchmarks for accuracy and efficiency. V3 is actually an older DeepSeek platform, and DeepSeek claims to have significant improvements in benchmark performance across multiple metrics.
It also claims to have improved the style and content quality of Chinese writing features, improved multi-round interactive rewrites, optimized translation quality and letter writing, enhanced reporting analysis requests and output more detailed output, and improved the accuracy of function calls, fixing issues in previous V3 versions.
The startup’s AI service has sparked a debate about whether cutting-edge platforms can be built for far less than the billions of dollars invested by US companies to build data centers. It also highlights the company’s intention to stay ahead of its competitors, especially those from Silicon Valley, such as OpenAI and Google.
Previously, DeepSeek surpassed OpenAI’s ChatGPT to become the most popular free app in Apple’s US App Store.
DeepSeek’s achievements also include the performance of the initial R1 model, which seems to be on par with OpenAI’s best model, but at a fraction of the cost. The cost part was particularly shocking to the industry and triggered a sell-off in AI and technology-related stocks in the US market. This is because the best companies in Silicon Valley have invested huge amounts of money in their artificial intelligence projects, but have only achieved similar results
Samsung and NVIDIA Collaborate to Advance AI in Mobile Networks
Samsung Demonstrates Significant Advances in AI-RAN Technology and Ecosystem Development, Unleashing the Full Potential of Software-Based Networks with NVIDIA AI PlatformSamsung Electronics today announced a collaboration with NVIDIA to advance AI-RAN technology. The collaboration reflects Samsung’s commitment to fostering a strong ecosystem and diverse available computing platforms. The effort aims to support the smooth and easy application of AI in mobile networks by expanding the central processing unit (CPU) ecosystem and strengthening collaboration with graphics processing unit (GPU) companies.
To maximize the power of AI and incorporate it into the radio access network (RAN), Samsung has made significant technological advances since early 2024 by leveraging its internal expertise in AI and radio. One of the key milestones achieved was interoperability between Samsung’s O-RAN-enabled virtualized RAN (vRAN) and NVIDIA accelerated computing, which was achieved at Samsung Research Labs late last year. Samsung successfully demonstrated a proof of concept of how NVIDIA’s accelerated computing can be seamlessly integrated into software-based networks to help enhance AI capabilities.
This achievement further enhances Samsung’s advancement in unique innovations combining AI and RAN. Building on this, Samsung can integrate its vRAN (virtual distributed unit, vDU) with NVIDIA’s accelerated computing into commercial off-the-shelf (COTS) servers with Samsung vRAN software installed, enabling seamless delivery of AI-RAN.
In addition, the two companies will continue to explore the best combination of AI-RAN solutions, leveraging Samsung vRAN with NVIDIA Grace CPU and/or GPU-based AI platforms using Compute Unified Device Architecture (CUDA) technology. All of these options are optimal for every network deployment environment, from rural and suburban areas to densely populated cities.
At MWC 2025, Samsung demonstrated its leadership in AI-For-RAN innovation with two AI-RAN demonstrations. Both demonstrations were endorsed by the AI-RAN Alliance and developed in collaboration with multiple members, including NVIDIA. The demonstrations included AI-based physical uplink shared channel (PUSCH) estimation and non-uniform modulation, showcasing innovative ways to integrate AI into mobile networks.
“AI is changing the telecom landscape, and Samsung is helping operators build the network architecture and environment needed to enable AI with our proven AI-driven vRAN,” said June Moon, executive vice president of R&D for the Networks Business at Samsung Electronics. “This collaboration with NVIDIA reflects our ongoing efforts to expand the GPU and CPU ecosystem, and we look forward to exploring new opportunities in the future.” »
“AI-RAN is a critical technology that will significantly improve network utilization, efficiency, and performance while enabling new AI services,” said Ronnie Vasishta, senior vice president of telecommunications at NVIDIA. “Samsung is a leader in AI-RAN development. Its expertise and vRAN software will be invaluable to our customers.”
As a founding member of the AI-RAN Alliance, which was established in 2024, Samsung is actively collaborating with academic institutions and industry leaders such as NVIDIA to advance AI-RAN technology. As vice chair-elect of the board and Working Group 3 (AI-on-RAN), Samsung is leading the industry’s transformation to next-generation AI networks.
Samsung’s end-to-end software network architecture provides the best foundation for easy deployment and adoption of AI at every layer of the network. By doing so, Samsung will be able to support operators with flexible networks, enhance their competitive advantage, and maintain leadership in the AI era. This advancement paves the way for leveraging network infrastructure not only for mobile communications but also for general workloads, providing a data center-like network architecture that opens up new business opportunities
China’s new AI model “Manus” attracts global attention, challenging OpenAI and Google
Manus is an advanced AI agent designed to think, plan, and perform real-world tasks independently. It can create websites, plan trips, analyze stocks, and more, all with a single user prompt. Just weeks after the launch of DeepSeek, China has unveiled another powerful artificial intelligence (AI) model, Manus, highlighting the country’s accelerating momentum to join the AI race. Developed by Chinese startup Monica, Manus is comparable to top AIsystems created by OpenAI, Google, and Anthropic. The company claims the model is a general AI that can perform tasks autonomously without human supervision.
What is Manus?
Manus is an advanced AI agent designed to think, plan, and perform real-world tasks independently. It can create websites, plan trips, analyze stocks, and more — all with a single user prompt. Unlike standard AI chatbots that provide answers, Manus takes comprehensive actions to complete tasks. For example, if asked to create a report on climate change, it will conduct research, write a report, create charts, and compile everything into a final document without further instructions. Chinese startup Monica has just launched Manus, the world’s first general AIagent. #Manus achieves state-of-the-art (SOTA) performance on all three difficulty levels, outperforming #OpenAI’s DeepResearch on the GAIA real-world problem-solving benchmarkWhat makes Manus so unique?
Manus was launched on March 6 and has quickly gained global attention. According to its creators, it outperforms OpenAI’s DeepResearch on the GAIA benchmark, a measure of AI performance. The demo video released by Monica shows Manus interacting with the internet, collecting data, and performing complex tasks in real time. It can browse websites, take screenshots, record online activity, and generate reports, spreadsheets, or presentations. This level of automation has many calling it a major leap forward in AI technologyMain features of Manus
Manus runs independently in the cloud, and it continues to perform assigned tasks even if the user disconnects the device. This feature ensures that long-term projects can proceed uninterrupted.
Unlike many AI models, Manus actively browses the web, interacts with websites, and displays its workflow in real time. This helps users understand how AI collects and processes information.
It learns from user interactions to provide customized results. Over time, it adapts to user preferences, improving the relevance and quality of responses.
The AI can access platforms like X (formerly Twitter), Telegram, and others to collect and process data. It can even manage multiple screens at once, as shown in its official video. Manus does more than just produce text-based results. It can create detailed reports, interactive presentations, and even code-based outputs like data visualizations and spreadsheets.How to use Manus AI
Manus functions similarly to AI chatbots like ChatGPT, but with greater autonomy. Users simply enter a task, such as “create a 7-day Bali itinerary within budget,” and Manus starts researching, collecting data, and formulating responses. The AI compiles all the relevant information and provides a complete itinerary with links, maps, and travel suggestions. If the user loses connection, the AIcontinues working in the cloud and notifies them when the task is complete.
Availability and future plans
Currently, Manus is available through an invite-only web preview. Monica has not announced a public release date, but hinted that it may be available soon. The company also plans to open source the model in the coming months, allowing developers to integrate it into their own projects. This move could lead to rapid improvement and widespread adoption of the technology.
Tencent releases new AI model, claims to be faster than DeepSeek-R1
Tencent on Thursday unveiled a new artificial intelligence model that it says can answer queries faster than global hit DeepSeek’s R1, the latest sign that the startup’s success at home and abroad is putting pressure on its larger Chinese rivals.Tencent said in a statement that Hunyuan Turbo S can answer queries in under a second, differentiating it from “DeepSeek R1, Hunyuan T1 and other slow-thinking models that need to ‘think for a while before answering.’” Tencent added that in tests of knowledge, math and reasoning, Turbo S’s capabilities were comparable to DeepSeek-V3, which powers DeepSeek’s AI chatbot that has surpassed OpenAI’s ChatGPT in app store downloads.DeepSeek did not immediately respond to a request for comment.The success of DeepSeek’s R1 and V3 models, the first time a Chinese company has received widespread acclaim and adoption in Silicon Valley, has also prompted Chinese tech giants such as Tencent to scramble to launch new versions of the AI models they began developing after OpenAI‘s ChatGPT came out in late 2022. Last month, just days after DeepSeek-R1 shook up the global tech order and triggered a sell-off in AI stocks outside of China, e-commerce giant Alibaba (9988.HK), opened in a new tab, released the Qwen 2.5-Max model, claiming that it outperformed DeepSeek-V3 in all aspects. Tencent also said that the new Turbo S is many times cheaper to use than previous generations, highlighting how DeepSeek’s open source and low-price strategy has forced other leading Chinese AI companies to charge users less.