विज्ञान

(Reuters) -Meta is cutting around 600 positions out of the several thousand roles in its Superintelligence Labs, the Facebook owner said on Wednesday as it looks to make its artificial intelligence unit more flexible and responsive. The job cuts will affect Facebook Artificial Intelligence Research (FAIR) unit, as well as teams focused on product-related AI and AI infrastructure, the company said. The newly formed TBD Lab, which comprises "a few dozen" researchers and engineers developing Meta's next-generation foundation models, will not be affected. Axios first reported the job cuts announcement, citing a company memo. Chief AI Officer Alexandr Wang said fewer team members would streamline decision-making and increase the responsibility, scope and impact of each role. The company said it was encouraging affected employees to apply for other jobs within Meta. Meta on Tuesday struck a $27 billion financing deal with Blue Owl Capital, the company's largest-ever private capital agreement, to fund its biggest data center project. Some analysts said the deal will allow Meta to achieve its massive AI ambitions by shifting much of the upfront cost and risk to external capital, while retaining a smaller ownership share in the project. The parent of Facebook and Instagram reorganized its AI efforts under Superintelligence Labs in June after senior staff departures and a poor reception for its open-source Llama 4 model. CEO Mark Zuckerberg personally led an aggressive hiring spree for the unit to revitalize Meta's AI efforts. Superintelligence Labs includes Meta's foundations, product and FAIR teams as well as TBD Lab that is focused on developing the company's next generation of AI models. The company began investing in AI in 2013 by launching FAIR unit and recruiting Yann LeCun, its chief AI scientist, to lead the effort and building a global research network focused on deep learning. (Reporting by Jaspreet Singh in Bengaluru; Editing by Vijay Kishore and Arun Koyyur)

(The article has been published through a syndicated feed. Except for the headline, the content has been published verbatim. Liability lies with original publisher.)

By Max A. Cherney SAN FRANCISCO (Reuters) -Google said it has developed a computer algorithm that points the way to practical applications for quantum computing and will be able to generate unique data for use with artificial intelligence. The new algorithm called Quantum Echoes, which runs on the company's quantum chip, is 13,000 times faster than the most sophisticated classical computing algorithm on supercomputers, Google said. In the future, the Quantum Echoes algorithm may be able to help measure molecular structure in molecules which could aid in drug discovery and help material science through identifying new types of materials, company executives told a media briefing last week. Alphabet's Google is among several major tech firms including Amazon and Microsoft investing in quantum computing – which promises to speed up computing and solve problems out of reach for today's machines. Last year, Google unveiled its quantum chip, Willow, that the company said is able to overcome a crucial problem with "qubits", the building blocks of quantum computing. The development of the algorithm was roughly equivalent in significance to the chip, the executives said. The algorithm is also verifiable with other quantum computers or through experiments. Verifiable data means that it can lead to practical applications. "If I can't tell you the data is correct, if I can't prove to you the data is correct, how can I do anything with it?" Google staff research scientist Tom O'Brien said. For artificial intelligence, Google's engineers hope to be able to use the algorithm to help create new data sets for uses in areas such as life sciences where good data sets do not exist to train AI models with. Google published details about the Quantum Echoes algorithm in the scientific journal Nature on Wednesday. (Reporting by Max A. Cherney in San Francisco; Editing by Edwina Gibbs)

(The article has been published through a syndicated feed. Except for the headline, the content has been published verbatim. Liability lies with original publisher.)

By Max A. Cherney SAN FRANCISCO (Reuters) -Google said it has developed a computer algorithm that points the way to practical applications for quantum computing and will be able to generate unique data for use with artificial intelligence. The new algorithm called Quantum Echoes, which runs on the company's quantum chip, is 13,000 times faster than the most sophisticated classical computing algorithm on supercomputers, Google said. In the future, the Quantum Echoes algorithm may be able to help measure molecular structure in molecules which could aid in drug discovery and help material science through identifying new types of materials, company executives told a media briefing last week. Alphabet's Google is among several major tech firms including Amazon and Microsoft investing in quantum computing – which promises to speed up computing and solve problems out of reach for today's machines. Last year, Google unveiled its quantum chip, Willow, that the company said is able to overcome a crucial problem with "qubits", the building blocks of quantum computing. The development of the algorithm was roughly equivalent in significance to the chip, the executives said. The algorithm is also verifiable with other quantum computers or through experiments. Verifiable data means that it can lead to practical applications. "If I can't tell you the data is correct, if I can't prove to you the data is correct, how can I do anything with it?" Google staff research scientist Tom O'Brien said. For artificial intelligence, Google's engineers hope to be able to use the algorithm to help create new data sets for uses in areas such as life sciences where good data sets do not exist to train AI models with. Google published details about the Quantum Echoes algorithm in the scientific journal Nature on Wednesday. (Reporting by Max A. Cherney in San Francisco; Editing by Edwina Gibbs)

(The article has been published through a syndicated feed. Except for the headline, the content has been published verbatim. Liability lies with original publisher.)

By Harshita Mary Varghese (Reuters) -AT&T added more wireless subscribers than expected in the third quarter, as bundled plans and heavy promotions around the latest iPhone launch helped it attract more customers in a fiercely competitive market. Shares of the company slipped about 2% in morning trading on Wednesday after the U.S. wireless carrier slightly missed estimates for quarterly revenue on lower-than-expected equipment sales in its mobility business. The quarter ended September is a critical period for U.S. wireless carriers, with competition peaking around Apple's annual iPhone launch. Like all other carriers, AT&T has also rolled out lucrative promotions for the iPhone 17 series to lock in new subscribers and encourage existing users to upgrade to pricier plans. Equipment revenue from its mobility unit grew 6.1% in the quarter to $4.79 billion, missing Visible Alpha estimates of $4.93 billion. Operating costs in the division rose 3.8%, driven by higher expenses from selling more pricier phones and increased spending on marketing and promotions. AT&T said that the fourth quarter typically brings a seasonal uptick in customers switching carriers and higher device upgrade activity during the holiday period. Analysts at MoffettNathanson said higher upgrade rate in the fourth quarter would likely be a larger impact for AT&T and could significantly pressure margins going forward. "The wireless industry has benefited from an inordinately low upgrade rate in recent years. Just seeing a normalization would be a stiff headwind for churn, ARPU, and margins," the brokerage said. Each upgrade cycle costs the carriers a small fortune in subsidies and typically leads to higher churn as customers are more likely to switch providers when they are getting a new handset. AT&T added 405,000 monthly bill paying wireless subscribers, compared to expectations for 334,100 additions, according to FactSet. By packaging wireless and fiber broadband services together for a discount, AT&T has encouraged more customers to take multiple offerings and reduce customer exits. More than 41% of fiber households have also opted for mobile plans. Third-quarter total broadband net adds of 558,000 were the highest for AT&T in more than eight years. Revenue from its business wireline unit fell 7.8% in the quarter, hurt by declines in legacy voice and data services. On an adjusted basis, AT&T earned 54 cents per share, largely in line with estimates, according to LSEG data. Total third-quarter revenue stood at $30.7 billion, missing estimates of $30.87 billion. (Reporting by Harshita Mary Varghese in Bengaluru; Editing by Arun Koyyur and Sriraj Kalluvila)

(The article has been published through a syndicated feed. Except for the headline, the content has been published verbatim. Liability lies with original publisher.)

By Akash Sriram and Abhirup Roy (Reuters) -Tesla is expected to post a surge in third-quarter results later on Wednesday, thanks to U.S. buyers rushing to take advantage of an expiring $7,500 federal electric vehicle tax credit. But investors and analysts may be more focused on Chief Executive Elon Musk's outlook, including whether new, cheaper versions of its Model 3 and Model Y Tesla can keep U.S. customers buying and attract new ones in Europe and Asia.  The new trims – called Standard – are $5,000 to $5,500 cheaper than their predecessors as Tesla shrank the battery, switched to a less-powerful motor and stripped out myriad features – from rear touchscreens to seat-back pockets. The company also temporarily has slashed lease prices on the higher-priced Premium versions. The cheaper Standard variants as well as Tesla's offers and discounts to stave off competition globally throughout the year have pressured the company's once-enviable margins, worrying investors. Sales of Tesla's aging lineup declined for the first time last year and analysts expect an 8.5% fall this year, in part because of Musk's far-right political rhetoric. ROBOTAXIS ROLLOUT UPDATE Musk also will be expected to update on robotaxis, the project that he sees as the key to Tesla's next stage of growth. Musk has said Tesla's robotaxis will serve half the population of the U.S. by year-end. "What are the relevant metrics – fleet size, cumulative miles, and territories – you expect in Q4 and 2026?" was the top question for the CEO in a note by analysts at Cantor Fitzgerald on Tuesday.  While Musk has pivoted Tesla's focus to robotics and artificial intelligence and much of the company's $1.4 trillion valuation hangs on that bet, most of its current revenue and profit comes from vehicle sales.  Tesla shares, which were down 1.6% on Wednesday, have risen nearly 10% this year, propelled by a record-breaking $1 trillion pay package for Musk that the board proposed last month. The stock is still largely among the laggards in the 'Magnificent 7' group of mega-cap stocks. Analysts expect Tesla to report a revenue of $26.24 billion in the quarter ended September, up 4.2% from a year earlier, according to data compiled by LSEG. Tesla also will report on whether regulatory credits that it sold to makers of gasoline-powered cars to satisfy pollution standards have faded away after policy changes by the Trump administration. Automotive gross margin, excluding regulatory credits, is estimated to be 15.6%, according to 19 analysts polled by Visible Alpha. That is lower than 17.05%, a year earlier. (Reporting by Akash Sriram in Bengaluru and Abhirup Roy in San Francisco; editing by Peter Henderson and Stephen Coates)

(The article has been published through a syndicated feed. Except for the headline, the content has been published verbatim. Liability lies with original publisher.)

(Corrects fourth bullet to attribute data to Visible Alpha, not LSEG) By Akash Sriram and Abhirup Roy (Reuters) -Tesla is expected to post a surge in third-quarter results later on Wednesday, thanks to U.S. buyers rushing to take advantage of an expiring $7,500 federal electric vehicle tax credit. But investors and analysts may be more focused on Chief Executive Elon Musk's outlook, including whether new, cheaper versions of its Model 3 and Model Y Tesla can keep U.S. customers buying and attract new ones in Europe and Asia.  The new trims – called Standard – are $5,000 to $5,500 cheaper than their predecessors as Tesla shrank the battery, switched to a less-powerful motor and stripped out myriad features – from rear touchscreens to seat-back pockets. The company also temporarily has slashed lease prices on the higher-priced Premium versions. The cheaper Standard variants as well as Tesla's offers and discounts to stave off competition globally throughout the year have pressured the company's once-enviable margins, worrying investors. Sales of Tesla's aging lineup declined for the first time last year and analysts expect an 8.5% fall this year, in part because of Musk's far-right political rhetoric. ROBOTAXIS ROLLOUT UPDATE Musk also will be expected to update on robotaxis, the project that he sees as the key to Tesla's next stage of growth. Musk has said Tesla's robotaxis will serve half the population of the U.S. by year-end. "What are the relevant metrics – fleet size, cumulative miles, and territories – you expect in Q4 and 2026?" was the top question for the CEO in a note by analysts at Cantor Fitzgerald on Tuesday.  While Musk has pivoted Tesla's focus to robotics and artificial intelligence and much of the company's $1.4 trillion valuation hangs on that bet, most of its current revenue and profit comes from vehicle sales.  Tesla shares have risen nearly 10% this year, propelled by a record-breaking $1 trillion pay package for Musk that the board proposed last month. The stock is still largely among the laggards in the 'Magnificent 7' group of mega-cap stocks. Analysts expect Tesla to report a revenue of $26.24 billion in the quarter ended September, up 4.2% from a year earlier, according to data compiled by LSEG. Tesla also will report on whether regulatory credits that it sold to makers of gasoline-powered cars to satisfy pollution standards have faded away after policy changes by the Trump administration. Automotive gross margin, excluding regulatory credits, is estimated to be 15.6%, according to 19 analysts polled by Visible Alpha. That is lower than 17.05%, a year earlier. (Reporting by Akash Sriram in Bengaluru and Abhirup Roy in San Francisco; editing by Peter Henderson and Stephen Coates)

(The article has been published through a syndicated feed. Except for the headline, the content has been published verbatim. Liability lies with original publisher.)

LONDON (Reuters) -Spending on new UK data centres is set to rise to 10 billion pounds a year by 2029, a more than five-fold increase from 2024, analysis by construction data firm Barbour ABI has found. A total of 1.75 billion pounds was spent on data centres in the UK last year, with 2.38 billion forecast to be spent in 2025. AI-driven demand is fuelling data centre investment. Tech giants are seen investing 25 billion pounds into the UK over the next five years, Barbour ABI found, and almost 100 new data centre projects are planned. Growth is being supported by government initiatives such as AI Growth Zones which aim to streamline the planning process for new digital infrastructure. London and surrounding areas have traditionally dominated the data centre sector, but development is increasingly expanding nationwide, according to Barbour ABI. The largest planned UK data centre project is a $13 billion "hyperscale" data centre in North East England, proposed by U.S. private equity group Blackstone. Last month, Britain and the United States agreed a technology pact with top U.S. firms including Microsoft, Nvidia and Google, among those pledging investments in the UK. Global data centre demand and planned projects have surged since ChatGPT was released in late 2022, as big money and governments bet generative AI will revolutionise the way we work and live. (Reporting by Lucy Raitano, editing by Ed Osmond)

(The article has been published through a syndicated feed. Except for the headline, the content has been published verbatim. Liability lies with original publisher.)

By Stephen Nellis SAN FRANCISCO (Reuters) -Biotech firm Metagenomi said it is using artificial intelligence chips from Amazon.com's cloud computing unit to power some of its work in developing gene-editing technologies. The work represents one of the first major uses of Amazon Web Services chips beyond chatbots and other products powered by large language models, with Metagenomi saying AWS's Inferentia chips had proven far more cost-effective than products from rivals like Nvidia. Emeryville, California-based firm Metagenomi is one of a number of companies working to develop tools to inject genetic material into the human body where it can edit genes to treat diseases. To do that, Metagenomi sifts through the natural world looking for proteins that might help with the challenge of precisely delivering a treatment to any gene in any cell of the human body. When it finds one that looks close, it uses AI to generate many similar examples to find just the right ones. "So we did this impressive thing where we generated over a million different proteins from this rare class of enzyme that we use for doing gene editing. And in that case, it was really a clear cost advantage for using the Inferentia platform," Chris Brown, head of discovery for Metagenomi. "Unless you cast a broad enough net in the beginning, you actually just miss what can be the most important discoveries and systems entirely," he added. The Inferentia chips – which were introduced in 2019 to power AI features for Amazon's Alexa virtual assistant – were able to do the same job as rival chips at roughly half the price, Brown said. (Reporting by Stephen Nellis in San Francisco; Editing by Edwina Gibbs)

(The article has been published through a syndicated feed. Except for the headline, the content has been published verbatim. Liability lies with original publisher.)

By Stephen Nellis SAN FRANCISCO (Reuters) -Biotech firm Metagenomi said it is using artificial intelligence chips from Amazon.com's cloud computing unit to power some of its work in developing gene-editing technologies. The work represents one of the first major uses of Amazon Web Services chips beyond chatbots and other products powered by large language models, with Metagenomi saying AWS's Inferentia chips had proven far more cost-effective than products from rivals like Nvidia. Emeryville, California-based firm Metagenomi is one of a number of companies working to develop tools to inject genetic material into the human body where it can edit genes to treat diseases. To do that, Metagenomi sifts through the natural world looking for proteins that might help with the challenge of precisely delivering a treatment to any gene in any cell of the human body. When it finds one that looks close, it uses AI to generate many similar examples to find just the right ones. "So we did this impressive thing where we generated over a million different proteins from this rare class of enzyme that we use for doing gene editing. And in that case, it was really a clear cost advantage for using the Inferentia platform," Chris Brown, head of discovery for Metagenomi. "Unless you cast a broad enough net in the beginning, you actually just miss what can be the most important discoveries and systems entirely," he added. The Inferentia chips – which were introduced in 2019 to power AI features for Amazon's Alexa virtual assistant – were able to do the same job as rival chips at roughly half the price, Brown said. (Reporting by Stephen Nellis in San Francisco; Editing by Edwina Gibbs)

(The article has been published through a syndicated feed. Except for the headline, the content has been published verbatim. Liability lies with original publisher.)

(Reuters) -FalconX said on Wednesday it will buy crypto investment management firm 21shares, as the digital assets trading firm looks to expand its exchange-traded funds offerings. The deal comes just over a month after Wall Street's top regulator removed the last remaining hurdle to dozens of new spot ETFs tied to cryptocurrencies ranging from solana to dogecoin. The terms of the deal were not disclosed. The Trump administration has aligned itself with the digital assets sector via long-awaited regulatory clarity, marking a stark turnaround from the years of struggle and a legal battle that paved the way for the approval of the first bitcoin ETF in January 2024. "We're witnessing a powerful convergence between digital assets and traditional financial markets, as crypto ETPs (exchange-traded products) open new channels for investor participation through regulated, familiar structures," said FalconX CEO Raghu Yarlagadda. Founded in 2018 by Hany Rashwan and Ophelia Snyder, 21shares now manages over $11 billion in assets across dozens of products.  FalconX plans to leverage 21shares' expertise in crypto ETFs and its brokerage platform to advance the adoption of digital asset investment products.  Although updated standards from the Securities and Exchange Commission are expected to unleash a flurry of crypto ETFs, the U.S. government shutdown could curb the agency's ability to review and approve these filings. Concerns are also mounting over highly leveraged funds tracking crypto and related companies. Earlier this month, the sector saw its largest selloff ever after U.S. President Donald Trump renewed trade tensions with China. FalconX, which was valued at $8 billion in a 2022 funding round, has facilitated over $2 trillion in trading volume and has a client base exceeding 2,000 institutions. (Reporting by Ateev Bhandari and Arasu Kannagi Basil in Bengaluru; Editing by Sahal Muhammed)

(The article has been published through a syndicated feed. Except for the headline, the content has been published verbatim. Liability lies with original publisher.)