Search this site

Meta will pay up to $18 billion and introduce major restrictions for teenage Facebook and Instagram users, potentially setting a new benchmark for the financial and regulatory costs facing social-media companies. Meta has agreed to pay up to $18 billion and introduce sweeping restrictions for teenage users of Facebook and Instagram, bringing one of the biggest legal challenges facing the social-media industry to a dramatic conclusion. The agreement resolves claims brought by nearly all U.S. states accusing Meta of designing features that encouraged young users to spend excessive amounts of time on its platforms. Meta has denied wrongdoing. Under the settlement, the company will introduce significant changes to how teenagers use Facebook and Instagram, including restrictions on certain features and additional protections intended to limit compulsive use. The financial cost is substantial. But for investors, the bigger question is whether the settlement changes the operating rules for the wider social-media industry. Facebook and Instagram are built around engagement. The longer users spend scrolling, watching videos and interacting with content, the more opportunities platforms have to display advertisements. That model has helped Meta build one of the world's largest advertising businesses. Restrictions affecting how younger users interact with social-media platforms could therefore create a difficult balance. Companies must continue making their services attractive enough to retain users while complying with increasingly demanding safety requirements. The impact may not stop with Meta. Other major platforms, including TikTok and YouTube, face similar scrutiny over how their products are used by children and teenagers. Meta's agreement could provide regulators elsewhere with a template for demanding additional safeguards from competing platforms. The settlement also removes a major source of uncertainty. Meta had been facing potentially enormous penalties if the dispute continued through court. Resolving the case allows the company to absorb a defined financial cost while adjusting its products under agreed rules. Meta remains an exceptionally profitable company, meaning even a settlement of this size is unlikely to threaten the business itself. The longer-term significance lies elsewhere. For years, social-media companies have been rewarded for finding increasingly effective ways to keep people engaged. Regulators are now placing more limits on how that strategy can be applied to younger users. Meta's settlement suggests those limits may increasingly come with a very large price tag.

Bank of America now spends more than $250 million annually on GLP-1 drugs, forcing corporate America to weigh higher healthcare costs against the possibility of healthier employees and long-term savings. Weight-loss drugs have become a multibillion-dollar business for pharmaceutical companies. For large employers, they are also becoming one of the fastest-growing items on the healthcare bill. Bank of America now spends more than $250 million a year providing GLP-1 drugs to its employees, according to CEO Brian Moynihan. That represents approximately 13% of the bank’s entire $2 billion annual healthcare budget. The figure is particularly striking because these medicines accounted for almost none of the company’s healthcare spending just four or five years ago. Demand has since accelerated as drugs such as Wegovy and Zepbound have moved beyond diabetes treatment and into the much larger obesity market. With around 211,000 employees, Bank of America has the scale to negotiate directly with drugmakers over prices. It also combines access to GLP-1 treatments with coaching and other health programmes designed to help employees achieve longer-lasting results. Moynihan believes the expense could eventually pay for itself. Obesity is connected to several costly long-term conditions, including cardiovascular disease. If the treatments improve employee health and reduce the need for other medical care, some of today’s spending could be recovered over time. That argument remains difficult to prove, however. GLP-1 drugs can cost thousands of dollars per patient each year, and some users regain weight after stopping treatment. Employers must therefore decide whether they are funding a temporary prescription or making a long-term commitment. Bank of America has chosen to maintain broad access, but many companies are taking a more cautious approach. Some have tightened eligibility requirements, introduced mandatory coaching or removed weight-loss coverage altogether as costs increased. This creates a new challenge for corporate America. GLP-1 medicines may improve employee health, reduce absenteeism and make benefits packages more attractive, but they can also consume a significant portion of an employer’s healthcare budget before the long-term savings are clear. Bank of America’s $250 million bill shows just how quickly the market has changed. Weight-loss drugs are no longer only a pharmaceutical success story. They are becoming a major financial decision for every large employer paying for them.

Adobe is changing CEOs after more than 18 years under Shantanu Narayen, handing new leadership the difficult task of protecting one of software’s strongest franchises as generative AI reshapes creative work. Adobe is preparing for its biggest leadership change in nearly two decades, and the timing could hardly be more significant. Long-time CEO Shantanu Narayen is moving to the role of executive chair after more than 18 years leading the company, with Anil Chakravarthy taking over as chief executive. Narayen helped transform Adobe from a company known primarily for packaged creative software into one of the world’s most successful subscription-software businesses. Products including Photoshop, Illustrator and Premiere Pro became essential tools across creative industries, while Adobe built a highly profitable recurring-revenue model around them. Artificial intelligence is now challenging that position. Generative AI has dramatically lowered the barrier to creating images, video and other digital content. New platforms can produce work from simple text instructions, potentially reducing the need for some of the traditional workflows that helped make Adobe’s software dominant. Investors have noticed. Adobe shares have fallen roughly 18% this year as Wall Street questions whether the company can turn AI into another growth engine quickly enough to defend its position against AI-native competitors. Adobe has not ignored the shift. Its Firefly generative-AI platform has been integrated across its creative products, allowing users to generate and modify images, video and other content directly within familiar Adobe applications. The opportunity is significant. Adobe already has an enormous base of professional users and decades of experience building tools for creators and businesses. If AI makes those products faster and more valuable, it could strengthen the company rather than replace it. But the competitive landscape is changing quickly. Chakravarthy therefore inherits a company that remains highly profitable and deeply embedded in the creative economy, but one facing perhaps the biggest technological disruption since Adobe was founded. That makes this leadership transition about much more than replacing a long-serving CEO. Adobe successfully navigated the transition from packaged software to subscriptions. Its next challenge is proving it can navigate another transformation, this time from traditional creative software into an increasingly AI-driven world.

China’s humanoid robots have learned to run, dance and perform backflips. Now manufacturers must prove that the machines can work reliably and generate enough value to justify their cost. Humanoid robots have spent the past several years running races, performing backflips and dancing for increasingly impressed audiences. The next challenge is considerably less glamorous: proving they can perform useful work at a competitive price. That question is taking centre stage at the World Robot Conference in Beijing, where more than 300 companies are displaying over 2,000 exhibits and 150 new products. China has become one of the most active markets for humanoid robotics, supported by substantial investment, a large manufacturing base and access to the components needed to build machines at scale. Companies including Unitree Robotics have generated considerable attention from investors, but widespread commercial use remains limited. Some robots are already completing relatively simple tasks, including delivering items in hotels and moving materials around factories. However, operating reliably in unpredictable environments is much harder than completing a choreographed demonstration. A robot working on an assembly line must recognise objects, respond to changes and operate safely alongside people for extended periods. It must also generate enough value to justify the cost of purchasing, programming and maintaining it. Those requirements remain difficult to meet. Industry experts warn that many humanoid robots are still too expensive for the tasks they can perform. Some estimates suggest that as many as 70% may initially be used to collect training data rather than perform productive work. The World Humanoid Robot Games will provide another test. Machines will compete in practical activities including packing goods, assembling products, charging electric vehicles and providing office services. The aim is to measure autonomy and reliability, not simply athletic ability. Humanoid robots still offer substantial long-term potential. A machine capable of using tools and navigating spaces designed for humans could eventually be deployed without factories or workplaces being completely redesigned. But technical ability alone will not create a successful industry. Manufacturers must reduce costs, improve reliability and demonstrate that their products can solve genuine business problems. The excitement surrounding humanoid robots has produced ambitious forecasts and high company valuations. The next phase will be more demanding. Investors will need to separate the machines designed to attract attention from those capable of becoming commercially useful.

In a truly heartwarming and emotional moment yesterday, Elon Musk's Neuralink demonstrated a breakthrough in 'reading the thoughts' of a volunteer implant wearer. Watch the moment here with thanks to TheTheorist_In This gives hope to sufferers of "locked in" syndrome who currently have very limited ways of communicating with the outside world. Its the kind of technological breakthrough we love to see here at WSE!

A recent cybersecurity investigation has highlighted an emerging challenge for AI developers: balancing strong safety guardrails with practical usability. As businesses adopt AI across critical industries, getting that balance right could become an increasingly important competitive advantage. Artificial intelligence companies have spent the past two years racing to build smarter models while adding increasingly sophisticated safety guardrails to prevent misuse. But one recent cybersecurity incident suggests those same protections could create an unexpected challenge as businesses decide which AI tools to rely on. During an investigation into a rogue AI agent that targeted AI development platform Hugging Face, researchers found that some leading U.S. AI models were unable to fully assist with parts of the defensive cybersecurity work because their built-in safety systems treated certain requests as potentially harmful. To continue the investigation, researchers turned to an open-source AI model developed by Chinese startup Zhipu AI. According to Reuters, the model was able to analyze logs and support the investigation without the same restrictions. OpenAI later expanded trusted access for Hugging Face, allowing approved security researchers broader capabilities for legitimate defensive work. The incident highlights a challenge that is becoming increasingly important as AI moves from demonstrations to real-world business applications. Safety guardrails are essential for reducing the risk of AI being used for cyberattacks, fraud and other malicious activities. At the same time, security professionals often need AI to examine the same kinds of code, attack methods and system vulnerabilities that criminals would use. Designing a model that can tell the difference between legitimate defence and malicious intent is proving to be far more difficult than simply making it more capable. For investors, that raises an interesting question. As enterprise AI adoption accelerates, companies won't judge models solely on benchmark scores or headline performance. They'll also evaluate how effectively those models fit into real workflows, whether that's software development, legal research, healthcare or cybersecurity. If legitimate users regularly encounter restrictions that slow critical work, businesses may begin looking for alternatives that offer greater flexibility while still maintaining appropriate safeguards. That doesn't mean companies should remove AI safety measures. Responsible guardrails remain essential as these systems become more powerful and widely adopted. Instead, the challenge is building smarter controls that can recognise trusted users and legitimate use cases without creating unnecessary friction. OpenAI's decision to broaden trusted access following the incident suggests the industry is already moving in that direction. The episode serves as a reminder that the AI race is becoming more nuanced. Building the most advanced model is only part of the equation. As organisations integrate AI into everyday operations, ease of use, reliability and thoughtful security controls may become just as important in determining which platforms earn long-term customer trust. For the companies developing AI, the next competitive advantage may not come from making models dramatically smarter. It may come from making them easier for legitimate users to use when it matters most.

Paramount Skydance’s $110 billion Warner Bros. Discovery deal is facing a costly new hurdle, with mounting fees and financing expenses showing investors just how expensive regulatory delays can become. Paramount Skydance wants the opponents of its Warner Bros. Discovery merger to provide a $1.88 billion bond, arguing that delays to the transaction could leave it with enormous and unrecoverable costs. Twelve US states and the Writers Guild of America are attempting to block the proposed merger, which values Warner Bros. Discovery at approximately $110 billion. They argue that combining the two media companies could weaken competition, increase prices and reduce opportunities for workers across film and television. Paramount disputes those claims and maintains that the combined company would be better equipped to compete with Netflix, Disney and other global streaming businesses. The disagreement is now becoming extremely expensive. Under the merger agreement, Paramount must begin paying Warner Bros. Discovery shareholders a “ticking fee” if the deal is not completed by September 30. The payment is worth approximately $7 million for every calendar day the transaction remains unfinished. A trial covering the states’ antitrust challenge is scheduled for March 2027. Paramount estimates that it could accumulate approximately $1.3 billion in unrecoverable ticking fees by the time the trial and final legal submissions are completed in April. If the merger remains delayed until June 2027, the company says its ticking fees could reach $1.7 billion. It could also face another $190 million in additional financing costs. Paramount is therefore asking the court to require the challengers to provide a $1.88 billion bond. If the merger is ultimately allowed to proceed, the company wants the ability to recover some of the losses caused by the delay. The states argue that Paramount knowingly accepted the ticking-fee arrangement when negotiating the acquisition and should not transfer that risk to public authorities. A judge will decide whether a bond is necessary and, if so, how large it should be. The request illustrates how quickly the economics of a major acquisition can change. A deal may appear attractive when it is announced, but financing expenses, legal challenges and contractual penalties can continue accumulating while regulators and courts consider its future. Paramount has already received approval for the merger in dozens of countries, but the remaining US lawsuit could prove to be its most expensive obstacle. Even if Paramount ultimately wins, waiting for that victory may cost the company billions.

HIVE has secured a five-year AI cloud contract as it converts infrastructure developed for cryptocurrency mining into a growing GPU-computing business. HIVE Digital Technologies is using the infrastructure it built for cryptocurrency mining to pursue a potentially more dependable source of revenue. The company’s BUZZ HPC division has signed a five-year AI cloud-services agreement worth approximately $350 million. The contract is expected to add about $70 million in annualised revenue once the necessary computing systems are fully operating. BUZZ plans to install a dedicated cluster containing 2,016 Nvidia Blackwell Ultra GPUs at the Bell AI Fabric facility in Merritt, British Columbia. The system will support demanding tasks including AI training and inference, although HIVE has not identified the customer beyond describing it as an investment-grade enterprise. The equipment is expected to begin operating during the fourth quarter of 2026. HIVE estimates that deploying the GPUs and supporting infrastructure will require approximately $185 million in capital spending. It will receive an upfront deposit of around $35 million from the customer, with financing expected to cover part of the remaining cost. The contract highlights a broader shift taking place across the cryptocurrency-mining industry. Crypto miners already control large data-centre sites, power agreements and cooling systems. Those assets can sometimes be adapted to host AI hardware, allowing operators to diversify away from the sharp price swings and changing economics of digital-asset mining. HIVE is particularly well placed to make that transition because the Merritt facility is powered by hydroelectricity. It will also use closed-loop liquid cooling, which the company says will avoid continuous water consumption while managing the heat created by high-performance GPUs. However, the strategy is not automatically low-risk. AI infrastructure is expensive, hardware can become outdated quickly and delivering the projected revenue depends on completing the facility on schedule. The identity of the customer has also not been disclosed, limiting the information available to investors assessing the agreement. Even with those uncertainties, a five-year contract provides considerably more revenue visibility than cryptocurrency mining normally offers. HIVE will also retain ownership of the Nvidia equipment after the agreement ends, potentially allowing it to continue generating income from the cluster. The company is no longer relying solely on digital currencies to justify its data centres and electricity capacity. It is attempting to turn those assets into the foundations of an AI cloud business, and this $350 million agreement is its clearest step in that direction so far.

Ryanair will use Google’s Gemini and DeepMind models to improve scheduling, maintenance and disruption management as it prepares to carry 300 million passengers annually. Ryanair plans to use Google’s artificial intelligence to help manage everything from crew schedules to aircraft maintenance. Europe’s largest airline by passenger numbers has signed a five-year agreement with Google Cloud. The partnership will bring Google Workspace and cloud services to approximately 35,000 Ryanair employees while expanding the airline’s use of Gemini and other DeepMind technology. Gemini Enterprise will be used to develop specialised AI agents capable of assisting with operational decisions, crew scheduling and disruption management. Ryanair also plans to use DeepMind’s WeatherNext and AlphaEvolve models to support fleet operations and maintenance planning. Airlines operate some of the most complicated logistical networks in business. A delayed aircraft can affect crews, airport slots and passengers across several countries. Even a small improvement in how quickly an airline reorganises its schedule can reduce costs and prevent disruption from spreading. Ryanair expects to carry 300 million passengers annually by 2034, up from roughly 200 million in its latest financial year. Reaching that target will require the airline to manage a larger fleet and more complicated network without allowing operating costs to grow at the same rate. The Google agreement will form part of a dual-cloud strategy. Ryanair already uses Amazon Web Services and intends to retain both providers, reducing its dependence on a single technology platform and lowering the risk that one outage could interrupt its operations. Financial terms of the partnership were not disclosed. Airlines have used algorithms for scheduling and pricing for years, but generative AI could allow their systems to respond more flexibly to unexpected events. Human controllers would still oversee important decisions, while AI handles more of the time-consuming analysis needed to present possible solutions. The technology will need to prove that it can perform reliably in an industry where incorrect information can quickly create expensive problems. Airlines will also need clear controls over which operational decisions AI systems can make independently. For Google, Ryanair provides an opportunity to demonstrate that Gemini can create measurable improvements outside the technology sector. For Ryanair, the goal is more direct: use AI to carry more passengers without making an already complicated airline even harder to run.

MapLight’s experimental schizophrenia drug succeeded in a Phase 2 trial, but its shares fell more than 66% as investors questioned whether the treatment could compete with Bristol Myers Squibb’s Cobenfy. In biotechnology, a successful clinical trial can transform a company overnight. MapLight Therapeutics just discovered that success alone is not always enough. The company reported positive Phase 2 results for ML-007C-MA, its experimental schizophrenia treatment. The twice-daily dose achieved the study’s main goal, showing a meaningful improvement in symptoms compared with a placebo. MapLight’s shares still collapsed by more than 66%. The dramatic reaction was not because the drug failed. Investors were questioning whether it performed well enough to compete with Cobenfy, Bristol Myers Squibb’s already-approved schizophrenia treatment. Patients receiving MapLight’s twice-daily medicine recorded an average 4.5-point improvement over placebo on a widely used scale measuring schizophrenia symptoms. Cobenfy produced improvements of 8.4 and 9.6 points in two pivotal trials before receiving FDA approval. Comparing results from separate clinical trials is never straightforward. The studies involved different patients and conditions, so the numbers cannot be treated as a direct competition. Nevertheless, investors appear to have used Cobenfy as the benchmark, and MapLight’s smaller improvement fell short of expectations. That reaction highlights one of the toughest realities of biotech investing. A drug does not simply need to work. It must offer enough value to compete with treatments already reaching patients. MapLight may still have advantages. Its treatment does not require patients to fast before taking it, while Cobenfy must be taken twice daily on an empty stomach and follows a specific dosing schedule. Side effects in MapLight’s trial were generally mild, and fewer patients discontinued treatment than in Cobenfy’s pivotal studies. Those differences could matter enormously in schizophrenia care. A medicine’s effectiveness outside a clinical trial depends partly on whether patients can tolerate it and follow its dosing instructions consistently. A treatment that produces a smaller improvement under trial conditions could still become commercially valuable if it is easier to use in everyday life. MapLight’s treatment remains at an early stage. The Phase 2 results will need to be confirmed in larger late-stage trials before the company can seek regulatory approval. That leaves considerable uncertainty around its effectiveness, safety and eventual commercial potential. The stock collapse may therefore say as much about investor expectations as it does about the medicine itself. MapLight proved that its drug may help patients. Wall Street wanted proof that it could challenge an established competitor. In biotech, that difference can erase two-thirds of a company’s value in a single morning.

Frontier Airlines' decision to adopt Starlink is about more than faster in-flight Wi-Fi. It highlights how SpaceX's satellite network is steadily evolving into critical infrastructure for the aviation industry and beyond. When most people think of SpaceX, they think of rockets. But one of the company's most valuable businesses may be orbiting the Earth long after each launch is over. This week, Frontier Airlines announced it will equip its fleet with SpaceX's Starlink satellite internet, becoming the first ultra-low-cost U.S. airline to adopt the service. For passengers, it promises faster Wi-Fi capable of streaming videos, browsing the web, and staying connected throughout a flight. For investors, however, the bigger story isn't Frontier. It's Starlink's growing role as critical infrastructure for entire industries. Starlink was originally introduced as a way to bring high-speed internet to rural and underserved communities. Today, its ambitions stretch much further. Airlines, cruise ships, cargo vessels, emergency responders, mining companies, and governments are increasingly turning to the satellite network to deliver reliable internet where traditional infrastructure falls short. The aviation industry has become one of Starlink's fastest-growing opportunities. Reliable in-flight internet has long been a challenge. Older satellite systems often struggled with slow speeds, patchy coverage, and limited bandwidth, particularly as more passengers connected to the network at the same time. As expectations have grown, airlines have been looking for a solution that delivers a more consistent experience. Starlink's low-Earth orbit satellite network is designed to do exactly that. By operating thousands of satellites much closer to Earth than traditional communications satellites, the system can deliver lower latency and significantly faster internet speeds. That benefit extends well beyond passengers streaming movies. Frontier said the new system will also support pilots, flight attendants, maintenance crews, and ground operations by improving communication and access to operational data. In other words, Starlink is becoming part of the airline's digital infrastructure, helping improve efficiency as well as the passenger experience. For years, airlines largely competed on ticket prices, especially in the ultra-low-cost segment. Today, customer expectations have changed. Travelers increasingly expect to stay connected wherever they are, whether they're answering work emails, joining a video call, or simply messaging friends and family during a flight. As a result, high-speed connectivity is evolving from a premium feature into a competitive necessity. For SpaceX, each new airline agreement strengthens a much larger business strategy. Every aircraft equipped with Starlink expands the network's commercial footprint and demonstrates that the technology can serve industries far beyond residential broadband. What began as a satellite internet service is steadily becoming a communications platform supporting transportation, logistics, enterprise operations, and public services around the world. Investors often associate SpaceX with ambitious missions to the Moon and Mars. Yet one of the company's most significant growth stories may be unfolding much closer to home. Starlink is no longer simply connecting remote communities. It is quietly becoming part of the infrastructure that keeps modern industries connected.

The U.S. has committed $4.84 million to support a proposed $150 million rare-earth project in Madagascar. The early funding could help unlock a new supply chain for manufacturers across America and Europe. A relatively small U.S. investment could help unlock a much larger rare-earth project in Madagascar. The U.S. International Development Finance Corporation has committed up to $4.84 million to Harena Rare Earths’ Ampasindava project. The funding will support early work including laboratory testing, environmental programmes and the operation of a pilot plant. Compared with the project’s estimated $150 million cost, the initial commitment is modest. Its importance lies in what could follow. If the early development work is successful, Harena may become eligible for substantially larger financing as the project moves towards construction. That could help create a new source of rare earths for manufacturers in the U.S. and Europe. Rare earths are a group of specialised materials used in permanent magnets for electric vehicles, electronics, wind turbines and advanced industrial equipment. Demand is rising, but building a reliable supply chain requires far more than opening a mine. The materials must also be separated, refined and converted into components manufacturers can use. Ampasindava could provide an important starting point. The project contains neodymium, praseodymium, dysprosium and terbium, four valuable rare earths used in high-performance magnets. Harena expects it could eventually produce around 4,000 tonnes of rare-earth oxides annually, including approximately 1,700 tonnes of high-value magnet materials. Production is targeted for mid-2028, although several obstacles remain. Harena still needs an exploitation permit, full construction financing and agreements covering where its output will be processed. The company is considering potential partnerships with U.S. and European groups, including MP Materials, USA Rare Earth and Solvay. Those relationships could prove just as important as the mine itself. Extracting rare earths in Madagascar would create a new source of supply, but processing them elsewhere would be necessary to build a complete chain connecting the deposit with Western manufacturers. For investors, the project shows how government support can help reduce the risks surrounding early-stage mining developments. Public funding for testing and environmental work can make it easier for private capital to assess whether a project is technically and commercially viable. However, the $4.84 million commitment should not be mistaken for a guarantee. Mining projects frequently face permitting delays, rising construction costs and difficulties securing long-term buyers. Ampasindava must still prove that it can move from a promising deposit to a profitable operation. The opportunity is nevertheless significant. If Harena secures the necessary approvals, partners and financing, a relatively small initial commitment could help establish a new rare-earth supply route stretching from Madagascar to processing facilities and manufacturers across the U.S. and Europe. The mine itself may be located in Africa, but its potential value lies in the wider supply chain it could help create.