Tesla CEO Elon Musk (L) and Apple CEO Tim Cook are both worried about the same thing. Getty Images Tim Cook and Elon Musk have found common ground. Cook has called the memory shortage "a hundred-year flood" and has raised prices on Apple products. Musk agreed. He said memory chips are the "biggest price jump in anything I've ever seen." Tim Cook isn't the only CEO shocked by the soaring cost of memory. Elon Musk said he's never seen anything like it. The Tesla CEO posted a quote on X from the Apple CEO, who earlier this month had said the memory shortage is "a hundred-year flood" and that he'd "never seen anything like it in any area in over 40 years." Musk said he agreed with Cook. He wrote on Thursday that it's the "biggest price jump in anything I've ever seen." The soaring demand for chips, which are widely used in consumer electronics, and the resulting costs, have eaten into company margins, leading many to raise prices. The heart of the issue is the AI boom, which has strained memory chip production as deep-pocketed hyperscalers and data center buildouts increase demand. That's led to skyrocketing memory costs and lower supply for many consumer electronics companies, which have begun passing those added costs onto shoppers with price hikes. Apple took the rare step of raising the prices of MacBooks, iMacs, iPads, HomePods, and the Apple TV box on Thursday, boosting some by as much as $300. Microsoft's Xbox did the same. It announced Xbox game consoles would be $100-$150 more expensive beginning August 1. So what's the solution? Musk stated the obvious. "The production shortfall relative to demand is insane. MUCH higher production is needed," he wrote on X. The Tesla CEO has earlier talked about the potential impact of the memory chip shortage in the context of needing to ensure his company has the supply it needs to build out its AI ambitions. During Tesla's Q4 earnings call in January, Musk said Tesla could become "limited by supplier output of chips" and added that "memory is an even bigger limiter than AI logic." Partially to address this concern, Musk earlier announced the Tesla Terafab joint project with SpaceX and Intel, a multibillion-dollar initiative to build out a chip production facility "combining logic, memory and advanced packaging under one roof." Read the original article on Business Insider

An AWS data center Noah Berger/Getty Images via Amazon Web Services AWS is raising key AI cloud prices by 20%, after a 15% price increase in January. Soaring memory chip costs and strong AI demand is driving higher pricing for cloud customers. Memory shortages have sent Micron and SK Hynix surging to records. Amazon raised prices for several key AI cloud offerings, the latest sign that memory chip shortages are driving up the cost of some technology. Amazon Web Services recently announced price increases for EC2 Capacity Blocks for ML. This is a cloud service that lets companies reserve GPUs in advance. The changes mean hourly rates for renting several types of cloud servers will jump by roughly 20% starting in July. AWS had already raised prices for the same service by about 15% in January. "Amazon EC2 Capacity Blocks for ML reservation prices are updated periodically based on supply and demand," the company said in its announcement. Amazon didn't immediately respond to a request for comment on Friday. Similar price increases are happening in other parts of the tech industry, as tech giants pass memory price pressure on to customers. Apple raised prices this week, blaming soaring memory chip costs. Xbox did the same, and Elon Musk complained about unprecedented memory price increases. The AWS move is more consequential than your next MacBook or gaming console costing a couple of hundred dollars more. As the world's largest cloud provider, AWS underpins many software services, and millions of developers rely on the cloud service to offer apps and other tech products. Price increases of 15% and now 20% will likely ripple through these sectors in coming quarters. The price hikes reflect a broader shift in tech: AI is increasingly constrained by physical limitations, rather than software availability. Tight memory chip supply and strong GPU demand are raising costs for cloud providers. One of the biggest physical constraints right now is high-bandwidth memory, a critical component packaged alongside advanced AI chips. AI cloud services run on these chips and servers, so shortages and price increases like this have a big impact on data center expansion plans and, ultimately, the supply of AI. "As there is a limit to how much memory can be produced, then there is a limit to how many GPUs can be produced, which means that there's a limit to how many data centers can be built," Peter Berezin, chief economist at BCA Research, wrote on X on Friday. Berezin added that cloud providers can pass on higher infrastructure costs because customers have few alternatives when GPU capacity is scarce, giving hyperscalers AWS, Microsoft, Google, and Oracle greater pricing power. "While the memory shortage raises their costs, it also keeps the demand for compute above the available supply, which gives them greater pricing power over access to cloud computing," Berezin wrote on X. The same shortages pushing up AI cloud prices have propelled memory-chip makers such as Micron and SK Hynix to records, reflecting investor expectations that AI-driven demand will keep the market tight, and prices high, for years. Have a tip? Contact this reporter via email at ekim@businessinsider.com or Signal, Telegram, or WhatsApp at 650-942-3061. Use a personal email address, a nonwork WiFi network, and a nonwork device; here's our guide to sharing information securely. Read the original article on Business Insider

<blockquote cite="https://openai.com/index/previewing-gpt-5-6-sol/"><p>We're beginning a limited preview of the GPT‑5.6 series: Sol, our flagship model; Terra, a balanced model for everyday work; and Luna, a fast and affordable model. Terra has competitive performance to GPT‑5.5 while being 2x cheaper and Luna brings strong capability at our lowest cost. [...]</p> <p>We believe in broad access, and we plan to make GPT‑5.6 Sol, Terra, and Luna generally available in the coming weeks. As part of our ongoing engagement with the U.S. government, we previewed our plans and the models’ capabilities ahead of today’s launch. At their request, we are starting with a limited preview for a small group of trusted partners whose participation has been shared with the government, before releasing more broadly. [...]</p> <p>GPT‑5.6 is priced per 1M tokens across three model sizes: Sol is $5 input / $30 output; Terra is $2.50 input / $15 output; and Luna is $1 input / $6 output. GPT‑5.6 also introduces more predictable prompt caching, including support for explicit cache breakpoints and a 30-minute minimum cache life. For GPT‑5.6 and later models, cache writes are billed at 1.25x the model’s uncached input rate, while cache reads continue to receive the 90% cached-input discount.</p></blockquote> <p class="cite">&mdash; <a href="https://openai.com/index/previewing-gpt-5-6-sol/">OpenAI</a>, Previewing GPT‑5.6 Sol: a next-generation model</p> <p>Tags: <a href="https://simonwillison.net/tags/gpt">gpt</a>, <a href="https://simonwillison.net/tags/generative-ai">generative-ai</a>, <a href="https://simonwillison.net/tags/ai-security-research">ai-security-research</a>, <a href="https://simonwillison.net/tags/openai">openai</a>, <a href="https://simonwillison.net/tags/llms">llms</a>, <a href="https://simonwillison.net/tags/llm-release">llm-release</a>, <a href="https://simonwillison.net/tags/llm-pricing">llm-pricing</a></p>

OpenAI’s Jalapeño chip signals a deeper push into AI infrastructure, but cost savings and independence from Nvidia still depend on scale. The post OpenAI’s New Custom Chip: 5 Things You Should Know appeared first on TechRepublic.

The White House asked OpenAI to delay the rollout of its GPT-5.6 AI models, two weeks after Anthropic had to take its most advanced AI models offline.

This week, we discuss talking aloud to computers, Cannes, and “Engineering Creativity: Guac Is Extra."

AI can enable drones to fly longer since they require less onboard computing, said Brendan Englot, the director of the Stevens Institute for Artificial Intelligence. Corvus Robotics GNC uses Corvus Robotics drones to manage warehouse inventory in Indiana. AI-enhanced drones at GNC provide precise inventory insights, helping to reduce backorders. Drone use at GNC frees up staff for investigative roles, improving job satisfaction and retention. At a GNC warehouse in Whitestown, Indiana, workers see and hear four drones overhead every day as they complete tasks that few humans want to do: count and track full boxes of products across more than 2,000 pallets in the warehouse's aisles. The drones' regular flights allow GNC, a nutritional manufacturing company, to obtain a full view of its 250,000-square-foot reserve inventory once a month rather than once a quarter, said Bill Monk, GNC's vice president of distribution. More frequent monitoring helps the company achieve its goal of reducing backorders, which occur when its inventory system records items as in stock even though they're missing from shelves, or from the warehouse altogether. "It helps us cycle it faster. If you don't know where it is, you can't ship it," Monk told Business Insider. GNC started using the drones, which come from Corvus Robotics, two years ago. Since then, nonshipments have plummeted from several hundred units a day to around 98, Monk said. Now, GNC's inventory staffers have more time to investigate and follow up on discrepancies by validating and correcting problems flagged by the drone system. AI helps with drone positioning and insights GNC's inventory staff works with Corvus to schedule the drones so they cover specific areas during their 25- to 30-minute flights, said Jackie Wu, the founder and CEO of Corvus Robotics. "There's a lot of physical AI on board the vehicle that enables us to collect the data and move autonomously in these environments," Wu told Business Insider. "The AI is helping the customers with what to do with their inventory data, what to do with the discrepancies." Specifically, AI helps the drones get accurate positioning so they can move safely and reliably. Drone images and videos also help staff find misplaced products, pallets half-off the rack, and fallen boxes. Generally, drones can be networked to access the internet and remotely query large AI models, said Brendan Englot, the director of the Stevens Institute for Artificial Intelligence and a professor at Stevens Institute of Technology. He said that networking capabilities enable drones to operate with less onboard computing so they can fly longer before needing a recharge. And, since warehouses have stable lighting conditions and few or no windows, drones' computer vision algorithms can be tuned to a single environment, which is ideal for performance, Englot said. AI also allows non-experts to more easily ask questions or provide drone instructions in plain English, Englot said. "The language element of these models bridges the gap between an expert robotics engineer and an employee in the warehouse," he said. Drones reduce operational issues at GNC Before using the drones, GNC's inventory staff included 20 people, and it was difficult to find staff willing to handle the monotonous work and take on the undesirable shifts required for manual inventory counts, said Monk. Employee turnover has improved since introducing the drones, Monk said. Workers are happy to have the drones do inventory, as "it is not a very fun job and it takes a lot of pallets counted to find a needle in a haystack," Monk said. Tammy Lacher, a senior specialist in inventory control at GNC, agreed. "The drones handle the counting, and our team gets to do more investigative work. It takes a lot of the grind out of the job," Lacher said. "It's pretty cool to be working with this kind of technology every day." With the adoption of drones, most of GNC's original inventory team members were reassigned to different roles, like customer service or inventory accuracy in active product-picking areas, or left through attrition, said Monk. Now, one to two inventory team members focus on the reserve inventory area and regularly verify the accuracy of the drones' data. "We audit 2% of what the drones said was right, to make sure," Monk said. Once, a drone showed a location with 600 boxes, contradicting GNC data that documented 60 boxes. "Everybody thought the drone was wrong," Monk said, but it was not. A staff member had input the wrong box size, eliminating a zero at the end of a measurement. Drones have some limitations Drones have a few limitations. For GNC, one is their plastic-wrapped pallets, said Monk. GNC's reserve inventory is in 70" aisles, a narrow space for drones that are two-thirds of that width and have moving rotors. The drone sensors can see through plastic for inventory counting, but when plastic is cut and flapping, the blades can get caught, sending the drone down, he said. To get around this, staffers are trained to cut and remove plastic when removing boxes from a pallet. Workers typically scan the aisle for issues before the scheduled drone flies through, said Monk. He added that GNC's accuracy has increased significantly, but the company only uses drones for reserve inventory, which typically includes unopened boxes. The drones cannot look back to see into open boxes that have been picked through, but the software can provide case-level estimating, Wu said. Still, the drones have helped GNC in ways Monk didn't anticipate. "When you're not finding a problem for a quarter or six months, you spend a lot of time trying to solve it. The drone expedites that problem," he said. Read the original article on Business Insider

Ukraine has launched a new portal where soldiers, arms makers, and allies can access data from Russian military hardware. AP Photo Ukraine launched a new portal, "TrophyLab," to expose Russian and North Korean military secrets. The portal includes detailed information about weapons like drones, missiles, and armored vehicles. The goal is to increase critical information sharing between Ukraine and its Western partners. Ukrainian soldiers, defense companies, and foreign governments are rushing to sign up for a new portal where Kyiv is offering access to captured Russian military hardware — and what's inside. Yurii Myronenko, the inspector general for Ukraine's defense ministry, told Business Insider on Friday that 150 users have already registered for the new "TrophyLab" portal in its first week. Roughly a third of the requests came from abroad. This portal contains technical data from Russian weapons like drones, missiles, and armored vehicles, which approved users can access in a first-of-its-kind information-sharing arrangement between Ukraine and its Western partners. Myronenko, a key figure behind TrophyLab's development, described this project in an interview as an "unprecedented" effort to expose not only Russian military secrets, but also those collected from North Korean weapons. The hope among Ukrainian officials is that by consolidating this data and making it public, Kyiv and its allies will learn how to stop the Russian war machine — now and in the future. "The better we understand the enemy's equipment and weapons, the faster we can develop effective countermeasures," Myronenko said, speaking through a translator. A win-win for Ukraine and its allies A Ukrainian official inspects Russian weaponry in Kharkiv. Viacheslav Madiievskyi/Ukrinform via Reuters Connect Since the start of Russia's full-scale invasion in February 2022, Kyiv has regularly shared intelligence with the US, NATO, and its Western backers, including information about the weapons it faced on the battlefield. However, Myronenko said that "such decisions were made on a case-by-case basis by the highest military and political leadership, and it was made within the framework of existing agreements between the two countries." During the second half of last year, Ukraine decided to digitize data it had gathered from captured Russian weaponry so it could consolidate information from different units and enable rapid access to it. Ukraine wanted the platform to be available across its defense ecosystem, from research organizations, military units, and government agencies to defense-tech manufacturers, as well as foreign governments, arms makers, and laboratories, including those in NATO countries. "We are obliged to share this information with the Ukrainian defense manufacturers, Myronenko said. Kyiv procures a lot of local products, especially drones and electronic warfare systems. It's a win-win for Ukraine to share the information abroad, he said. Ukraine has a well-developed research apparatus capable of studying the captured hardware, but some foreign countries have advanced labs with a high level of specialization in areas such as optical equipment or satellite navigation. Ukraine's partners benefit from the information on Russian weapons, and Kyiv benefits from their research. A Russian reconnaissance drone on the TrophyLab site. Courtesy of the Ministry of Defense of Ukraine The portal was complicated and involved months of work, including legal considerations, organization and coordination among military units, and technical development. Ukraine's defense ministry officially launched TrophyLab on June 19. "We are opening access to captured Russian weapon technologies for our global partners. Every missile, drone, and vehicle seized on the battlefield is now a source of knowledge for the free world," Mykhailo Fedorov, Ukraine's defense minister, said in a statement. "What was meant to be the enemy's secret advantage is being dismantled to defend democracy," he added. Missiles, drones, and armored vehicles. The TrophyLab portal features more than 150 samples of captured weapons and military equipment across roughly 80 distinct categories, as well as over 225 completed research projects. And that is all expected to grow. The weapons in the portal include Russian ballistic and cruise missiles, one-way attack and reconnaissance drones, air defense systems, small arms, armored vehicles, artillery pieces, and electronic equipment, among other weaponry. Notably, the catalog of weapons also includes North Korea's KN-23, a short-range ballistic missile that Russia has used throughout the war, according to screenshots of the TrophyLab shared with Business Insider. It's unclear whether the portal contains other foreign weaponry. A catalog of Russian missiles and drones, including the North Korean KN-23. Courtesy of the Ministry of Defense of Ukraine "We are going to feature any kind of weapons or military equipment that is used against Ukrainian citizens," Myronenko said. "So all the weapons that are used against Ukrainians on the battlefield, they will be featured on this TrophyLab platform." In its first week, Ukraine's defense ministry has already received 150 registration requests for TrophyLab. Myronenko said that these include Ukrainian defense companies and military units, as well as foreign companies and governments. Foreign users make up 30% of the pool, and Kyiv expects the number to grow over time. Once users are vetted and approved, they will have round-the-clock access to the platform. They can also request physical samples of weaponry for study and even destroy them to obtain additional data. TrophyLab marks Ukraine's latest effort to share battlefield lessons and experience with its Western allies, as it has done extensively with drone warfare. Myronenko believes that blowing the lid on Russia's arsenal will benefit all involved. "It will be a great experience, both for Ukraine and the partners," he said. Read the original article on Business Insider

The Open Source Sustainability Initiative's goal is to help enterprises manage and secure aging open source projects while maintaining regulatory compliance.

Using Gemma 4, Ollama, OpenAI Agents SDK, and Tavily MCP to build a lightweight research agent The post From Local LLM to Tool-Using Agent appeared first on Towards Data Science.

President Donald Trump's administration confirms the earliest SAVE student-loan borrowers will have to find a new plan. Aaron Schwartz / AFP via Getty Images The Education Department said that SAVE borrowers won't have to find a new plan until September 29 at the earliest. It also said in a legal filing that most borrowers will have "more time than that" because the transition will occur in waves. It follows Trump's elimination of the SAVE student-loan repayment plan in March. The Trump administration confirmed the earliest date by which millions of student-loan borrowers will have to start switching repayment plans. In a legal filing on Thursday, the Department of Education said that borrowers will not be required to move off the SAVE plan until September 29 "at the earliest." Additionally, the department said that since it is transitioning borrowers off the plan in waves, most borrowers "will get even more time than that." Previously, the department said that starting July 1, borrowers would begin receiving notices of their 90-day timeline to transition to a new plan, and that if they did not take action, they would automatically be placed in the most expensive plan. This update is in response to a lawsuit filed in March that challenged the department's decision to eliminate the SAVE plan, a Biden-era plan that gave borrowers more affordable monthly payments and a shorter timeline to debt relief. Public Goods Practice, which represents the borrowers in the lawsuit, filed a motion on Tuesday calling for the court to pause that forced transition to a more expensive plan while a court decides the outcome of the rest of the case. The new legal filing comes just days before Trump's sweeping student-loan repayment overhaul takes effect on July 1, with new repayment plans and borrowing caps. Borrowers said they expect their monthly payments to increase under the new plans, and some have reported glitches with their accounts, including inaccurate payment projections. Have a story to share about student loans? Contact this reporter at asheffey@businessinsider.com, or fill out this form. Read the original article on Business Insider

My family had a lot of fun spending time in Argentina — mostly Buenos Aires — but we made a few mistakes before and during our trip. Jamie Davis Smith My family of five spent six days in Argentina, mostly spending our time in and around Buenos Aires. I wish we'd brought Argentine pesos and slept in so we could stay up later to enjoy local nightlife. We should've planned indoor activities for especially hot days and utilized public transit sooner. Over the years, I've spent a lot of time in South America. However, I had never made it as far south as Argentina until recently. In December, my husband and three of our kids spent six days in and around Argentina's capital, Buenos Aires, while en route to Antarctica. We loved exploring Buenos Aires, which is nicknamed "Paris of the South" because of the classic French-inspired architecture found throughout the city's wide, tree-lined cobblestone streets, and a café culture that rivals that of Paris. I especially loved feasting on the rich local cuisine, exploring its robust soccer culture, heading to the countryside just outside the city limits to become a gaucho (cowboy) for the day on a traditional Argentine farm, and exploring nearby canals reminiscent of Venice. However, my family's six-day trip could have been better if we had avoided a few mistakes. Taking advantage of public transportation earlier would've been helpful Using public transportation got us around the city faster. Jamie Davis Smith Before my trip, a friend who was born and raised in Buenos Aires cautioned me against taking the subway there. She said that the Buenos Aires Underground, or Subte, would be too hot to be tolerable during my summer visit because the stations and several train cars are not air-conditioned. She also warned that there was a high risk of being pickpocketed in busier stations. At first, I took her advice and relied on Uber to get around the city. However, taking rideshares everywhere quickly became expensive, and we often had to wait a while for cars to pick us up. About midway through the trip. I decided to give the Subte a try in hopes it'd get us around the city faster. The stations and many train cars were warm, but not as sweltering as we worried they'd be. I also used the same street smarts on the Subte that I use on the metro at home in Washington, DC, such as keeping my bag in front of me at all times, and I never felt unsafe. Ultimately, I found Buenos Aires's subway system, the oldest in South America, to be inexpensive, surprisingly modern, and easy to use. We would've saved a lot of time, money, and aggravation if we had taken a chance on the Subte starting on day one of our trip. Arriving with cash would've saved us some headaches I wish I'd had Argentine pesos on hand throughout my trip. Jamie Davis Smith Before I travel abroad, I always check whether credit cards are widely accepted wherever I'm going. In many countries, I never need cash at all. Based on a quick online search, I didn't anticipate any problems using my credit cards on our trip to Buenos Aires. However, at my family's first dinner in Buenos Aires — a scrumptious Argentine-style pizza restaurant — I realized I was wrong. When the bill came, I found out the hard way that many restaurants here don't accept credit cards. I desperately needed Argentine pesos and had to leave my husband and children at our table and go out in search of an ATM. Additionally, I had a hard time getting Ubers until a local told me that most drivers prefer to be paid in cash. To increase my chances of getting picked up, I chose the option to pay in cash in the app (an option I don't remember seeing anywhere else I've traveled). It seemed to help, but meant I needed even more Argentine pesos. To add insult to injury, the ATMs I found charged high fees and limited the amount of cash I could withdraw each day. Arriving with a wallet full of Argentine pesos would have saved me time, aggravation, and money. I wish I'd planned better to experience the local late-night culture In Argentina, it's common to eat dinner late. Jamie Davis Smith According to traditional travel wisdom, it's almost always a good idea to get up and go out early to beat the heat and visit popular attractions before crowds swell. Although I don't always follow that advice myself, I've generally found it to hold true in many places I've visited. In Buenos Aires, it turns out, getting up early wasn't always the best move. When we went out to eat at 8 p.m, the streets in some neighborhoods felt dead. Restaurants were often empty at that time, even on weeknights. Some parts of the city felt quiet during the day and most alive at night. Jamie Davis Smith Before arriving, I knew that locals like to eat dinner later in the day and stay up late. However, I didn't fully realize just how late. I discovered that it's not unusual for people in Buenos Aires to wait until 10 p.m. or later to have dinner. I wish I had slept in a day or two so I had the energy to stay out later and fully experience Buenos Aires' culture and nightlife. Planning more indoor activities would've helped us beat the heat We loved spending time outdoors in Buenos Aires, but some days got really hot. Jamie Davis Smith My family likes to be outdoors, and Buenos Aires has plenty of parks and colorful neighborhoods to explore. However, during our summer visit, several days were very hot, reaching over 90 degrees Fahrenheit. I wish I had planned some indoor activities as a backup plan for the hottest days. For example, my family could've explored the Evita Museum, the National Museum of Fine Arts, or the Museum of Latin American Art of Buenos Aires. We also could've booked cultural experiences, such as tango performances, that would have allowed us to cool down indoors for a few hours and see a different side of Argentine culture. Read the original article on Business Insider

As context windows grow longer, moving large model weights efficiently becomes critical to performance. A common way to address this is quantization, an... As context windows grow longer, moving large model weights efficiently becomes critical to performance. A common way to address this is quantization, an optimization technique that compresses model weights into a smaller data format. One quantization format is NVFP4, an innovative 4-bit floating point introduced with NVIDIA Blackwell architecture. That’s the approach behind our new Nemotron 3… Source

The 2026 FIFA World Cup is being played on some of the most heavily researched grass fields ever created. FIFA has spent millions of dollars working with turf scientists to engineer 16 stadium pitches across three countries and vastly different climates. We visited the University of Tennessee to see how robotic cleats, ball-launching machines, and stadium simulators helped prepare fields for the world's biggest sporting event. And we found out how World Cup grass is grown, tested, transported, and maintained before billions of viewers tune in. Read the original article on Business Insider

Instead of eliminating jobs for early-career cyber pros, AI is creating new opportunities for candidates with strong human decision-making skills.

Investment, safety and next-generation AI models suggest the conversation is shifting from robot demonstrations to real-world deployment.

With falling sales in the US and especially China, VW Group wants to restructure.

Apple raised prices on Macs, iPads, and other devices as AI data center demand drives up memory chip costs across consumer electronics. The post Apple Raises Prices on Macs, iPads as Data Centers Drive Memory Shortage appeared first on TechRepublic.

"The current push for AI adoption that we're seeing is directly coming from the financial incentives of AI firms," she added. Because of the massive capital expenditures, the hyperscalers and other AI firms are making a "deliberate push for AI everywhere — no matter whether the demand is there or if customers want it or not." The post Big Tech is spending trillions on AI. Investors now want proof it will pay off. appeared first on AI Now Institute.

Why memorizing for the exam doesn't mean you understand the subject The post Water Cooler Small Talk, Ep. 11: Overfitting in RAG evaluation appeared first on Towards Data Science.