Latest news with #GPT-5
Yahoo
2 days ago
- Business
- Yahoo
Nvidia's chips are among the world's hottest commodities. So why is the company likely trashing $4.5 billion worth of them?
Nvidia's blockbuster quarterly earnings last week came with a big dose of negative news: A $4.5 billion write down on chips originally destined for customers in China that will ultimately go undelivered. The company had developed the so-called H20 chips, which are less powerful than its top-of-the-line semiconductors, to comply with Biden Administration regulations against sending technology to foreign adversaries that could help their AI efforts. The Trump Administration, however, went a step further in early April and banned exports of even second-tier chips. 'We are taking a multibillion dollar write off on inventory that cannot be sold or repurposed,' said Nvidia CEO Jensen Huang on a May 28 earnings call. The details around the decision to write the value of the chips to zero—rather than sell them to other customers—wasn't explained. The company added, without elaborating, that it's 'exploring limited options' for the unused inventory, a move that would apparently fall short of selling the chips. The H20 was built specifically for the China market under the old export rules. Because of the chip's design and limited capabilities, it may be difficult to use in other countries. 'It doesn't really fit anywhere else without a lot of expensive tweaking,' says Arash Azadegan, a professor of supply chain management at Rutgers Business School. This tweaking could involve additional costs for Nvidia. Furthermore, some of the chips may 'not meet the performance needs of customers in other regions' or may be engineered 'specific to Chinese customer requests or requirements,' says Chad Autry, a University of Tennessee supply chain professor. Even if it could sell the chips by cutting their price, Nvidia would risk damaging its image as a seller of top-tier innovation. The company's new Blackwell GPUs, after all, power cutting-edge AI like OpenAI's yet-to-be-released GPT-5 model. 'Nvidia probably doesn't want to flood the market with discounted chips—it could mess with their pricing, confuse customers, and distract from their big push into the newer Blackwell lineup,' says Azadegan, referencing Blackwell's used in new products from Amazon's AWS, Microsoft Azure, Google Cloud, and Oracle, among others. Nvidia declined to comment beyond what its executives said during the company's earnings call last week. After China-based DeepSeek released its ChatGPT rival in January, major Chinese firms like Alibaba, ByteDance, and Tencent purchased Nvidia's H20 chips, Reuters reported in February. These companies have likely developed powerful AI capabilities using H20 chips, so there's little reason to believe that U.S. companies could not do the same if Nvidia sold H20 to them. But Nvidia and U.S.-based customers probably have no interest in this, says Matthew Bryson, a Wedbush Securities managing director who covers Nvidia. 'The reason that these products are going to China is because everyone would choose something better.' Bryson explains that the technology underpinning H20s 'would never have been made' if not intentionally engineered to prevent Chinese firms from building models similar to U.S.-made advanced AI applications. And if Nvidia sold the chips at a big discount to reflect their U.S. market value, it could 'cannibalize' sales of better products, he says. Nvidia had initially expected to have to write off $5.5 billion of chip inventory from its balance sheet. But it managed to salvage about $1 billion of H20 inventory by '[reusing] certain materials,' reducing the realized first-quarter writeoff to $4.5 billion, chief financial officer Colette Kress said on last week's earnings call. Unsurprisingly, she noted that as a result of the Trump Administration's new export rule, Nvidia will take a revenue hit from Trump's China ban in the first half of the year. The company booked a $2.5 billion loss in first quarter revenue and predicted a steeper $8 billion revenue loss in the second one. The question remains of what happens with this H20 stockpile. Supply chain experts interviewed by Fortune suspect it will be discarded. 'These chips will meet the 'cool lava lamp' you got from your aunt in a landfill somewhere,' Alan Amling, a professor of supply chain at the University of Tennessee's Haslam College of Business, wrote to Fortune via email. 'With so many other growth opportunities, the opportunity cost of repurposing, retesting and requalifying these chips was obviously too high a bar.' Because of how quickly the new export rules under Trump took effect, the undeliverable H20 chips are likely sitting in U.S.-friendly places like Taiwan and Hong Kong, said University of Tennessee's Autry. Taiwan-based TSMC manufactures much of Nvidia's inventory meant for China-based customers, and Nvidia's fulfillment partners are in nearby countries. Writing off the H20 chips had minimal short-term impact on Nvidia's otherwise booming business. Nvidia reported first quarter revenue of $44.1 billion (up 69% year-over-year) and $18.8 billion in net income, representing a healthy 42.6% net profit margin. Writing down the full value of the chips provides an immediate tax benefit by reducing the company's taxable income. Depending on Nvidia's future sales prospects and costs, this benefit could outweigh or come close to the financial benefit of selling the chips at a steep discount. 'Nvidia is going to be alright; investors of Nvidia are going to be alright,' says Rutgers' Azadegan. 'The real story,' he said, is that Nvidia's manufacturing partners like TSMC and suppliers, including Samsung and Micron, could be most impacted by the H20 inventory writeoff because they've built businesses by serving critical functions in Nvidia's supply chain. The magnitude of the impact to these partners will be determined in the months and years ahead, notes Azadegan. 'It's never on a dime that we can pivot.' Nvidia analyst Harsh Kumar of Piper Sandler is hopeful that Trump eventually undoes the blockade on H20 chips to China. If so, Nvidia could finally complete the H20 sales or deliver the chips to new customers in China, assuming the company holds onto the H20 inventory. It's unclear how realistic this scenario is after the Trump administration appealed a judge-imposed block on his 'Liberation Day' tariffs. Still, Trump's new export ban prevents Nvidia from delivering chips to China for national security reasons, so it may have little to no correlation with his tariff agenda. 'To me, it almost implied that there is a pathway that Jensen [Huang] and Nvidia see for this H20 chip to be sold back into China,' Kumar told Fortune, potentially helping to propel the company's already high-flying stock to greater heights in the future. This story was originally featured on Sign in to access your portfolio
Yahoo
27-05-2025
- Business
- Yahoo
One App With All of the AI Models You Actually Want to Use
The following content is brought to you by PCMag partners. If you buy a product featured here, we may earn an affiliate commission or other compensation. If you rely on AI for work, then you've likely noticed that many of the free tools are either inefficient or extremely limited in scope. Even OpenAI puts a cap on how much you can use GPT-4, and their paid subscriptions aren't cheap. The alternative is to get access to the same AI models through a different platform. That's what does. This all-in-one AI platform gives you access to GPT-5 and GPT-4 Turbo, Gemini, Claude, and more. It's also only $29.97 (reg. $234). puts all the AI tools you rely on into one platform. You can generate copy with GPT, craft images with Midjourney, and there's even AI for editing audio and video It works on a credit system, but the lifetime subscription gives you more than enough. With your 1,000,000 monthly credits, you could generate over 800,000 words, research 1,933 SEO keywords, upscale 241 images, convert 120,833 characters to speech, or transcribe up to 4,833 seconds of audio. These credits roll over every month if you don't use them all. And just by logging in daily, you can earn up to 450,000 additional credits every month. Don't rent tools you can own. Get a subscription to while it's on sale for only $29.97. Prices subject to change. PCMag editors select and review products independently. If you buy through StackSocial affiliate links, we may earn commissions, which help support our testing.
India Today
20-05-2025
- India Today
AI godfather Geoffrey Hinton who warned against supersmart AI says he trusts ChatGPT more than he should
Geoffrey Hinton, often described as the godfather of AI, has made headlines once again, but this time not for sounding the alarm, but for admitting a surprising degree of trust in the very technology he has warned the world about. In a recent CBS interview, Hinton revealed that he relies on OpenAI's GPT-4 model for his daily tasks and confessed, 'I tend to believe what it says, even though I should probably be suspicious.'advertisementThis statement stands out coming from a man who has spent years cautioning the world about the dangers of artificial intelligence. Hinton, who received the 2024 Nobel Prize in Physics for his groundbreaking work in neural networks, has previously warned that superintelligent AI could eventually manipulate humans or become uncontrollable. Yet, when asked which tool he uses most, his answer was simple: the interview, Hinton tested GPT-4 with a riddle: 'Sally has three brothers. Each of her brothers has two sisters. How many sisters does Sally have?' The correct answer is one because Sally is one of the two sisters. But GPT-4 got it wrong. 'It surprises me. It surprises me it still screws up on that,' Hinton said. Despite the error, he acknowledged he still finds himself trusting its answers more than he asked during the interview if he thinks the next version, GPT-5, would get it right, he said, 'Yeah, I suspect.'While Hinton may be right to expect improvements, he's not alone in his fascination. Since GPT-4 was introduced by OpenAI in March 2023, it has been widely adopted across industries for its ability to write code, summarise documents, and solve complex problems. However, OpenAI officially discontinued GPT-4 as a standalone product earlier this month. The latest models OpenAI has are GPT-4o and GPT-4.1, which are newer and more advanced models that boast faster response times, lower costs, and a broader range of capabilities including real-time audio and visual input. For Hinton, though, this trust in AI is layered with caution. He has long expressed concerns about the potential misuse of AI. From warning against AI spreading misinformation to posing a threat to humanity if machines ever surpass human intelligence, Hinton has always asked scientists and consumers to be vary of the technology. Hinton worked at Google's AI division for a decade. In 2023, he resigned to speak more openly about these dangers. He has warned of systems so persuasive that they could influence public opinion or deceive users, especially once AI starts understanding the world better than we do. advertisementAnd yet, there's a human side to Hinton that often goes overlooked, one that celebrates curiosity and mentorship. During a press conference after his Nobel win, he proudly acknowledged his students, especially Ilya Sutskever, co-founder of OpenAI. With a touch of dry humour, he said, 'I'm particularly proud of the fact that one of my students fired Sam Altman,' referring to the high-drama leadership crisis at OpenAI in his outlook on AI hasn't softened. In several public remarks, Hinton has compared AI's rise to the industrial revolution. Only this time, it's about intellectual rather than physical power. 'We have no experience in having things which are smarter than us,' he warned. He has advocated for responsible development and governance frameworks, saying the benefits of AI in fields like healthcare and climate science could be immense, but only if managed wisely.
Time of India
19-05-2025
- Time of India
One question that Geoffrey Hinton, the Godfather of AI says OpenAI's GPT-4 still gets wrong
Geoffrey Hinton , the Godfather of AI has admitted that he sometimes trusts his favorite chatbot more than he should. In a recent interview with CBS, Hinton revealed that he uses OpenAI 's GPT-4 and he tends to believe what GPT-4 tells him, even though he knows he should be more careful. 'I tend to believe what it says, even though I should probably be suspicious,' he said. Sharing an example from his own experience, he said that he once asked GPT-4 a simple riddle: "Sally has three brothers. Each of her brothers has two sisters. How many sisters does Sally have?" The correct answer is one as Sally herself is one of the two sisters, and each brother has the same two sisters. But GPT-4 answered two, which surprised Hinton. "It surprises me. It surprises me it still screws up on that," he said during the interview. by Taboola by Taboola Sponsored Links Sponsored Links Promoted Links Promoted Links You May Like Trading CFD dengan Teknologi dan Kecepatan Lebih Baik IC Markets Mendaftar Undo Geoffrey Hinton on the limits of AI Stating that while GPT-4 is very advanced, it's not perfect, he said 'It's an expert at everything. It's not a very good expert at everything.' Still, Hinton believes future AI models will improve. When asked if GPT-5 would get the riddle right, he replied, "Yeah, I suspect." OpenAI launched GPT-4 in 2023 and became popular for clearing difficult exams like the SAT, GRE, and bar exam. In May 2024, OpenAI released GPT-4o, which became the default model for ChatGPT. The company recently unveiled new models—GPT-4.1, GPT-4.1 Mini, and GPT-4.1 Nano that are designed to outperform previous iterations, offering developers more efficient and cost-effective AI solutions. One of the standout features of GPT-4.1 is its expanded context window, supporting up to 1 million tokens, significantly surpassing the 128,000-token limit of its predecessor, GPT-4o. This enhancement allows the model to process and understand larger datasets, making it particularly effective for complex tasks such as analysing extensive codebases or lengthy documents. Additionally, GPT-4.1 delivers notable gains in coding performance, showing a 21% improvement over GPT-4o and a 27% increase compared to GPT-4.5. AI Masterclass for Students. Upskill Young Ones Today!– Join Now
Yahoo
29-04-2025
- Business
- Yahoo
Beyond SaaS: Why capital is the new platform moat
SaaS has been a cash cow for platforms for more than a decade, but it was only ever meant to be a means to scaling growth - never a destination in itself. Yet the model's recurring revenues and predictability have led many platforms into a SaaS trap. Even though long-term SaaS revenue growth is slowing, they stick to the same formula. But AI raises another threat. A beautiful UI, clever onboarding flows, a 'holistic dashboard' can be copied and created in a weekend with AI tools like GPT-5 and Lovable. When everything can be built and replicated in just a few clicks, the only thing left to make you stand out is a fintech moat. And the strongest –and most obvious – moat is capital. Embedded finance gives platforms more than a feature; it's a system of distribution, data, and trust. AI may be able to copy your features, but it can't copy your lending relationships. It can't help a family business cover cash flow gaps, help an ambitious founder invest in new stock, or enable a restaurant to replace a broken fridge overnight. Embedded lending creates a stickiness that investors love to see and a revenue stream that doesn't rely on constant upselling. You only need to look at the companies pulling ahead to see this impact in action. For instance, Shopify Capital has issued over $2 billion in funding to its merchants since 2016. Merchants that take this funding grow 36% more than their peers, 76% come back for more, and Shopify has seen its revenue soar from $389 million to $8.88 billion since Capital's launch. At Toast, financial services, not software, are now its biggest revenue driver. Uber's embedded payments didn't just boost revenue by 10% in a year, they made the entire product better: 15% higher retention, 25% lower admin costs and 20% better ad conversions. These platforms didn't get lucky. They stopped thinking like feature factories and started giving real and considered thought to what their customers actually need. And all SMEs, no matter their size, sector, or region, need the same thing – money. Of course, this has been true since the early days of the industry, but it's been thrown into stark view in recent years due to the continued decimation of high street banks. The death of local relationships with lenders who truly 'get' your business has left SMEs with few – and dwindling – options. Outdated credit models and one-size-fits-all financing have created a staggering €400bn gap in SME funding in Europe alone. This isn't just bad for those businesses, it's bad for the economy. Small and medium-sized businesses make up the majority of the companies across the continent and beyond, and if they fail or stall, we all do. So if the benefits are so clear and repeatable, why aren't platforms rushing to jump in and provide the new forms of lending these businesses so desperately deserve? In the early days, embedded finance was clunky, expensive, and risky. It made sense for platforms to hold off to see how the market played out, protected by the SaaS frameworks they'd built around them. That excuse is gone. Today, infrastructure providers can handle the heavy lifting, managing everything from compliance to underwriting. Today, all platforms – from e-commerce tools to B2B marketplaces, vertical SaaS, and more – can embed lending, payments, and more without becoming a 'bank': all of the benefits without the risk. The real risk now is doing nothing. The platforms that win in the next decade won't just build features, they'll power businesses. That means offering capital, on your terms, inside your ecosystem, and ahead of the competition. Because if you're not getting your head out of the sand and embedding finance, someone else will – and when they do, SaaS won't save you. It will become the trap that eventually renders your platform irrelevant. Max Schertel, is the co-founder and CEO of finmid "Beyond SaaS: Why capital is the new platform moat" was originally created and published by Retail Banker International, a GlobalData owned brand. The information on this site has been included in good faith for general informational purposes only. It is not intended to amount to advice on which you should rely, and we give no representation, warranty or guarantee, whether express or implied as to its accuracy or completeness. You must obtain professional or specialist advice before taking, or refraining from, any action on the basis of the content on our site. Sign in to access your portfolio
