ASML shares drop sharply after warning on semiconductor recovery
Gloomy outlook from Europe’s most valuable tech company drags down stocks from Nvidia to Arm
Shares in ASML led a tech rout on Nasdaq on Tuesday after the chip equipment maker warned of a slower recovery in the semiconductor market, in results accidentally published a day early.
The Dutch chipmaker cut its outlook for next year after reporting orders that were only half as much as investors had expected for the third quarter.
Chief executive Christophe Fouquet warned of “customer cautiousness” and a “more gradual” recovery in all areas beyond artificial intelligence.
The disappointing results, which had been scheduled for release on Wednesday, were briefly published on its website on Tuesday before being deleted and then republished 30 minutes later, with ASML blaming a “technical error”.
ASML’s stock was down almost 17 per cent in early afternoon trading in New York as its gloomy outlook reverberated throughout the tech sector.
Shares in US chipmakers Nvidia — which on Monday hit a new all-time high — and AMD were down about 5 per cent, Broadcom was down 3.5 per cent and chip designer Arm was 6.5 per cent lower following the news.
ASML said that its total net sales for 2025 would be €30bn-€35bn, with a gross margin of between 51 per cent and 53 per cent. It had previously said that revenues could be as high as €40bn for the year, with 54 per cent to 56 per cent gross margins.
Europe’s most valuable tech company is the dominant provider of the precision chipmaking machines that Taiwan Semiconductor Manufacturing Company, Intel and Samsung Electronics use to produce their most advanced semiconductors.
Net bookings — a measure of orders placed by ASML’s customers — were €2.6bn for the third quarter, far lower than the more than €5bn that analysts had expected. Analysts at Stifel said in a note to clients that the order intake was “very weak”.
Challenges at Intel and Samsung appear to have contributed to ASML’s underwhelming results.
Intel is racing to cut costs by reducing headcount and delaying investments in new production facilities after it reported disappointing results in August. This month, Samsung acknowledged it was facing a “crisis” after falling behind in AI chipmaking, as a downturn in the memory market looms.
In a pre-recorded interview published on Tuesday, ASML’s finance chief Roger Dassen said “very specific competitive issues in the foundry business” had contributed to a slower recovery in those chip markets that were not benefiting from booming demand for AI computing infrastructure.
“The strong performance of AI clearly continues,” he added.
Dassen also said ASML expected its sales to China to fall next year, from almost half of revenues in the third quarter to “around 20 per cent”.
Shipments to China of ASML’s most advanced lithography machines have been restricted by the Dutch and US governments over the past year, in an effort to constrain Beijing’s ability to develop AI systems. Nonetheless, Chinese chipmakers have still been importing its older equipment as they expand production of the less sophisticated processors used in household appliances and industrial equipment.
LVMH quarterly sales drop as luxury group warns of ‘uncertain’ outlook
Sales at core fashion and leather goods unit down 5%, missing consensus
LVMH’s sales fell in the third quarter as the world’s largest luxury group contended with ebbing luxury demand in what it described as an “uncertain economic and geopolitical environment”.
Group revenues at the conglomerate controlled by French billionaire Bernard Arnault dropped 3 per cent to €19.9bn, below consensus estimates that sales would be up 1 per cent.
Sales in the core fashion and leather goods division, which is seen as a bellwether for the industry, fell 5 per cent year on year compared with the same period a year ago, underperforming Visible Alpha analysts’ consensus of 1 per cent growth.
The company said the decline in growth in the third quarter “mainly arose from lower growth seen in Japan, essentially due to the stronger yen”.
Pressure on sales in Asia outside of Japan continued to weigh down on LVMH, with sales in that region falling 16 per cent in the third quarter.
With more than 75 brands, LVMH spans luxury segments from the closely watched fashion and leather goods division to watches, jewellery and travel.
The UK should resist calls for infrastructure public ownership
Privatisation is not the solution for every situation but it can be highly effective in many cases
I listened on Monday with great interest at the UK government’s International Investment Summit, as ministers, investors and the best of British business shared their plans on how to reinvigorate the economy through infrastructure investment. They certainly face a challenge. In the UK, infrastructure is frequently the subject of public frustration. Something has to change. As the government considers reforms and how to fund much-needed upgrades to roads, water, public transport and digital services, there are three key lessons from my home country, the US, that could inform their approach.
The first is that the UK model works. It needs updating, not a complete overhaul. This is a timely issue given the crises surrounding Thames Water and the growing calls to move away from private ownership in the water sector, and perhaps infrastructure more broadly.
The US has resisted infrastructure solutions that the rest of the world has adopted, especially those involving the private sector. In the US, 90 per cent of the water sector is publicly owned by cities, municipalities, or regional authorities — roughly 50,000 entities in total. According to the American Society of Civil Engineers, the state of this infrastructure is far from ideal. Clean water infrastructure earned a grade of C+ and wastewater infrastructure a D+. Two million Americans don’t have access to clean drinking water.
Airports tell a similar story. In Europe, over 40 per cent of airports have some level of private ownership, while in the US, nearly all are government-owned. Their rating? Also D+.
The second lesson from the US is that believing a government can spend or tax its way out of an infrastructure shortfall is wishful thinking. Private capital has to play a role. The US’s $1.2tn infrastructure bill, for instance, covers only a third of the funding required simply to restore the nation’s systems. It does not include the cost of building the new ones so desperately needed to spur economic growth. With a US budget deficit exceeding 6 per cent of GDP and debt to GDP over 120 per cent, there is no appetite to invest further.
Here lies the third and final lesson for Britain: until the 1930s, much of the US’s infrastructure was privately owned and operated. The Depression and second world war created conditions where greater government intervention made sense. But the temporary shift has since fossilised, largely due to inertia and ideology. The UK should avoid making the mistake of turning its back on privatisation altogether — it’s not always the solution for every situation but it can be highly effective in many cases.
In my recent book, Build, I argue that the US needs a new model to invest in infrastructure — one that includes the private sector through public-private partnerships. The US should look to examples from the UK, Europe, Australia and even China, where such partnerships are already in place.
No serious advocate of privatisation suggests it should be embraced blindly. These are complex legal arrangements that succeed or fail based on regulatory provisions, financing mechanisms and enforcement. But thanks to decades of experience and advancements in technology, the partnerships can now be structured, monitored and enforced more effectively than ever before.
The truly grim reality is that governments cannot afford the infrastructure needs of the coming decades. This includes protecting national security, making infrastructure more resilient to climate change, developing new power generation systems to support renewable energy, integrating battery storage and meeting the growing energy demands of artificial intelligence technologies.
What I heard on Monday was a focus on opportunity and maximising the UK’s strengths. Ultimately, practical, long-term solutions are more likely to come from thoughtful reform that applies real-world lessons, engages the private sector and utilises private capital. Ideological slogans can be politically expedient but they are not going to reinvigorate the UK economy in the way that it wants or needs.
Apple study exposes deep cracks in LLMs’ “reasoning” capabilities
Irrelevant red herrings lead to "catastrophic" failure of logical inference.
or a while now, companies like OpenAI and Google have been touting advanced "reasoning" capabilities as the next big step in their latest artificial intelligence models. Now, though, a new study from six Apple engineers shows that the mathematical "reasoning" displayed by advanced large language models can be extremely brittle and unreliable in the face of seemingly trivial changes to common benchmark problems.
The fragility highlighted in these new results helps support previous research suggesting that LLMs use of probabilistic pattern matching is missing the formal understanding of underlying concepts needed for truly reliable mathematical reasoning capabilities. "Current LLMs are not capable of genuine logical reasoning," the researchers hypothesize based on these results. "Instead, they attempt to replicate the reasoning steps observed in their training data."
Mix it up
In "GSM-Symbolic: Understanding the Limitations of Mathematical Reasoning in Large Language Models"—currently available as a pre-print paper—the six Apple researchers start with GSM8K's standardized set of over 8,000 grade-school level mathematical word problems, which is often used as a benchmark for modern LLMs' complex reasoning capabilities. They then take the novel approach of modifying a portion of that testing set to dynamically replace certain names and numbers with new values—so a question about Sophie getting 31 building blocks for her nephew in GSM8K could become a question about Bill getting 19 building blocks for his brother in the new GSM-Symbolic evaluation.
This approach helps avoid any potential "data contamination" that can result from the static GSM8K questions being fed directly into an AI model's training data. At the same time, these incidental changes don't alter the actual difficulty of the inherent mathematical reasoning at all, meaning models should theoretically perform just as well when tested on GSM-Symbolic as GSM8K.
Instead, when the researchers tested more than 20 state-of-the-art LLMs on GSM-Symbolic, they found average accuracy reduced across the board compared to GSM8K, with performance drops between 0.3 percent and 9.2 percent, depending on the model. The results also showed high variance across 50 separate runs of GSM-Symbolic with different names and values. Gaps of up to 15 percent accuracy between the best and worst runs were common within a single model and, for some reason, changing the numbers tended to result in worse accuracy than changing the names.
This kind of variance—both within different GSM-Symbolic runs and compared to GSM8K results—is more than a little surprising since, as the researchers point out, "the overall reasoning steps needed to solve a question remain the same." The fact that such small changes lead to such variable results suggests to the researchers that these models are not doing any "formal" reasoning but are instead "attempt[ing] to perform a kind of in-distribution pattern-matching, aligning given questions and solution steps with similar ones seen in the training data."
Don’t get distracted
Still, the overall variance shown for the GSM-Symbolic tests was often relatively small in the grand scheme of things. OpenAI's ChatGPT-4o, for instance, dropped from 95.2 percent accuracy on GSM8K to a still-impressive 94.9 percent on GSM-Symbolic. That's a pretty high success rate using either benchmark, regardless of whether or not the model itself is using "formal" reasoning behind the scenes (though total accuracy for many models dropped precipitously when the researchers added just one or two additional logical steps to the problems).
The tested LLMs fared much worse, though, when the Apple researchers modified the GSM-Symbolic benchmark by adding "seemingly relevant but ultimately inconsequential statements" to the questions. For this "GSM-NoOp" benchmark set (short for "no operation"), a question about how many kiwis someone picks across multiple days might be modified to include the incidental detail that "five of them [the kiwis] were a bit smaller than average."
Adding in these red herrings led to what the researchers termed "catastrophic performance drops" in accuracy compared to GSM8K, ranging from 17.5 percent to a whopping 65.7 percent, depending on the model tested. These massive drops in accuracy highlight the inherent limits in using simple "pattern matching" to "convert statements to operations without truly understanding their meaning," the researchers write.
In the example with the smaller kiwis, for instance, most models try to subtract the smaller fruits from the final total because, the researchers surmise, "their training datasets included similar examples that required conversion to subtraction operations." This is the kind of "critical flaw" that the researchers say "suggests deeper issues in [the models'] reasoning processes" that can't be helped with fine-tuning or other refinements.
The illusion of understanding
The results of this new GSM-Symbolic paper aren't completely new in the world of AI research. Other recent papers have similarly suggested that LLMs don't actually perform formal reasoning and instead mimic it with probabilistic pattern-matching of the closest similar data seen in their vast training sets.
Still, the new research highlights just how fragile this kind of mimicry can be when the prompt in question pushes it in a direction that doesn't precisely match any training data. It also highlights the inherent limitations in trying to perform high-level reasoning without any underlying model of the logic or world behind it. As Ars' Benj Edwards put it in a July story about AI video generation:
One of the reasons OpenAI's GPT-4 turned heads in text synthesis is that the model finally reached a size where it was large enough to have absorbed enough information (in training data) to give the impression that it might be able to genuinely understand and model the world when, in reality, a key aspect of its success is that it "knows" far more than most humans and can impress us by combining those existing concepts in novel ways. With enough training data and computation, the AI industry will likely reach what you might call "the illusion of understanding" with AI video synthesis eventually...
We're likely seeing a similar "illusion of understanding" with AI's latest "reasoning" models, and seeing how that illusion can break when the model runs in to unexpected situations.
AI expert Gary Marcus, in his analysis of the new GSM-Symbolic paper, argues that the next big leap in AI capability will only come when these neural networks can integrate true "symbol manipulation, in which some knowledge is represented truly abstractly in terms of variables and operations over those variables, much as we see in algebra and traditional computer programming..." Until then, we're going to get the kind of brittle "reasoning" that can lead AI models to fail mathematical tests in ways that calculators never do.
Universal Plans Early Summer Launch of New Park With Marquee Harry Potter Ride
Wizard-battle attraction is expected to be a major draw as Universal takes on Disney
Universal is readying its attack on Disney’s DIS 0.16%increase; green up pointing triangle theme-park stronghold in Florida.
The company is planning for its 750-acre Universal Epic Universe park to be fully opened as early as Memorial Day weekend of next year—earlier than expected—with tickets on sale as soon as this month, according to people familiar with the matter.
If successful, Universal’s new park could be a major broadside in the theme-park wars, which have intensified in recent years as live entertainment has become more central to generating growth at big media companies like Comcast—Universal’s CMCSA 0.61%increase; green up pointing triangle parent company—and Disney.
Universal said it hasn’t yet confirmed an opening date. The timing of the planned opening could change.
Meeting that Memorial Day target would enable Universal to take advantage of a full summer-vacation season and would be a coup in the company’s long-running rivalry with nearby Walt Disney World. Universal had long said the park was likely to open in summer 2025, though fans and theme-park-industry watchers grew worried earlier this year when the company rolled out a series of updates that listed the opening date as simply 2025.
In the early stages of the park’s opening, Universal plans to require that visitors purchase multiday passes in order to enter Epic Universe, travel agents briefed on the plans said. Those passes would also include admission to Universal’s other Orlando parks, including Universal Studios Florida, these agents said. Ticket holders would be able to spend only one of the days at Epic Universe, in part because demand is anticipated to be high and the company wants to avoid overcrowding, they said.
The approach could prod visitors into spending multiple days at Universal’s properties and make it harder for some visitors to justify buying passes to Disney’s parks during the same Florida trip, which is a common move for many families.
Universal Epic Universe could snatch about one million visitors from Disney during 2025 and 2026 and boost overall attendance at Universal’s three parks in the area by more than eight million, according to estimates made this year by MoffettNathanson.
“They want people there for more than a weekend,” said Jennifer St. Gelais, owner of Practically Perfect Vacations, a Massachusetts-based travel agency that has received updates from Universal about ticketing plans.
Both Universal and Disney have reported slowing growth in their theme-park divisions after an explosion of pent-up demand following the Covid-19 pandemic.
Revenue and income at Universal’s theme parks both fell in the second quarter from a year earlier, as many tourists were choosing to travel internationally or go on cruises instead of to theme parks, said Michael Cavanagh, president of Universal parent Comcast.
One of the most ambitious elements of Universal Epic Universe is a new Harry Potter-themed area set in the 1920s Paris of the Fantastic Beasts movies. The area is slated to include full-scale replicas of street facades.
The section is anchored by a new attraction based on the Ministry of Magic from the Harry Potter movies—a Capitol-type building of sorts for the wizarding world. Guests visiting the attraction, known as Harry Potter and the Battle at the Ministry, will ride in multidirectional elevator cars and watch as Potter and his wizard pals battle villain Dolores Umbridge and her allies, the Death Eaters, according to preview videos released by Universal and some of the people familiar with the plans.
The attraction is one of the most technically complex Universal has ever built, and seeks to compete on quality with the best rides in the world.
The Ministry of Magic ride is aimed to be what is known in the industry as an “intent-to-visit” attraction—a ride that visitors consider to be worth the price of admission just on its own—said Jim Hill, who runs a theme-park-industry publication and hosts several theme-park podcasts.
“It’s a moonshot,” Hill said.
Epic Universe, which Comcast announced in 2019 and was delayed more than a year by the pandemic, is set to include Super Mario Bros. and How to Train Your Dragon themed areas in addition to Harry Potter. Two other areas will include attractions based on new characters and story lines: Celestial Universe—which has an outer-space theme—and Dark Universe, based on monsters from classic movies like “Frankenstein” and “The Wolf Man.”
Universal is hoping that tourists at Epic Universe will be drawn in by large-scale, marquee attractions, like the Donkey Kong roller coaster in Super Nintendo World. That area builds on the success of Universal Pictures’ and Illumination’s “The Super Mario Bros. Movie,” which was last year’s highest-grossing animated film, earning $1.36 billion at the global box office.
In August, Disney announced a slew of new rides and expansions to its Florida theme parks that theme park chief Josh D’Amaro described as finalized ideas that would all eventually be built, rather than “blue sky” ideas that are still in the conceptual stage. He described one such expansion as the biggest “in the park’s entire history.”
More recently, Disney has announced promotions at its Florida parks that extend through the early part of next year’s summer season, including a low-price water park pass and $200-a-night hotel discounts for visitors who buy park passes in packages that include stays at Disney-owned hotels.
Disney Chief Executive Bob Iger said at a May investor conference that when Universal expanded in the past, it brought more visitors to Orlando.
“It’s not something that should be distracting to us or anxiety-provoking,” Iger said of Universal’s new park.
LONDON, Oct 15 (Reuters) - Portugal's main power utility EDP EDP.LS initiated merger discussions with London-listed peer SSE SSE.L earlier this year, but talks did not progress, three people with familiar with the matter told Reuters.
The proposed deal would have created one of Europe’s largest utilities, with the companies' combined market value exceeding $44 billion, though still smaller than leaders Iberdrola IBE.MC and Enel ENEI.MI.
EDP made the approach before the summer, driven by the prospect of creating an asset portfolio that would be broader in scope, two of the people said. SSE’s management declined the proposal, preferring to focus on investing to grow the company independently, the people added.
Talks ended around June, one of the people said, speaking on condition of anonymity because they were not authorized to speak publicly.
EDP declined to comment. SSE said it does not comment on market speculation.