TheChek News Flow

Huawei improves AI chip production in boost for China’s tech goals

Chinese company improves ‘yield’ of latest semiconductor, despite US efforts to prevent manufacturing advances

Huawei has significantly improved the amount of advanced artificial intelligence chips it can produce, in a key breakthrough that supports China’s push to create its own advanced semiconductors.

The Chinese conglomerate has increased the “yield” — the percentage of functional chips made on its production line — of its latest AI chips to close to 40 per cent, according to two people with knowledge of the matter. That represents a doubling from 20 per cent about a year ago.

The move represents an important advance for Huawei, which has been rolling out its latest Ascend 910C processors, which offer better performance than its previous 910B product.

The improved yield means that Huawei’s production line for Ascend chips has become profitable for the first time, according to the people with knowledge of its business. The company has a goal to further improve yields to 60 per cent, in line with the industry standard for similar chips.

The breakthrough is a step forward for China’s hopes to build computing infrastructure that can support its burgeoning AI industry, despite US export controls designed to hamper the country’s ability to develop sensitive technologies.

The effort has state support, with Beijing urging local tech companies to buy more of Huawei’s AI chips and shift away from $3.3tn US chipmaker Nvidia, which remains the market leader in China by far.  

Huawei founder Ren Zhengfei told Chinese President Xi Jinping last week that the worries China had about a “lack of core and soul” had eased, adding “I firmly believe a greater China will rise faster”, the People’s Daily reported.

The phrase “lack of core and soul” dates back to a 1999 comment by a former China technology minister about the country’s information industry, with “core” referring to semiconductors and “soul” referring to operating systems.

Huawei’s recent progress is also significant to achieving China’s goal of reaching full independence for advanced chip production.

The world’s leading chip manufacturer, Taiwan Semiconductor Manufacturing Company, was forced to stop making Ascend chips and advanced smartphone chips in 2020, after Washington blocked Huawei from accessing manufacturing that used US technology.

Austin Lyons, semiconductor analyst with consultancy Creative Strategies, compared Huawei’s production milestone to TSMC’s estimated 60 per cent yield for production of Nvidia’s H100 AI processor, a similarly sized chip. On that basis, it is possible that a rival product such as Huawei’s would be commercially viable at a 40 per cent yield, he said.

Huawei partnered with the sanctioned Chinese fabrication group Semiconductor Manufacturing International Corp to relaunch its Ascend chip.

SMIC currently uses its so-called N+2 process, which is capable of producing advanced chips without extreme ultraviolet technology. China is currently banned from purchasing EUV lithography machines, the most cutting-edge chipmaking equipment from Dutch group ASML.  

Shenzhen-based Huawei plans to produce 100,000 910C processors and 300,000 910B chips this year, said people with knowledge of its plans. This compares with 200,000 910B and no mass production of 910C in 2024.

The figures suggest that Nvidia will still sell more AI chips in China than Huawei, despite the US company only being able to sell Chinese customers its H20 chips, a less powerful version of its H100 chips designed to adhere to Washington export controls.

The consultancy SemiAnalysis has estimated that Nvidia made $12bn selling 1mn of its H20 chips to China last year.

Huawei faces challenges to convince more customers to abandon Nvidia. One person close to the business pointed to Nvidia’s Cuda software, which is known for being easier to use and capable of faster data processing than Huawei’s offerings.

AI companies and Huawei researchers have also said that the Ascend 910B did not work well for large-scale model training, because of problems with inter-chip connectivity and memory issues.

Huawei has been trying to improve these issues by working with partners to resolve software bugs and increase memory capacity in its latest 910C series. 

However, Huawei has still emerged as the frontrunner to challenge Nvidia in the market for so-called “inference” chips, the hardware used to run AI models once they have been trained.

Prospective customers for the Ascend chip have also cited difficulties securing supplies, with Huawei prioritising orders for large state-run cloud providers like China Mobile.

Huawei currently accounts for more than three-quarters of the overall output of AI chips in China, said one of the people with knowledge of its business. The smaller rivals have struggled to compete with Huawei to get enough capacity at SMIC’s leading nodes, the person added. 

Huawei declined to comment.

Weather forecasting takes big step forward with Europe’s new AI system

Predictions for up to 15 days ahead expected to improve tracking of extreme events

Improved weather prediction using artificial intelligence promises to take a big step forward with the launch of a new European system, which can outperform conventional forecasting methods for up to 15 days ahead.

While tech companies and meteorological offices around the world are already applying AI to the weather, the European Centre for Medium-range Weather Forecasts (ECMWF) said its operational model broke new ground by making global predictions freely available to everyone at any time.

“This milestone will transform weather science and predictions,” said Florence Rabier, director-general of ECMWF, an intergovernmental organisation. “Making the AI Forecasting System operational produces the widest range of parameters using machine learning available to date.”

An experimental version tested over the past 18 months showed the system was about 20 per cent more accurate on key predictions than the best conventional methods, which feed millions of worldwide weather observations into supercomputers and crunch them with physics-based equations.

The new European system could predict the track of a tropical cyclone 12 hours further ahead, giving valuable extra warning time for severe events, said Florian Pappenberger, ECMWF director of forecasts.

The world experienced its hottest temperatures on record in 2024, and Europe has become the fastest warming continent, triggering extreme weather events. The agency has been at the forefront of observations and public awareness about the effects of climate change.

Other medium-range AI forecasting systems under development include GenCast and GraphCast from Google DeepMind, Pangu-Weather from Huawei, FourCastNet from Nvidia and FuXi from Shanghai Academy of AI for Science and Fudan University. All were trained on a database of weather observations compiled by the ECMWF over 40 years.

Comparing the accuracy of competing AI forecasting systems was hard, Pappenberger said, because their relative performance differed according to the variables and timescales assessed. Scores published by the ECMWF give some idea of performance but do not identify an overall champion.

But Pappenberger noted that its system stood out for predicting many more features than standard temperature, precipitation and wind. For example, it also forecasts solar radiation and wind speeds at 100 metres — the height of a typical turbine — helpful for the renewable energy sector.

Although ECMWF forecasts are freely available, the agency does not issue severe weather alerts nor tailor-made predictions to industry users, leaving the specialised forecasts to national or local authorities and private companies.

ECMWF and a group of European national met offices have created an open-source technical framework for AI weather systems called Anemoi, after the Greek god of the winds. The underlying machine-learning architecture is based on the same “graph neural network” as Google DeepMind’s forecasting models.

Peter Battaglia, research director at DeepMind, said it was “impressive” to see how the ECMWF had adapted to the AI wave that had reshaped the field in recent years, and the latest open model would add to the pool of knowledge.

The ECMWF plans to improve its system further by increasing its spatial resolution and moving from the present version, which generates one forecast at a time, to “ensemble forecasting” — or creating a collection of 50 forecasts simultaneously with slightly different starting conditions to provide a range of possible outcomes.

In the future, said Kirstine Dale, chief AI officer at the UK Met Office, a mix of physics-based and data-based simulations would be needed for “their combined strengths to provide accurate, fast, reliable and trustworthy forecasts”.

Today the boundaries of reliable day-to-day weather forecasts in Europe are six to seven days ahead for precipitation and wind, and up to 14 or 15 days for temperature, said Pappenberger.

“Machine learning models have a fair chance of extending that because they may be able to extract something out of the data that we may not currently represent well enough in physics-based models.”

Europe must cut electricity taxes, warn Iberdrola and EDP bosses

Energy company chiefs say EU must ‘urgently’ remove various taxes and simplify bills

Europe must cut taxes on electricity if it wants to help struggling industries become more competitive, the heads of two of the continent’s largest power companies have said. 

Ignacio Galán, the chief executive of Spanish power group Iberdrola, said the EU needed to “urgently” move to remove the various taxes and levies on electricity that have built up over time.

“In the US, these costs amount to 10 per cent of consumer bills,” he noted, adding that in Iberdrola’s home market of Spain, taxes can make up closer to 40 per cent of bills. 

He said the EU should find a way of making taxes “level” in the bloc compared with the US. 

Separately, Miguel Stilwell d’Andrade, chief executive of the Portuguese energy company EDP, said that the EU needed to “simplify what goes into the energy bill for customers”. 

He added that electricity bills were a convenient way for governments to raise tax because they were impossible to evade without being cut off from power.

“It’s a sure thing that you can collect very easily,” he said, adding that the extras tacked on to the price of electricity can be “in many places the biggest part of the bill”.

The two men spoke ahead of the launch of the EU’s Clean Industrial Deal policy on February 26, a package of measures designed to help the bloc’s under-pressure industries cope with the costs of the energy transition.

Dan Jørgensen, EU energy commissioner, said at a Financial Times event in Brussels earlier this month that the policy would look at reducing the “non-energy” component of bills. 

In a joint article for the FT last month, Ursula von der Leyen and Christine Lagarde, the presidents of the European Commission and European Central Bank respectively, promised measures to reduce energy taxes. 

Electricity prices in the EU were two to three times higher than in the US, said Mario Draghi, the former head of the European Central Bank, in a report on EU competitiveness last year, and must come down if companies have any hope of rivalling their peers in the US and Asia.  

A large chunk of the bill is made up of various taxes and levies imposed by national governments, which in the first half of last year rose from 18.5 per cent to 24.3 per cent of bills, compared with the previous year, according to Eurostat. 

These taxes are a key source of revenue for national governments, amounting to 4.2 per cent of their total tax revenues in 2021, according to the European Commission. Long-running negotiations among EU countries to rejig energy taxes have so far made little progress. 

Kristian Ruby, secretary-general of industry group Eurelectric, said electricity is still taxed “more than gas” in the EU and that removing electricity taxes “is a no-brainer to ensure more affordable prices”.

Galán said that the EU also needed to speed up its planning and permitting process and to encourage companies to invest more in electricity grids. 

He contrasted Iberdrola’s business in the US and the UK, where it plans to invest £24bn between 2024 and 2028 in improvements to the electricity grid, with Spain, where he said there was a cap on investment in networks.

“So look at the difference. In Britain they are asking for more and more, and in Spain we are capped, we cannot invest more. That’s why the money is going there (to the UK).” 

The London Stock Exchange has been left behind by its parent

While its equities platform suffers an exodus, the LSEG itself is a financial success story — with an identity problem

The London Stock Exchange was hit by more bad news last week when Glencore, the mining group, said it was considering moving its primary listing to another exchange in search of a higher valuation. Glencore could join an exodus of companies such as Flutter and the building materials group CRH.

The exchange can take comfort from one thing: it is not going to lose the listing of a much larger company, the ninth most valuable in the FTSE 100 with a market capitalisation of £62bn at the start of this week. That venture has turned itself in the past decade from a takeover target into the European champion of a global industry: the London Stock Exchange Group.

The gulf in fortunes between the London Stock Exchange and its parent LSEG is striking. The flow of setbacks for the former has not impeded the steady rise in the latter’s financial value. The explanation is simple: share listing and trading only account for 3 per cent of LSEG’s revenues due to its diversification over the past decade.

LSEG’s transformation from a stock exchange to a global hub for securities trading and clearing, along with data and analytics, is hard to fault as a defence strategy. In 2016, it agreed a merger of equals with Deutsche Börse, but this was blocked by the EU. It is worth far more than the latter now, and is six times the market value of Euronext, which owns the Amsterdam and Paris exchanges.

But LSEG’s financial success story leaves it with an identity problem. The group is not considering moving to the New York Stock Exchange, but it might be worth more if it did, thanks to its highly rated global trading and data businesses. Another venture with a less weighty history in a similar financial position would be tempted to change its name or sell the lagging business.

As things stand, the exchange and LSEG are stuck in an odd relationship. LSEG gains lustre from a business that dates back to the 17th century, while the exchange benefits from being owned by a group that is not vulnerable to takeover bids. But it would be simpler for both if their financial prospects were more aligned.

Investors will this week hear LSEG’s response when David Schwimmer, chief executive since 2018, presents its 2024 results. His predecessor Xavier Rolet started the transformation but Schwimmer, a former Goldman Sachs banker, has advanced it. He carried out its sharpest pivot, the $27bn acquisition in 2019 of the financial data provider Refinitiv.

LSEG has turned Refinitiv into a somewhat stronger challenger to Bloomberg under its Workspace brand, helped by an alliance with Microsoft. The deal also gave it a 51 per cent stake in Tradeweb Markets, a US bond and fixed-income derivatives platform that has grown rapidly. It is much less famous than the London Stock Exchange, but has higher revenues.

There is a broader lesson for exchanges and market platforms: investors often attach the highest values to the least exciting operations. LCH, which is one of the world’s largest bond and derivatives clearing houses, is among LSEG’s most highly valued assets. While equity listings get more attention, bond trading and clearing are quietly reliable.

Please use the sharing tools found via the share button at the top or side of articles. Copying articles to share with others is a breach of FT.com T&Cs aLSEG’s antithesis is Euronext, which relies far more heavily on equities. Euronext’s market value is lower but, on measures such as the total number of new listings and trading volumes on its seven stock exchanges, it has been beating London. Stéphane Boujnah, chief executive, claimed last month that losing listings to New York was just “a London problem”.

Boujnah would clearly love LSEG to lose patience with the London Stock Exchange and sell it to him. That is implausible, given the importance of the exchange to London as a financial centre, no matter how small a share it is of LSEG’s revenues. LSEG is in a much better position to buy Euronext itself, but that would raise competition issues as well as European hackles.

The truth is that LSEG did well to remove itself from the old game of stock exchange musical chairs, and to build a broader business under a related brand. It would probably not be thanked by shareholders for turning back: there are many data, analytics and trade-related operations to build or buy instead.

But the London Stock Exchange and its parent should not drift too far apart. While it may not matter much to LSEG’s financial valuation that its original asset is having trouble, the exchange remains the heart of the brand.

Private equity investor’s partner says the region’s market ‘isn’t as crowded as in the US’

US private equity group Thoma Bravo has raised its first dedicated European fund, as one of the industry’s most prolific investors hunts for software takeovers that are less pricey and competitive than in its home market.

The firm said on Tuesday that it had closed a €1.8bn fund for regional software deals, after establishing a London office in 2023.

While the firm has deployed €14bn of cash across deals in Europe during the past 14 years — including high-profile transactions like the buyout of the UK cyber security group Darktrace — the new fund cements its strategy in the region.

“We think Europe is better value,” said Irina Hemmers, a partner at Thoma Bravo who leads its Europe office.

Particularly in software, where the firm has carved out a niche, “the market isn’t as crowded as in the US . . . the ability for us to differentiate is even greater,” she said in an interview with the Financial Times.

Thoma Bravo’s European presence is modest compared with its overall $166bn in assets under management. The group has about 10 staff in Europe, including seven members of the investment team, out of about 100 investors globally, Hemmers said.

The new Europe fund will focus on mid-market software deals with enterprise values of between €150mn and €1bn, Hemmers said, with an emphasis on risk and compliance, healthcare, and tools for chief financial officers. Last year, Thoma Bravo closed its €400mn buyout of German corporate compliance group EQS.

The European push comes as some investors say US technology stocks are overvalued, offering few opportunities for substantial gains.

Hemmers cautioned that the macroeconomic outlook for Europe was still challenging, but that the software sector had fared better as bigger companies pushed to automate and digitise their operations.

“The macro context is not great and I think that’s why a lot of generalist investors are down on Europe,” she said. “Software is somewhat isolated from that.”

In the US, Thoma Bravo is set for a $4bn windfall after the initial public offering of cyber security specialist SailPoint in a transaction that would mark a large gain from a takeover struck at the height of the pandemic dealmaking frenzy.

Hemmers said that the US market for new listings was ahead of Europe, where many private equity groups are hoping for a revival that will allow them to exit some of their record unsold assets.

“We’re all watching the IPO markets hopefully opening,” she said. “We still need to see evidence about the European markets.”

The U.S. government is leaning hard on tech companies to make more commitments to building their businesses in the country, and Big Tech is falling in line. On Monday, Apple laid out its own plans in that area: It will spend $500 billion over the next four years in areas like high-end manufacturing, engineering, and education covering technologies like artificial intelligence and chip making.

Big projects will include a new factory in Houston, Texas, to produce servers that support Apple’s in-house AI effort, Apple Intelligence; doubling the value of Apple’s U.S. Advanced Manufacturing Fund to $10 billion; a new academy in Michigan to train people to work in next-generation factories; and more R&D.

Some of this is not “new” news. Apple has worked for years with thousands of suppliers across the U.S. in areas like chip making — currently 24 factories across 12 states — alongside directly employing people in the country. Globally, Apple employs 164,000 people, according to recent filings. It does not break out how many of them are in the U.S. specifically. It said today it plans to hire another 20,000 people in the next four years. But again, it does not specify if these people will be in the U.S. or elsewhere.

Nevertheless, Apple’s news is significant because of what it underscores. First, there is the bigger effort that the U.S. has been making to expand its economic footing, specifically to remove some of the reliance that the U.S. currently has on ecosystems outside of the U.S. itself, such as China for manufacturing. The U.S. is waging a fairly drastic effort to shift investment in line with that, for example, by floating new tariffs on certain goods in an effort to drive more national production.

The magic number is $500 billion: It’s also the amount that SoftBank, Oracle, and OpenAI are apparently committing to their own major AI data center project.

Apple, as a major consumer electronics company, relies heavily on production outside of the U.S. The exercise of laying out plans to invest within the U.S. will not completely replace that, now or ever, but becomes a bone — a very valuable bone — that it can throw to show that it’s making efforts too.

Second, the focus on artificial intelligence in Apple’s news today should be noted. The major server factory that it will be building will be focused on building machines that can handle AI compute. Similarly, the ecosystem fund and training budget are largely focused on skills and manufacturing of hardware that will be used in AI systems.

Of note: It is not clear what kinds of tax breaks (if any) companies will get on the investments such as the ones Apple listed today. That will be top of mind for companies, their investors, and hopefully the U.S. public. Apple did note that it “remains one of the largest U.S. taxpayers, having paid more than $75 billion in U.S. taxes over the past five years, including $19 billion in 2024 alone.”

The news today, in any case, is being represented as Apple’s own commitment to growing America’s industry profile in the world.

“We are bullish on the future of American innovation, and we’re proud to build on our long-standing U.S. investments with this $500 billion commitment to our country’s future,” said Tim Cook, Apple’s CEO, in a statement. “From doubling our Advanced Manufacturing Fund, to building advanced technology in Texas, we’re thrilled to expand our support for American manufacturing. And we’ll keep working with people and companies across this country to help write an extraordinary new chapter in the history of American innovation.”

One of the bigger specific projects announced today will be a new 250,000-square-foot AI server manufacturing facility in Houston — taking on building services that up to now have been manufactured in other countries. Ground breaks later this year, and it will be completed by 2026, it said.

The project is important not just in value but also intention: Apple is doubling down on how it believes AI will be used within its products and services. So the project is coming along with an expansion of server capacity in Apple’s other data centers in North Carolina, Iowa, Oregon, Arizona, and Nevada.

“Teams at Apple designed the servers to be incredibly energy efficient, reducing the energy demands of Apple data centers,” Apple said, although it also claimed these are already run on renewable energy.

The manufacturing fund, in contrast, will be used to help finance expansions for its partners, including a “multibillion-dollar commitment” to TSMC for advanced silicon made in the latter company’s Fab 21 facility in Arizona. Apple said it is Fab 21’s largest customer.

Apple has not specified how much it has earmarked for educational initiatives aimed at training workforces — although the costs of building factories or investing in frontier-level research and development are likely to be substantial.

The first effort in that vein will be a new Apple Manufacturing Academy in Detroit, it said, where “Apple engineers, along with experts from top universities such as Michigan State,” will work in consultation with SMBs to help them implement “AI and smart manufacturing techniques.” There are a large number of smaller businesses in that region that have worked in concert in other legacy industries like automotive, and it will be worth watching to see how and if they make the transition as envisioned.