logo
Nvidia not permitted to sell its most advanced chips to China

Nvidia not permitted to sell its most advanced chips to China

Demand in China has begun surging for a business that, in theory, shouldn't exist: the repair of advanced Nvidia artificial intelligence chipsets that the U.S. has banned the export of to its trade and tech rival.
Around a dozen boutique companies now offer repair services, according to two such firms in the tech hub of Shenzhen which say they predominantly fix Nvidia's H100 graphics processing units (GPUs) that have somehow made their way to the country, as well as A100 GPUs and a range of other chips.
Even before it was launched, the H100 was banned from sale in China in September 2022 by U.S. authorities keen to rein in Chinese technological development, particularly advances that its military could use. Its predecessor, the A100, was also banned at the same time after being on the market for over two years.
'There is really significant repair demand,' said a co-owner of a firm that has been fixing Nvidia's gaming GPUs for 15 years and began working on AI chips in late 2024.
Business has been so good that the owners created a new company to handle those orders, which now repairs up to 500 Nvidia AI chips per month. Its facilities, as shown in social media advertising, include a room which can accommodate 256 servers, simulating customers' data centre environments to conduct testing and validate repairs.
The rapid growth of the repair industry from late last year supports the view that there has been a significant amount of smuggling of Nvidia chipsets into China. Tenders have shown that the government and the military have made purchases of the U.S. firm's banned AI chips.
Concern about large-scale smuggling of high-end Nvidia products into China has prompted both Republican and Democratic lawmakers to introduce bills that would require the tracking of chipsets so that their location can be verified after they are sold. U.S. President Donald Trump's administration also backed the idea this week.
The thriving repair industry also highlights how Nvidia's advanced GPUs remain in high demand despite new, albeit less powerful, products from Chinese tech giant Huawei .
Though the buying, selling and repair of Nvidia GPUs is not illegal in China, sources for this article were reluctant to draw scrutiny from U.S. or Chinese authorities and declined to be identified.
Nvidia cannot legally provide repair or replacement items for restricted products in China. In contrast, sources said if an Nvidia GPU in another nation has a defect and is under warranty, which is normally three years, the company usually replaces it.
An Nvidia spokesperson said only the company and authorised partners 'are able to provide the service and support that customers need. Using restricted products without approved hardware, software, and technical support is a nonstarter, both technically and economically.'
REPAIR DEMAND MAY NOT FADE
Nvidia has only just been allowed to recommence sales of its H20 AI chipset, which has been specifically developed for China to comply with U.S. restrictions. Switching over to H20 chipsets is, however, not necessarily a simple or good option for Chinese entities.
Price is an issue as one H20 server with eight GPUs inside will likely cost more than 1 million yuan ($139,400), industry sources say. H20 chipsets, which have increased memory bandwidth, have been specifically designed for AI inference work, but firms involved in the training of large language models would likely prefer H100 chipsets which are better suited to that task.
Industry sources said some of the H100 and A100 GPUs in China have been crunching data around the clock for years now, leading to an increase in failure rates. Depending on how frequently a GPU is used and how often it is maintained, an Nvidia GPU generally lasts two to five years before needing to be repaired, they said.
According to the first source, his company charges between 10,000 yuan and 20,000 yuan ($1,400 to $2,800) to fix a GPU depending on the complexity of the problem.
The second Shenzhen-based repair service provider – which shifted from GPU rentals to repairs this year – says it can repair up to 200 Nvidia AI chips each month, charging about 10% of the GPUs' original selling price per repair.
Services generally include software testing, fan repair, printed circuit board and GPU memory fault diagnostics and repair, as well as the replacement of broken parts.
In the meantime, smuggling of high-end Nvidia chips continues. Traders of chips in China say customer demand is pivoting to top-of-the-line B200 chips which Nvidia began shipping to other countries in larger quantities this year.
A server with eight B200 GPUs costs more than 3 million yuan in China, they said.
Orange background

Try Our AI Features

Explore what Daily8 AI can do for you:

Comments

No comments yet...

Related Articles

The new chips designed to solve AI's energy problem
The new chips designed to solve AI's energy problem

Mint

time9 minutes ago

  • Mint

The new chips designed to solve AI's energy problem

'I can't wrap my head around it," says Andrew Wee, who has been a Silicon Valley data-center and hardware guy for 30 years. The 'it" that has him so befuddled—irate, even—is the projected power demands of future AI supercomputers, the ones that are supposed to power humanity's great leap forward. Wee held senior roles at Apple and Meta, and is now head of hardware for cloud provider Cloudflare. He believes the current growth in energy required for AI—which the World Economic Forum estimates will be 50% a year through 2030—is unsustainable. 'We need to find technical solutions, policy solutions and other solutions that solve this collectively," he says. To that end, Wee's team at Cloudflare is testing a radical new kind of microchip, from a startup founded in 2023, called Positron, which has just announced a fresh round of $51.6 million in investment. These chips have the potential to be much more energy efficient than ones from industry leader Nvidia at the all-important task of inference, which is the process by which AI responses are generated from user prompts. While Nvidia chips will continue to be used to train AI for the foreseeable future, more efficient inference could collectively save companies tens of billions of dollars, and a commensurate amount of energy. There are at least a dozen chip startups all battling to sell cloud-computing providers the custom-built inference chips of the future. Then there are the well-funded, multiyear efforts by Google, Amazon and Microsoft to build inference-focused chips to power their own internal AI tools, and to sell to others through their cloud services. The intensity of these efforts, and the scale of the cumulative investment in them, show just how desperate every tech giant—along with many startups—is to provide AI to consumers and businesses without paying the 'Nvidia tax." That's Nvidia's approximately 60% gross margin, the price of buying the company's hardware. Nvidia is very aware of the growing importance of inference and concerns about AI's appetite for energy, says Dion Harris, a senior director at Nvidia who sells the company's biggest customers on the promise of its latest AI hardware. Nvidia's latest Blackwell systems are between 25 and 30 times as efficient at inference, per watt of energy pumped into them, as the previous generation, he adds. To accomplish their goals, makers of novel AI chips are using a strategy that has worked time and again: They are redesigning their chips, from the ground up, expressly for the new class of tasks that is suddenly so important in computing. In the past, that was graphics, and that's how Nvidia built its fortune. Only later did it become apparent graphics chips could be repurposed for AI, but arguably it's never been a perfect fit. Jonathan Ross is chief executive of chip startup Groq, and previously headed Google's AI chip development program. He says he founded Groq (no relation to Elon Musk's xAI chatbot) because he believed there was a fundamentally different way of designing chips—solely to run today's AI models. Groq claims its chips can deliver AI much faster than Nvidia's best chips, and for between one-third and one-sixth as much power as Nvidia's. This is due to their unique design, which has memory embedded in them, rather than being separate. While the specifics of how Groq's chips perform depends on any number of factors, the company's claim that it can deliver inference at a lower cost than is possible with Nvidia's systems is credible, says Jordan Nanos, an analyst at SemiAnalysis who spent a decade working for Hewlett Packard Enterprise. Positron is taking a different approach to delivering inference more quickly. The company, which has already delivered chips to customers including Cloudflare, has created a simplified chip with a narrower range of abilities, in order to perform those tasks more quickly. The company's latest funding round came from Valor Equity Partners, Atreides Management and DFJ Growth, and brings the total amount of investment in the company to $75 million. Positron's next-generation system will compete with Nvidia's next-generation system, known as Vera Rubin. Based on Nvidia's road map, Positron's chips will have two to three times better performance per dollar, and three to six times better performance per unit of electricity pumped into them, says Positron CEO Mitesh Agrawal. Competitors' claims about beating Nvidia at inference often don't reflect all of the things customers take into account when choosing hardware, says Harris. Flexibility matters, and what companies do with their AI chips can change as new models and use cases become popular. Nvidia's customers 'are not necessarily persuaded by the more niche applications of inference," he adds. Cloudflare's initial tests of Positron's chips were encouraging enough to convince Wee to put them into the company's data centers for more long-term tests, which are continuing. It's something that only one other chip startup's hardware has warranted, he says. 'If they do deliver the advertised metrics, we will open the spigot and allow them to deploy in much larger numbers globally," he adds. By commoditizing AI hardware, and allowing Nvidia's customers to switch to more-efficient systems, the forces of competition might bend the curve of future AI power demand, says Wee. 'There is so much FOMO right now, but eventually, I think reason will catch up with reality," he says. One truism of the history of computing is that whenever hardware engineers figure out how to do something faster or more efficiently, coders—and consumers—figure out how to use all of the new performance gains, and then some. Mark Lohmeyer is vice president of AI and computing infrastructure for Google Cloud, where he provides both Google's own custom AI chips, and Nvidia's, to Google and its cloud customers. He says that consumer and business adoption of new, more demanding AI models means that no matter how much more efficiently his team can deliver AI, there is no end in sight to growth in demand for it. Like nearly all other big AI providers, Google is making efforts to find radical new ways to produce energy to feed that AI—including both nuclear power and fusion. The bottom line: While new chips might help individual companies deliver AI more efficiently, the industry as a whole remains on track to consume ever more energy. As a recent report from Anthropic notes, that means energy production, not data centers and chips, could be the real bottleneck for future development of AI. Write to Christopher Mims at

Muizzu's U-turn: India in, China out? Maldives seemingly pivots again as PM Modi offers big incentives
Muizzu's U-turn: India in, China out? Maldives seemingly pivots again as PM Modi offers big incentives

Time of India

time22 minutes ago

  • Time of India

Muizzu's U-turn: India in, China out? Maldives seemingly pivots again as PM Modi offers big incentives

NEW DELHI: Maldives, which has had a pro-China tilt in the past, seems to be warming up to India with President Mohamed Muizzu praising Prime Minister Narendra Modi as a "wonderful person". Coming from someone who rose to power on an "India Out" campaign, followed by his cabinet colleagues making derogatory remarks against PM Modi, shows a significant shift in the geopolitical dynamics of the Global South. The two leaders held talks on Friday to repair bilateral ties as the also discussions focused on strengthening cooperation in trade, defence, and infrastructure. Strategic U-turn After coming to power by promising to reduce Indian influence and build closer ties with China—including a first state visit to Beijing and agreements for Chinese military assistance—Muizzu now calls India a "trusted friend" and rolled out the red carpet for PM Modi, breaking with protocol to personally receive him at the airport. "He is a wonderful person who is very fond of building relationships between India's neighbours. The Maldives and India have a very good relationship that goes back centuries, and with PM Modi's leadership, cooperation between the two governments is going to be even more prosperous in the days ahead," he said on the last day of PM Modi's trip. What could be the reasons behind change in stance? The Maldivian economy is under severe stress with a substantial budget deficit and dwindling foreign reserves, and the support offered by India is pivotal for Muizzu's government to stabilize finances. by Taboola by Taboola Sponsored Links Sponsored Links Promoted Links Promoted Links You May Like No One Knew About Her Bags—Now They're Going Viral Handmakers Report Read Now Although Maldives still owes China large debts, New Delhi has emerged as a key partner to avoid potential economic default. During PM Modi's visit, India announced a fresh $ 565 million line of credit (LoC) for the Indian Ocean neighbour and signed an agreement to reduce its annual debt repayment burden by 40%. 'To give a new impetus to our development partnership, we have decided to provide a line of credit of $ 565 million, or approximately Rs 5,000 crore, to Maldives. It will be used for projects related to infrastructure development in line with the priorities of the people of Maldives,' PM Modi said, as he hailed Maldives as a "true friend". India's sustained efforts However, this diplomatic shift hasn't happened overnight, and it certainly can't be attributed to PM Modi's visit alone. Amid early concerns about losing a key ally in the region, India chose restraint over rhetoric. New Delhi maintained steady engagement, with PM Modi being the first world leader to congratulate Muizzu after his election victory. This gesture, combined with consistent diplomatic outreach and sustained economic support, laid the groundwork for rebuilding trust. Back in 2019, India had extended an $800 million Line of Credit (LoC) to the Maldives during the tenure of the pro-India Maldivian Democratic Party (MDP) government. What PM Modi's trip has done, is to bring India's efforts to light, with Muizzu acknowledging it. "We all have seen how India has helped the Maldives in the past, and nobody will doubt how India will be a very crucial partner going forward," he said. "India is one of the major tourism countries that helps the Maldives with tourism. With PM Modi's visit, it is going to increase a lot. I am sure that people-to-people exchange between the two countries will greatly boost because of this visit," he added. Analysts note that Muizzu's pivot is a pragmatic recognition of Maldives' geographic and economic realities—India remains indispensable for trade, connectivity, and tourism. While China's support remains, India's proximity and willingness to provide urgent financial and developmental assistance have proved decisive in influencing Muizzu's recalibration

Trump plays golf in Scotland while protesters take to the streets and decry his visit
Trump plays golf in Scotland while protesters take to the streets and decry his visit

New Indian Express

time39 minutes ago

  • New Indian Express

Trump plays golf in Scotland while protesters take to the streets and decry his visit

EDINBURGH: President Donald Trump played golf Saturday at his course on Scotland's coast while protesters around the country took to the streets to decry his visit and accuse United Kingdom leaders of pandering to the American. Trump and his son Eric played with the U.S. ambassador to Britain, Warren Stephens, near Turnberry, a historic course that the Trump family's company took over in 2008. Hundreds of protesters gathered on the cobblestone and tree-lined street in front of the U.S. Consulate about 100 miles (160 kilometers) away in Edinburgh, Scotland's capital. Speakers on a makeshift stage told the crowd that Trump was not welcome and they criticized British Prime Minister Keir Starmer for striking a recent trade deal to avoid stiff U.S. tariffs on goods imported from the U.K. Protests were planned in other cities as environmental activists, opponents of Israel's war in Gaza and pro-Ukraine groups loosely formed a 'Stop Trump Coalition." 'I think there are far too many countries that are feeling the pressure of Trump and that they feel that they have to accept him and we should not accept him here,' said June Osbourne, 52, a photographer and photo historian from Edinburgh who protested wearing a red cloak and white hood, recalling "The Handmaid's Tale." Osbourne held up picture of Trump with 'Resist' stamped over his face. The dual-U.S.-British citizen said the Republican president was "the worst thing that has happened to the world, the U.S., in decades.'

DOWNLOAD THE APP

Get Started Now: Download the App

Ready to dive into a world of global content with local flavor? Download Daily8 app today from your preferred app store and start exploring.
app-storeplay-store