China's DeepSeek faces questions over claims after shaking up global tech
The Hangzhou-based startup's announcement that it developed R1 at a fraction of the cost of Silicon Valley's latest models immediately called into question assumptions about the United States's dominance in AI and the sky-high market valuations of its top tech firms.
Some sceptics, however, have challenged DeepSeek's account of working on a shoestring budget, suggesting that the firm likely had access to more advanced chips and more funding than it has acknowledged.
'It's very much an open question whether DeepSeek's claims can be taken at face value. The AI community will be digging into them and we'll find out,' Pedro Domingos, professor emeritus of computer science and engineering at the University of Washington, told Al Jazeera.
'It's plausible to me that they can train a model with $6m,' Domingos added.
'But it's also quite possible that that's just the cost of fine-tuning and post-processing models that cost more, that DeepSeek couldn't have done it without building on more expensive models by others.'
In a research paper released last week, the DeepSeek development team said they had used 2,000 Nvidia H800 GPUs – a less advanced chip originally designed to comply with US export controls – and spent $5.6m to train R1's foundational model, V3.
OpenAI CEO Sam Altman has stated that it cost more than $100m to train its chatbot GPT-4, while analysts have estimated that the model used as many as 25,000 more advanced H100 GPUs.
The announcement by DeepSeek, founded in late 2023 by serial entrepreneur Liang Wenfeng, upended the widely held belief that companies seeking to be at the forefront of AI need to invest billions of dollars in data centres and large quantities of costly high-end chips.
It also raised questions about the effectiveness of Washington's efforts to constrain China's AI sector by banning exports of the most advanced chips.
Shares of California-based Nvidia, which holds a near-monopoly on the supply of GPUs that power generative AI, on Monday plunged 17 percent, wiping nearly $593bn off the chip giant's market value – a figure comparable with the gross domestic product (GDP) of Sweden.
While there is broad consensus that DeepSeek's release of R1 at least represents a significant achievement, some prominent observers have cautioned against taking its claims at face value.
Palmer Luckey, the founder of virtual reality company Oculus VR, on Wednesday labelled DeepSeek's claimed budget as 'bogus' and accused too many 'useful idiots' of falling for 'Chinese propaganda'.
'It is pushed by a Chinese hedge fund to slow investment in American AI startups, service their own shorts against American titans like Nvidia, and hide sanction evasion,' Luckey said in a post on X.
'America is a fertile bed for psyops like this because our media apparatus hates our technology companies and wants to see President Trump fail.'
In an interview with CNBC last week, Alexandr Wang, CEO of Scale AI, also cast doubt on DeepSeek's account, saying it was his 'understanding' that it had access to 50,000 more advanced H100 chips that it could not talk about due to US export controls.
Wang did not provide evidence for his claim.
Tech billionaire Elon Musk, one of US President Donald Trump's closest confidants, backed DeepSeek's sceptics, writing 'Obviously' on X under a post about Wang's claim.
DeepSeek did not respond to requests for comment.
But Zihan Wang, a PhD candidate who worked on an earlier DeepSeek model, hit back at the startup's critics, saying, 'Talk is cheap.'
'It's easy to criticize,' Wang said on X in response to questions from Al Jazeera about the suggestion that DeepSeek's claims should not be taken at face value.
'If they'd spend more time working on the code and reproduce the DeepSeek idea theirselves it will be better than talking on the paper,' Wang added, using an English translation of a Chinese idiom about people who engage in idle talk.
He did not respond directly to a question about whether he believed DeepSeek had spent less than $6m and used less advanced chips to train R1's foundational model.
In a 2023 interview with Chinese media outlet Waves, Liang said his company had stockpiled 10,000 of Nvidia's A100 chips – which are older than the H800 – before the administration of then-US President Joe Biden banned their export.
Users of R1 also point to limitations it faces due to its origins in China, namely its censoring of topics considered sensitive by Beijing, including the 1989 massacre in Tiananmen Square and the status of Taiwan.
In a sign that the initial panic about DeepSeek's potential impact on the US tech sector had begun to recede, Nvidia's stock price on Tuesday recovered nearly 9 percent.
The tech-heavy Nasdaq 100 rose 1.59 percent after dropping more than 3 percent the previous day.
Tim Miller, a professor specialising in AI at the University of Queensland, said it was difficult to say how much stock should be put in DeepSeek's claims.
'The model itself gives away a few details of how it works, but the costs of the main changes that they claim – that I understand – don't 'show up' in the model itself so much,' Miller told Al Jazeera.
Miller said he had not seen any 'alarm bells' but there are reasonable arguments both for and against trusting the research paper.
'The breakthrough is incredible – almost a 'too good to be true' style. The breakdown of costs is unclear,' Miller said.
On the other hand, he said, breakthroughs do happen occasionally in computer science.
'These massive-scale models are a very recent phenomenon, so efficiencies are bound to be found,' Miller said.
'Given they knew that this would be reasonably straightforward for others to reproduce, they would have known that they would look stupid if they were b*********** everyone. There is a team already committed to trying to reproduce the work.'
Falling costs
Lucas Hansen, co-founder of the nonprofit CivAI, said while it was difficult to know whether DeepSeek circumvented US export controls, the startup's claimed training budget referred to V3, which is roughly equivalent to OpenAI's GPT-4, not R1 itself.
'GPT-4 finished training late 2022. There have been a lot of algorithmic and hardware improvements since 2022, driving down the cost of training a GPT-4 class model. A similar situation happened for GPT-2. At the time it was a serious undertaking to train, but now you can train it for $20 in 90 minutes,' Hansen told Al Jazeera.
'DeepSeek made R1 by taking a base model – in this case, V3 – and applying some clever methods to teach that base model to think more carefully,' Hansen added.
'This teaching process is comparatively cheap when compared to the price of training the base model. Now that DeepSeek has published details about how to bootstrap a base model into a thinking model, we will see a huge number of new thinking models.'
Hashtags
Try Our AI Features
Explore what Daily8 AI can do for you:
Comments
No comments yet...
Related Articles
Qatar Tribune
9 hours ago
- Qatar Tribune
Fewer trees, higher prices: China tariff delay too late to save US holidays
Agencies U.S. shoppers in search of artificial Christmas trees and holiday decorations this year will face higher prices and fewer options, as tariffs on Chinese imports prompt retailers to cut back orders while gauging how much households are willing to spend. A 90-day extension to a tariff reprieve – agreed to by Washington and Beijing on Aug. 11 – will allow retailers to rush in some last-minute shipments, but most holiday purchases are already done. Retailers typically import seasonal goods in advance because many products need six-month lead times. 'We're going to have a lower supply year,' said Chris Butler, CEO of National Tree Company, a New Jersey-based artificial tree importer supplying Walmart, Home Depot, Lowe's and Amazon. The company, which sources roughly half its trees from China and the rest from Vietnam, Cambodia, and Thailand, will hike prices by 10% to 20% on its Carolina pine, Nordic spruce, and Dunhill fir trees, Butler said. China is the biggest exporter of Christmas decorations to the U.S., accounting for 87% of such imports last year, worth roughly $4 billion, according to the United States International Trade Commission data. 'We're not overbuying (from our suppliers) because we're not sure about consumer demand and don't want expensive inventory on our books,' Butler retailers are more keen than usual to have National Tree ship directly to consumers rather than buying them as inventory, reducing the stores' risk on their balance sheets. Butler's rival, Mac Harman, CEO of California-based Balsam Hill, expects about 15% fewer trees in the market this season.'Even with the extended 90 days, it's too late for any of us to add orders,' he said. Retailers started cutting orders after U.S. President Donald Trump flip-flopped on China tariffs – raising them to 145% in April, then cutting them to 30% a month later – because they are unwilling to buy trees at elevated prices, said Harman. He sources trees from around 80 suppliers, half of which are in China. Still, the latest pause has netted Balsam Hill some $2.5 million in savings, he said. A Walmart spokesperson said the company was confident in its inventory position heading into the holiday. The reduced demand for fake trees, a key Christmas purchase, signals a muted shopping season. Higher prices on essentials like diapers and dish soap have already strained budgets. Denim maker Levi Strauss said last month it will offer a leaner holiday selection. Isaac Larian, CEO of MGA Entertainment that makes Bratz dolls, said a 30% tariff on toys 'is still too much.' The company has raised prices, he said. Only goods shipped by air would benefit from the delay in higher tariffs, said Chris Rogers, head of supply chain research, S&P Global Market Intelligence. Companies including Apple that have upcoming product launches will gain from the certainty that there will be a 30% tariff, he said. Some suppliers, shippers, and retailers, however, started rushing extra orders after the moratorium was announced, industry sources most footwear makers simplified holiday orders due to tariff uncertainty, a few are placing new orders to add variety to their inventory, said Matt Priest, CEO of industry group Footwear Distributors and Retailers of America (FDRA).It may not be simple to ramp up orders, though. Imports face manufacturing bottlenecks as brands that diversified to other countries after initial tariffs now face delays before new manufacturers can scale up, said Dave Tu, president of DCL Logistics, which imports for clients like GoPro. By and large, this tariff extension changes little for holiday imports. For most companies, said FDRA's Priest, holiday inventory 'is what it is.'
Qatar Tribune
9 hours ago
- Qatar Tribune
China extends probe into EU dairy products
Agencies China extended on Monday its anti-subsidy investigation into European Union dairy imports by six months, showing little sign of let up in a two-front trade war with Brussels and Washington that threatens to upend its export-orientated growth model. The Ministry of Commerce said it had prolonged the investigation period to February 21, 2026, citing the case's complexity. Beijing's decision to extend its probe into some EU cheese, milk and cream products comes after it rolled over in June an investigation into European pork, of which it is also a major buyer, and in July announced duties on EU brandy producers - though major cognac makers were spared provided they sell at or above a minimum price. China is seeking a lasting trade truce with the United States and EU to preserve its export engine as it struggles with the prospect of overhauling its economic model, analysts say, with domestic demand still in the doldrums. Trade tensions between China and the EU erupted in 2023 when the European Commission - which oversees the bloc's trade policy - launched an anti-subsidy probe into China-made electric vehicles (EVs), accusing Beijing of flooding the market with state-backed exports. In April this year, a European Commission spokesperson said the EU and China had agreed to look into setting minimum prices of Chinese-made electric vehicles instead of tariffs imposed by the EU last year. The two sides have yet to reach a deal. 'Beijing is still hoping to come to terms with the EU on a long list of trade conflicts,' said Even Rogers Pay, an analyst at Beijing-based Trivium China who specializes in agriculture. 'This investigation - along with the investigation into EU pork, which was extended in June - are significant bargaining chips in the ongoing negotiations around the EU's tariffs on Chinese new energy vehicles,' she said. China originally announced the dairy probe a year ago, a day after Brussels revised duties on Chinese-made electric vehicles but refrained from abandoning them as Beijing had urged.
Qatar Tribune
9 hours ago
- Qatar Tribune
As Trump builds tariff walls, China and Asean are building bridges
Agencies Amid escalating tensions between the United States and China and lingering trade tensions, China and Asean have gone full steam ahead on upgrading their landmark free trade pact. China and the Association of Southeast Asian Nations – two key players in the Global Southaccount for a quarter of the world's population and a fifth of global gross domestic product, contributing to over 30 per cent of global economic growth. In July, Chinese Foreign Minister Wang Yi announced that both sides would submit the '3.0 version' of the China-Asean Free Trade Agreement for approval in October. This signals a push to future-proof Asia's trade flows amid renewed US protectionism and intensifying global competition. With negotiations over, the pact is expected to be signed by the end of 2025. The initial version of the free trade pact was Asean's first with an external partner. The latest upgrade is set to expand cooperation across nine areas, such as digital and green economies. Since the agreement's inception in 2010, Asean's trade with China has almost tripled – from US$235.5 billion in 2010 to US$696.7 billion in 2023. China has been the bloc's largest trading partner for 16 consecutive years. Asean has also become China's top trading partner for the past five years. Last year, bilateral trade almost reached US$1 trillion. In the first four months of 2025, trade reached US$330.86 billion, a 9.2 per cent year-on-year jump. Upgrading the pact benefits both sides. It gives Asean countries a chance to deepen regional integration and reinforce supply chains amid rising geopolitical uncertainty. A 2023 HSBC survey found that 92 per cent of Indonesian, 87 per cent of Filipino and 89 per cent of Vietnamese companies expect supply chain growth with China through 2026. For Beijing, Southeast Asia is indispensable. The upgraded deal could help China diversify export markets, secure vital raw materials – such as minerals from Indonesia and Thailand – and encourage manufacturers to offshore production to lower costs and bypass Western restrictions. This consolidates China's position as a stable regional economic partner for Asean while counterbalancing US influence and boosting intra-Asian trade. For the US, the upgraded agreement reveals the limits of its tariff offensive. US tariffs are prompting more countries to boost economic ties with China, which is increasingly seen as a steadier source of investment, capital and financing. Indonesia reportedly secured a tariff cut from 32 to 19 per cent. Cambodia had faced a 36 per cent rate, which was also reduced to 19 per cent. However, US President Donald Trump's tariffs threaten Southeast Asia unevenly, ranging from 10 per cent on Singapore to 40 per cent on war-stricken Myanmar and Laos. A shift in public opinion is also under way. The 2024 ISEAS-Yusof Ishak Institute survey, a barometer of regional elite opinion, found that most respondents would pick China over the US if they had to choose. As Washington steps back from multilateralism, Beijing is positioning itself as the defender of world trade norms, challenging Western dominance. However, major hurdles remain. Asean and China must navigate complex regulations on logistics, warehousing and product certification. This demands major infrastructure investment, harmonised data governance and stronger cybersecurity cooperation, efforts that hinge on institutional trust, coordinated reforms and sustained policy support. Equally pressing is the shifting trade balance. Asean accounted for 15 per cent of China's trade in 2023, up from 10 per cent in 2010, according to McKinsey, while China made up 20 per cent of Asean's trade in 2023, a 12 per cent increase over the same time period. Asean imports from China have surged faster than exports, pushing the bloc's trade deficit above US$190 billion. The updated agreement could deepen these imbalances. The relocation of Chinese manufacturers to Southeast Asia is another concern. Facing US technology restrictions, Chinese firms have shifted production to Asean countries to bypass tariffs and avoid the 'Made in China' label. While some countries like Vietnam and Thailand stand to gain from this, despite recent obstacles, a surge of cheaper Chinese imports could also threaten less competitive local producers and further widen trade long-standing South China Sea disputes continue to fuel unease. Past Chinese restrictions on Filipino banana exports show how maritime disputes can affect trade ties. Beijing's renewed commitment to finalising a code of conduct for the South China Sea with Asean brings optimism. Still, overlapping claims and periodic flare-ups could sour goodwill. Disputes must be carefully managed to avoid eroding trust. Overall, the upgraded trade deal offers opportunities for both China and Asean. If both sides manage risks, the pact will accelerate Southeast Asia's rise and solidify China's place at the centre of it.
