DeepSeek claims AI model trained for just $294,000, challenging U.S. rivals

/in /tarafından

Chinese AI developer DeepSeek has disclosed that its reasoning-focused R1 model cost just $294,000 to train—dramatically below the hundreds of millions reportedly spent by U.S. leaders such as OpenAI. The figure, revealed in a Nature article co-authored by founder Liang Wenfeng, is the company’s first public estimate of training costs and is likely to reignite debate over China’s position in the global AI race.

According to the paper, R1 was trained on a cluster of 512 Nvidia H800 chips over 80 hours. DeepSeek acknowledged for the first time that it also owns Nvidia A100 GPUs, which were used in preparatory phases before training shifted to the China-specific H800s. The H800 was designed to comply with U.S. export restrictions that bar Nvidia from selling its more powerful H100 and A100 chips to China.

The cost revelation is striking: OpenAI CEO Sam Altman has said foundational models cost “much more” than $100 million to train, though OpenAI has never published detailed figures. DeepSeek’s claim of drastically lower costs fueled January’s investor selloff in global tech stocks, amid fears it could disrupt the market dominance of Nvidia and other AI giants.

Skepticism remains. U.S. officials have suggested DeepSeek may have obtained H100 chips despite restrictions, while U.S. companies have questioned whether its development relied on model distillation—a technique where one AI model learns from another. DeepSeek has admitted using Meta’s open-source Llama models and said its training data may have included content generated by OpenAI systems, though it insists this was incidental.

DeepSeek defends distillation as an efficient way to cut costs and expand access to AI by reducing the enormous energy and resource demands of large-scale training. Analysts note this could accelerate the spread of competitive AI models outside the U.S., though questions about intellectual property and national security will remain central to the debate.

Apple’s iPhone 17 launch draws long queues in Beijing, Pro Max tipped as bestseller

/in /tarafından

Apple’s new iPhone 17 launch attracted large crowds in Beijing on Friday, with around 300 customers lining up outside the flagship Sanlitun store to collect pre-ordered devices. The turnout suggests a promising start for Apple in China, its second-largest market, where it has faced declining shipments and fierce competition from local rivals Xiaomi and Huawei.

Among those queuing, 35-year-old Shuke Wang picked up the iPhone 17 Pro Max, which starts at 9,999 yuan ($1,406) and is expected by analysts to be the top-selling model of the series. Wang praised the redesign but noted the orange version looked “too flashy.” Apple highlights the Pro Max’s extended battery life as a key feature.

The base iPhone 17 offers a brighter, more scratch-resistant screen and an upgraded front-facing camera optimized for horizontal selfies. Meanwhile, the iPhone Air model introduces support for eSIM in China—pending regulatory approval from the country’s telecom giants—though Apple has not opened pre-sales for it. Analysts view the Air as a testing ground for slim designs that may eventually feed into foldable iPhones, though compromises in battery, camera, and audio quality could dampen its appeal among Chinese consumers.

Apple’s shipments in China fell 6% year-on-year in Q3, according to Counterpoint Research, but analysts predict a rebound. Omdia expects iPhone shipments in China to climb 11% in the second half of 2025, helping Apple to a 5% global full-year growth. The Pro Max model, driven by its major redesign, is projected to outperform last year’s 16 Pro Max and dominate Apple’s sales in China by 2026.

Nvidia takes $5B stake in Intel, forging alliance on future AI chips

/in /tarafından

Nvidia announced a $5 billion investment in Intel, acquiring roughly 4% of the struggling chipmaker and pledging to jointly develop new chips for PCs and data centers. The deal comes just weeks after the U.S. government took an extraordinary 10% stake in Intel to shore up the company amid mounting concerns about its competitiveness.

Intel shares surged 23% on the news, while Nvidia’s stock rose nearly 4%. Nvidia will pay $23.28 per share, slightly below Intel’s prior closing price but above what Washington paid earlier this month. The investment makes Nvidia one of Intel’s largest shareholders and marks a pivotal moment in the U.S. effort to counterbalance Asia’s dominance in chip production.

Under the pact, Intel will supply central processors and advanced packaging for joint products that combine Intel CPUs with Nvidia GPUs, linked by Nvidia’s high-speed proprietary technology. The companies pledged to build “multiple generations” of such products, though Nvidia stopped short of committing to use Intel’s foundries for its own chips—a key issue for Intel’s turnaround.

The partnership could reshape the competitive landscape. Analysts say it poses the most immediate risk to AMD, which competes with Intel in supplying data center CPUs, and a longer-term threat to TSMC, which currently manufactures Nvidia’s flagship processors. Broadcom, whose chip-to-chip interconnect technology underpins many AI systems, may also feel pressure.

“This is a massive game-changer for Intel and effectively resets its position of AI-laggard into a cog in future AI infrastructure,” said Gadjo Sevilla, senior analyst at eMarketer. Some analysts even speculate the deal could be the first step toward an eventual breakup or acquisition of Intel by U.S. chipmakers.

Intel’s new CEO, Lip-Bu Tan, has vowed to streamline operations and build capacity more cautiously, only when demand is clear. Nvidia CEO Jensen Huang emphasized the administration was not directly involved in the partnership but noted Washington would welcome the collaboration.

For Intel, the deal adds to a growing cash reserve after a $2 billion investment from SoftBank and $5.7 billion from the U.S. government. For Nvidia, the alliance gives it a foothold in Intel’s deep enterprise and government networks, while cementing its dominance in AI infrastructure.