Yazılar

Delta Electronics Warns AI Boom Is Driving Costs Higher

/in /tarafından

Taiwan’s Delta Electronics, a major supplier of power and cooling systems for AI data centres, has warned that surging artificial intelligence infrastructure demand is pushing operating costs higher. The company expects rising oil prices, material shortages, and broader inflation linked to AI expansion to increase pressure in the coming quarters.

Delta, whose customers include Nvidia, Google, and Meta, said production capacity remains tight as global AI datacentre construction accelerates. To meet demand, the company is expanding operations across China, Thailand, the United States, and Taiwan.

The firm previously announced capital expenditure of T$46.1 billion ($1.46 billion) in 2025 and now says spending will rise even further this year. This reflects how critical energy management, cooling systems, and infrastructure have become as hyperscale cloud providers and AI companies scale hardware deployments.

Despite cost concerns, Delta posted strong first-quarter results. Revenue climbed 34% year-over-year to T$159.35 billion ($5.02 billion), while gross profit surged 56% to T$59 billion ($1.86 billion), largely fueled by AI datacentre growth.

Delta Electronics’ stock has risen nearly 125% this year, significantly outperforming Taiwan’s broader market. The company’s outlook suggests that while AI remains a massive growth driver, supply chain constraints and inflation may increasingly shape profitability across the sector.

Nvidia B300 Servers Hit $1M in China as US Curbs Tighten Supply

/in /tarafından

Nvidia’s advanced B300 AI servers are reportedly selling for nearly 7 million yuan, around $1 million, in China as stricter US export controls and anti-smuggling crackdowns sharply reduce supply. According to industry sources, prices have almost doubled from roughly 4 million yuan late last year, creating a major scarcity premium in the Chinese grey market.

The B300 server, equipped with eight B300 GPUs, costs around $550,000 in the United States, but Chinese demand for high-end AI computing has pushed prices far beyond that level. Chinese technology companies are aggressively seeking cost-efficient hardware to power AI inference and token generation, while many remain cautious about directly holding Nvidia systems due to sanctions concerns.

Reuters reports that pressure increased after US authorities prosecuted Supermicro co-founder Wally Liaw in March, disrupting key black-market supply channels. Nvidia emphasized that B300 systems are restricted from sale in China and warned that unauthorized diversion would receive no support or service from the company.

Some Chinese firms unable to afford direct purchases are instead turning to rentals, with short-term annual contracts reaching 190,000 yuan per month. At the same time, domestic players like Huawei are trying to capitalize on Nvidia’s restricted access, challenging Nvidia’s estimated 55% Chinese AI chip market share.

The surge highlights how geopolitical restrictions are reshaping China’s AI infrastructure market, driving up costs while accelerating local competition in advanced computing hardware.

Nvidia B300 Servers Hit $1 Million in China Amid US Export Crackdown

/in /tarafından

Nvidia’s advanced B300 AI servers are now reportedly selling for nearly $1 million each in China, almost double their U.S. price, as tighter American export restrictions and anti-smuggling enforcement create severe supply shortages. According to industry sources, the scarcity has transformed the B300 into one of China’s most expensive and sought-after AI computing assets.

The B300, equipped with eight GPUs and designed for high-performance AI inference, normally costs around $550,000 in the United States. In China, however, prices have surged to roughly 7 million yuan due to shrinking grey-market channels and rising demand from major Chinese technology firms racing to expand AI model deployment.

China’s growing need for AI infrastructure is accelerating the premium. Local firms are under pressure to secure hardware capable of efficiently processing tokens, a key monetization factor for generative AI systems. At the same time, many companies are cautious about directly owning restricted Nvidia systems because of potential exposure to U.S. sanctions.

The market disruption intensified after U.S. legal action against individuals tied to Nvidia partner Supermicro, further constraining unofficial supply routes. As a result, some Chinese companies are shifting from direct purchases to rentals, with monthly leasing costs reaching as high as 190,000 yuan.

This environment is also creating strategic opportunities for domestic rivals such as Huawei, which aims to capture market share as uncertainty around Nvidia’s H200 and B300 exports continues. Despite sanctions, Nvidia still holds a dominant position in China’s AI chip market, but prolonged restrictions may accelerate local alternatives and reshape competitive dynamics.