Jefferies says cheaper AI models to boost infrastructure demand; Favours memory chip makers
New Delhi, June 26
The emergence of low-cost artificial intelligence models is unlikely to slow AI investments. Instead, it could increase demand for computing infrastructure, according to a report by global brokerage Jefferies.
The report said the launch of Chinese AI company Z.ai's GLM-5.2 model marks another key development in the AI industry. It said the model has intensified competition with leading Western AI firms while reducing inference costs.
"The past week has seen another DeepSeek moment," the report said, referring to the growing competition from Chinese AI developers.
Jefferies said GLM-5.2 delivers performance close to leading enterprise AI models at a much lower operating cost. Lower costs could encourage wider AI adoption across businesses.
The brokerage also noted that falling token costs are prompting more companies to deploy AI models on their own servers instead of relying on public cloud platforms. This helps improve data security and reduce cloud dependence.
"GLM-5.2 proves enterprises no longer have to sacrifice intelligence for privacy. We are seeing a massive acceleration in companies pulling their AI workloads out of the public cloud and back onto local corporate servers," the report said.
According to Jefferies, Chinese AI models have rapidly increased their share of global usage on OpenRouter, reflecting growing acceptance of lower-cost AI alternatives.
The report said lower AI costs could eventually increase overall demand for computing power through the economic principle known as Jevons Paradox, where greater efficiency leads to higher overall consumption.
Jefferies believes this trend will benefit AI hardware suppliers, especially memory chip makers, as higher computing demand is expected to support stronger Dynamic Random Access Memory (DRAM) demand and pricing.
The brokerage also said there is currently "zero sign of AI capex slowing," indicating that hyperscalers and AI developers continue to invest heavily in data centres and computing infrastructure.
Jefferies said the biggest long-term risk remains whether companies will generate sufficient returns on their large AI investments. However, it added that these concerns remain theoretical for now as investment momentum continues.
Reflecting its positive outlook, Jefferies added South Korean memory maker SK Hynix and Japanese flash memory company Kioxia to its model portfolios. It has also increased its weighting in Samsung Electronics while reducing exposure to internet companies such as Alphabet and Alibaba.
The report also highlighted strong AI-driven investment trends in Taiwan. It said the country's economy, exports and semiconductor capital expenditure continue to benefit from the global AI infrastructure expansion led by companies such as TSMC.
— ANI
Reader Comments
The Jevons Paradox point is spot on. When technology becomes cheaper, usage explodes. Remember how affordable smartphones led to a data consumption boom in India? Same thing could happen with AI. Memory chip makers like SK Hynix and Samsung could indeed be big beneficiaries 🚀
While the trend looks promising, I'm a bit cautious. The report says there's "zero sign of AI capex slowing," but we've seen hype cycles before. Remember the dot-com boom? Massive infrastructure investment that didn't always pay off. Let's see if enterprise adoption actually materializes this time.
Good to see Jefferies highlighting Chinese AI models like GLM-5.2. Competition is healthy, and lower costs will help Indian SMEs adopt AI without breaking the bank. The move toward local server deployments for privacy is especially relevant given India's data protection laws coming into effect.
The report mentions Taiwan benefiting from AI infrastructure, but what about India? We have a strong IT services sector and growing chip design talent. With the right policies, we could capture some of this DRAM and computing demand. Just wish Jefferies had a word on Indian players specifically.
As someone working in IT infrastructure, I've seen this shift firsthand. Companies are pulling workloads from public cloud due to cost and security concerns. Lower AI model costs will accelerate that. Memory chip demand will definitely rise - but I worry about the geopolitical risks with supply chains concentrated in Korea and Taiwan.
We welcome thoughtful discussions from our readers. Please keep comments respectful and on-topic.