The logo of an Apple Store is seen reflected on the glass exterior of a Samsung flagship store in Shanghai, China Monday, Oct. 20, 2025. Wang Gang | Feature China | Future Publishing | Getty Images
The cost of your smartphone might rise, analysts are warning, as the AI boom clogs up supply chains and a recent change by Nvidia to its products could make it worse. AI data centers, on which tech giants globally are spending hundreds of billions of dollars, require chips from suppliers, like Nvidia, which relies on many different components and companies to create its coveted graphics processing units. But other companies like AMD , the hyperscalers like Google and Microsoft , and other component suppliers all rely on this supply chain. Many parts of the supply chain can't keep up with demand, and it's slowing down components that are critical for some of the world's most popular consumer electronics. Those components are seeing huge spikes in prices, threatening price rises for the end product and could even lead to shortages of some devices. "We see the rapid increase in demand for AI in data centers driving bottlenecks in many areas," Peter Hanbury, partner in the technology practice at Bain & Company, told CNBC.
Where is the supply chain clogged?
One of the starkest assessments came from Alibaba CEO Eddie Wu, CEO of Chinese tech giant Alibaba . Wu, whose company is building its own AI infrastructure and designs its own chips, said last week that there are shortages across semiconductor manufacturers, memory chips and storage devices like hard drives. "There is a situation of undersupply," Wu said, adding that the "supply side is going to be a relatively large bottleneck." He added this could last two to three years. Bain and Co.'s Hanbury said there are shortages of hard disk drives, or HDDs, which store data. HDDs are used in the data center. These are preferred by hyperscalers,: big companies like Microsoft and Google. But, with HDDs at capacity, these firms have shifted to using solid-state drives, or SSDs, another type of storage device. However, these SSDs are key components for consumer electronics. The other big focus is on a type of chip under the umbrella of memory called dynamic random-access memory or DRAM. Nvidia's chips use high-bandwidth memory which is a type of chip that stacks multiple DRAM semiconductors.
watch now
Memory prices have surged as a result of the huge demand and lack of supply. Counterpoint Research said it expects memory prices to rise 30% in the fourth quarter of this year and another 20% in early 2026. Even small imbalances in supply and demand can have major knock on effects on memory pricing. And because of the demand for HBM and GPUs, chipmakers are prioritizing these over other types of semiconductors. "DRAM is certainly a bottleneck as AI investments continue to feed the imbalance between demand and supply with HBM for AI being prioritized by chipmakers," MS Hwang, research director at Counterpoint Research, told CNBC. "Imbalances of 1-2% can trigger sharp price increases and we're seeing that figure hitting 3% levels at the moment – this is very significant."
Why are there issues?
Building up capacity in various areas of the semiconductor supply chain can be capital-intensive. And it's an industry that's known to be risk-averse and did not add the capacity necessary to meet the projections provided by key industry players, Bain & Co.'s Hanbur said. "The direct cause of the shortage is the rapid increase in demand for data center chips," Hanbury said. "Basically, the suppliers worried the market was too optimistic and they did not want to overbuild very expensive capacity so they did not build to the estimates provided by their customers. Now, the suppliers need to add capacity quickly but as we know, it takes 2-3 years to add semiconductor manufacturing fabs."
Nvidia at the center
A lot of attention is on Nvidia given it dominates when it comes to the chips that are being put into AI data centers. It is a huge customer of high bandwidth memory, for example. And its products are manufactured by TSMC which also has other major customers like Apple. But analysts are focused on a change Nvidia has made to its products that has the potential to add major pressure to consumer electronics supply chains. The U.S. giant is increasingly shifting toward using a type of memory in its products called Low-Power Double Data Rate (LPDDR). This is seen as more power efficient than the previous Double Data Rate, or DDR memory. The problem is, Nvidia is increasingly using the latest generation of LPDDR memory, which is also used by high-end consumer electronics makers such as Samsung and Apple . Typically, the industry would just be dealing with demand for this product from a handful of big electronics players. But now Nvidia, which has huge scale, is entering the mix. "We also see a bigger risk on the horizon is with advanced memory as Nvidia's recent pivot to LPDDR means they're a customer on the scale of a major smartphone maker — a seismic shift for the supply chain which can't easily absorb this scale of demand," Hwang from Counterpoint Research said.
... continue reading