Will Demand For Local AI Prolong the Memory Shortage?

The memory shortage has been blamed on AI data center demand. But might the growing interest in running AI from your home make things worse? 

The CEO of Phison, a Taiwanese memory controller and NAND flash vendor, is predicting the shift toward running AI programs on local hardware risks pushing the supply gap out for several years, and even possibly up to a decade. 

“The AI demand is not going to slow down,” Khein Seng Pua told PCMag in an interview at Nvidia’s GTC event in San Jose, California. 

Pua spoke to us as news emerged that one of the leading memory suppliers, SK Hynix, is already projecting the memory shortage will last four or five years, rather than merely two. SK Hynix is pointing to a scarcity of wafers, the foundational base that chips are built on. 

However, Pua says growing interest in OpenClaw, the open-source autonomous AI agent that you can run over a PC, is another reason why the memory industry will struggle to address the demand in coming years. 

“In these few weeks, in China, OpenClaw is getting popular. It’s getting crazy, right? Don’t you believe that users will eventually install OpenClaw on-premise?” he asked. Pua noted that a recent Phison memory shipment to a Chinese PC vendor quickly sold out because Chinese customers were buying Intel’s “Panther Lake” laptops to run OpenClaw. 

At GTC, Nvidia’s CEO Jensen Huang even said of OpenClaw, “This is as big of a deal as HTML. This is as big of a deal as Linux. We have now a world-class open agentic framework.”

However, Pua sees a near future where a growing number of users will want to run more advanced AI models, including for video generation, that need more memory for the best performance, far beyond merely 16GB of RAM or 512 GB of storage. “In two years, the PC can definitely run AI inference, maybe OpenClaw or something,” he said. But this will create more consumer demand for memory when AI data centers will continue to gobble up the supplies.

Companies including Nvidia, AMD and Apple are starting to offer mini PCs and laptops with 128GB of RAM, capable of running larger AI models, but the products can range from $3,000 to around $5,000. Above Nvidia’s DGX Spark products are shown at GTC. (PCMag)

In the short-term, Pua said the memory shortage is forcing smartphone and PC vendors to settle for lower storage configurations. He predicts by Q3 more PCs will arrive with only 256GB of storage, down from 1TB. During this time, Nvidia will likely start shipping the company’s new “Vera Rubin” chips for data centers, which will absorb even more SSD chips, further worsening the memory crunch. 

Get Our Best Stories!

Your Daily Dose of Our Top Tech News

Sign up for our What’s New Now newsletter to receive the latest news, best new products, and expert advice from the editors of PCMag.

Sign up for our What’s New Now newsletter to receive the latest news, best new products, and expert advice from the editors of PCMag.

By clicking Sign Me Up, you confirm you are 16+ and agree to our Terms of Use and Privacy
Policy.

Thanks for signing up!

Your subscription has been confirmed. Keep an eye on your inbox!

“256 (GB) for PC users, is that enough or not? No,” he said. “But we think for these two years, you have no solution.” He expects the same memory tightening to occur for smartphones too, with configurations moving to as low as 64GB in storage.

Although the memory industry is investing in new fabs, Pua estimates the extra production will only amount to a 40% increase when they start coming online around 2028. “After two years, it’s 40%, but demand is already 2x,” he said. The memory industry will invest again only to see demand increase “3x,” he added. 

As a result, Pua expects the electronics industry will need “three cycles” of memory investment before supplies can truly meet the demand, meaning the shortage might span as long as 10 years. “So first cycle (of investment), second cycle, third cycle it may go (away). I’m saying may,” he emphasized. 

Phison showing their products running AI models at Nvidia’s GTC. (PCMag)

For now, Phison can only serve 30% to 40% of its customers’ needs. The shortage has been so severe that a Japanese elevator company called two weeks ago, asking for help securing a mere 500 units of SSD supplies. But to resolve the shortage, Pua says it’s ultimately up to the leading memory suppliers, such as SK Hynix, Micron and Samsung, to invest in new capacity. 

That’s why Pua has been pointing to the growing demand for running AI locally when talking to the memory suppliers. “They have to invest more because of market need, but when they invest more, they say ‘Oh, how about next year, market oversupply?’ I’m telling them ‘No need to be afraid because edge device AI is coming. So please invest more,’” he said.

Phison has already been trying to lay the groundwork by developing memory systems that can run AI programs even on lower specs. “It means you start educating the users doing AI on-premise. But unfortunately, 16GB with 1TB of SSD is only ‘good-to-use,’ not the best performance,” he said. “Why we’re using low capacity is because of the shortage. We enable this market. Two years later, when DRAM has a new output, this will go to 64 (GB), and 2TB or 4TB.”

“If you are the memory company CEO, are you going to take this story or not?” he asked, pointing to the revenue opportunity for the industry if tens of millions of PC shipments switch to higher memory specs.

Pua also half-jokingly said he only sees a slow down in AI investments if World War III occurs. But a more realistic scenario is oil prices skyrocketing due to the ongoing US-Iran war, although Phison’s CEO expects President Trump will de-escalate before things spiral out of control. “He cares more about Wall Street,” Pua said. 

About Our Expert

Michael Kan

Senior Reporter

---

Experience

I’ve been a journalist for over 15 years. I got my start as a schools and cities reporter in Kansas City and joined PCMag in 2017, where I cover satellite internet services, cybersecurity, PC hardware, and more. I’m currently based in San Francisco, but previously spent over five years in China, covering the country’s technology sector.

Since 2020, I’ve covered the launch and explosive growth of SpaceX’s Starlink satellite internet service, writing 600+ stories on availability and feature launches, but also the regulatory battles over the expansion of satellite constellations, fights with rival providers like AST SpaceMobile and Amazon, and the effort to expand into satellite-based mobile service. I’ve combed through FCC filings for the latest news and driven to remote corners of California to test Starlink’s cellular service.

I also cover cyber threats, from ransomware gangs to the emergence of AI-based malware. Earlier this year, the FTC forced Avast to pay consumers $16.5 million for secretly harvesting and selling their personal information to third-party clients, as revealed in my joint investigation with Motherboard.

I also cover the PC graphics card market. Pandemic-era shortages led me to camp out in front of a Best Buy to get an RTX 3000. I’m now following how President Trump’s tariffs will affect the industry. I’m always eager to learn more, so please jump in the comments with feedback and send me tips.

Read Full Bio