Intel CEO’s AI Hardware Shortage Forecast: What It Means for DDR4 and DDR5 RAM Prices

What Tan said, and where​

Tan spoke on October 12 at the inaugural DBS Asia Future Forward Conference in Singapore. DBS had billed the event as a day-long programme with AI at its centre. His session was a panel with Broadcom CEO Hock Tan, moderated by DBS CEO Tan Su Shan, on where the next trillion dollars of AI value could be created.

The widely repeated wording comes from a post on X by Theresa Perrin. TechRadar quotes her as saying Tan observed that "AI hardware will stay in short supply for the next 2 to 3 years." TechRadar adds that Tan tied this to "persistent constraints in the underlying tech supply chain." Several outlets reported the same gist. Investing.com, for example, said Tan stated that hardware will remain in short supply for the next two to three years because of the rapid AI buildout.

I found no official transcript or recording. Treat the exact phrasing as secondhand, as reported by press and social posts. One aggregator noted that no verbatim transcript of the panel has been publicly confirmed.

It's about AI hardware, not your DIMMs​

Tan's reported forecast covers AI hardware broadly. It doesn't say a consumer RAM kit will be scarce or expensive for three years. TechRadar's reading is that if AI supply stays tight, consumer RAM will be even thinner on the ground. That is plausible, and it's the outlet's inference rather than a Tan statement.

Other reports of the same panel add context that cuts against a simple "RAM doom until 2029" headline:

• Investing.com reported that Tan said the memory chip segment could be among the first to lose momentum. One market commentary linked that remark to Micron and SanDisk being among the worst-performing stocks that day. That link is the commentator's interpretation.
• Broadcom's Hock Tan, per CNBC's account relayed by Wolf Street, said "Memory shortage today might not be that way a year from now" as production capacity expands. He also expected wafer shortages to ease by 2028 or 2029.
• His view was that overall demand for compute infrastructure exceeds supply. He said that gap could last through 2029 and possibly into 2030.

Those are Hock Tan's comments, not Lip-Bu Tan's, and they're only as reliable as the secondary reports. Still, the panel's tone was mixed on memory. Tight supply was expected, but at least one executive suggested memory might be the first segment to ease.

The bigger theme was money​

Most coverage of the panel focused on financing. Lip-Bu Tan said the front-line challenge is capital and its cost, and that in the longer run people want to see a return on investment. Hock Tan argued that capital, not chips, memory or power, could become the biggest constraint. Lip-Bu Tan said he saw no significant slowdown in the next two to three years.

Read together, the two-to-three-year line is a statement that AI demand stays strong. It isn't a pure supply-chain warning.

The Barclays price figures​

TechRadar pairs Tan's remarks with a Barclays report on memory pricing. Wccftech's coverage of that report gives the detail:

Wccftech's report also says DDR5 spot prices rose 8% month over month in October. The Barclays report itself wasn't available to me. These numbers come from press coverage of it, relayed via a post on X.

Two cautions apply.

• Contract prices are not retail prices. They are what large buyers such as OEMs pay chipmakers. A 24% rise in DDR4 contract pricing does not mean a 24% rise on a retail DIMM. Retail often lags or diverges.
• "Through 2028" is not "until 2029." Barclays reportedly sees Micron's DRAM supply staying constrained through calendar 2028. TechRadar's later claim that supply stays short until 2029 is its own extrapolation. I couldn't verify its figure of DDR5 being up 544% in Germany, so I'm leaving it out.

TrendForce's public price pages show current spot and contract data. They do not confirm Barclays' specific three-month figures, so I won't use them as corroboration.

What PC buyers and IT admins can do​

This is general industry reasoning, not from the sources.

1. Budget for memory as a volatile line item. If you refresh fleets in 2027, price RAM configurations close to purchase time rather than from last year's quotes.
2. Watch for de-speccing. Wccftech notes Barclays flagged customers cutting memory specifications as a stopgap. Check RAM capacity on laptop and desktop models before bulk orders, since an identically named SKU may ship with less.
3. Right-size rather than overbuy. Audit actual memory use on your Windows 11 fleet before standardizing on larger configurations.
4. Don't panic-buy on a forecast. A CEO's panel remark is a signal, not a price guarantee. Other panel reporting suggests memory could ease sooner than the broader hardware picture.

Reading the forecasts with a critical eye​

Tan runs a company with a stake in AI-driven CPU demand. According to Wccftech, Intel's Data Center and AI division grew revenue 59% year over year in the second quarter. A prolonged AI buildout suits that narrative. Hock Tan's Broadcom is also deeply involved in financing AI customers. Conference remarks from executives are not neutral market forecasts.

On the other hand, the reported analyst outlooks and the Barclays pricing data point the same way. They suggest tightness will last well into 2027 and probably 2028. The most defensible takeaway is this: expect continued pressure on memory and other AI-adjacent components. The duration and the size of consumer price effects remain uncertain.

Bottom line​

• Tan reportedly said AI hardware will stay short for two to three years. That forecast is about AI hardware overall, and it is unconfirmed by any transcript.
• Barclays-attributed contract-price gains of 24% for DDR4 and 20% for DDR5 over three months are real signals. They don't translate directly to retail.
• Other accounts of the panel suggest memory could cool before other parts of the AI supply chain. Financing, not only supply, is a central worry.
• For Windows buyers, the practical response is careful budgeting and spec-checking, not alarm.