Diving Deeper into the Component Shortage: DRAM and SRAM

Over the last year or two, demand for new technology has outpaced what many suppliers can make. Component demand rose with the spread of the Internet of Things and with growth in the automotive, mobile and industrial markets, EPS News reported in November 2017.

Cars carried more electronics, phone makers chased more capable devices, and tight supplies of capacitors and resistors were expected to last into mid-2018, EE Times reported in October. An October issue on shortages and manufacturing covered the wider picture; this one looks at the memory parts in short supply.

New Technology and Underlying Demand

Underlying demand drove much of the shortage. As electronics spread into more everyday products and companies competed to stay ahead, demand outpaced supply, and lead times that had been 12 to 15 weeks a year earlier reached 30 weeks or more, Supply Chain Connect reported in October 2017.

Apple’s product cycle added pressure. Analysts told Reuters in June 2017 that Apple buys about 18% of the world’s annual supply of NAND flash chips, and some memory chip prices had doubled or tripled from a year earlier.

Electronics makers usually built up inventory in the first half of the year to avoid being squeezed by Apple’s fall iPhone launch, Reuters reported, but the shortage in the first half of 2017 left many scrambling. NAND flash demand had exceeded supply every quarter since the third quarter of 2016, according to TrendForce.

Suppliers were also wary of overbuilding. Paul Romano, chief operating officer of the distributor Fusion Worldwide, told EPS News that component makers wanted to meet all demand without expanding into a market that then contracts. The industry remembered the early 2000s, when the dot-com market bottomed out and left the electronics supply chain with $13 billion of excess semiconductor inventory in 2001, EE Times recalled.

The DRAM Shortage

DRAM, the low-cost, high-density main memory in computers, servers and phones, came under pressure from both data centers and phones in 2017.

Server DRAM demand rose as Google, Facebook, Amazon and Microsoft built new data centers, and mobile DRAM prices were expected to rise 10 to 15% in the fourth quarter of 2017, according to TrendForce. TrendForce also expected the major DRAM suppliers to add little production capacity in 2018.

Flash memory was tight for a reason of its own. The move from 2D to 3D NAND manufacturing went more slowly than planned for suppliers other than Samsung, which TrendForce named as the main reason NAND supply stayed tight through 2017.

How SRAM Differs

Static random-access memory, or SRAM, is a close relative of DRAM. It stores each bit in bistable latching circuitry built from four or six transistors, where DRAM uses one transistor and one capacitor.

SRAM is faster than DRAM and costs more, so it is used where speed matters more than cost.

The shortage of 2017 reached well beyond memory. Lead times stretched and prices rose for passives and discretes as well, down to some basic transistors and diodes, and some memory makers were refusing new orders because of their backlog, Supply Chain Connect reported.

What Manufacturers and Planners Should Do

If the components a design needs are among those in short supply, waiting is the wrong move. Planners and buyers need to get ahead of demand, because 30 days is nowhere near enough time to place an order.

Lead times ran far past that. One distributor executive told Supply Chain Connect that lead times on some devices had gone from 10 to 14 weeks to 25, 30, 35 and even 40 weeks.

For memory, that meant placing orders months before a build rather than weeks, and raising any constrained part with the manufacturing partner while the design could still change. Kodiak’s part in that conversation is an early warning when a memory part on a build runs short, and an approved alternate source where the design allows one.