TrendForce: DRAM bit growth, a key barometer for chip market demand that measures memory produced, will rise 8.3% in 2023, the weakest on record
they are using extra exposure to make up for the lack of EUV,” says Douglas Fuller, an expert on the Chinese semiconductor industry. “But it is understood that the yield is terrible.” https://www.ft.com/... @ru_unforgiven2 : “Desktop processor shipments dropped to their lowest level in nearly three decades in the second quarter, according to Mercury Research. Total processor shipments experienced their largest year-over-year falloff since about 1984.” https://www.bloomberg.com/...
The article also bundles in the demand side: Mercury Research's finding that desktop processor shipments hit a near three-decade low, with total processor shipments falling the most since about 1984. That PC slump is what made weak bit growth a leading indicator rather than a footnote — and it is the trough against which the later recovery is measured.
First-order effects
Samsung, SK Hynix, and Micron head into 2023 with record-weak bit growth and collapsing pricing; Samsung's answer, per the related coverage, is to spend $30B+ in 2023 to grow capacity into the downturn.
PC-centric memory demand is the immediate casualty — Mercury Research's processor shipment data shows the desktop market shrinking to levels unseen in nearly 30 years.
Second-order effects
Samsung's counter-cyclical $30B+ capacity push pressures rivals to match spending or cede share, deepening the oversupply that drove the 2022 price collapse even as bit growth stalls.
The 2023 trough becomes the baseline for the AI-era repricing: TrendForce's later estimate that 2026 DRAM revenue will exceed 4x the 2023 trough, reaching ~$231B on HBM demand, only makes sense against this record-low growth year.
Third-order effects
The pattern — record-weak bit growth followed by a supply shortfall — points to the structural shortage now visible in coverage that global DRAM supply will meet only ~60% of demand through 2027, with memory climbing to ~40% of low-end smartphone manufacturing costs by mid-2026.
Memory's boom-bust cadence is tightening into what the corpus frames as a memory supercycle: the same industry that over-produced into 2022 under-invested for the AI demand wave, making capacity decisions in trough years the dominant variable for the rest of the decade.
The trend: DRAM is cycling from its weakest bit-growth year on record into an AI-driven supercycle in which trough-year capacity cuts, not demand, set the industry's ceiling.
Numerous chipmakers, particularly out of Korea and Taiwan, are sending warning signs about the health of the global economy. https://www.bloomberg.com/...
South Korea's exports are poised to drop next year if not within the next several months as demand for memory chips sours https://www.bloomberg.com/... https://twitter.com/...
The US might get that soft landing, but if chip demand is blinking red, global demand is sinking. Which may have second order effects that will eventually ripple through all economies https://www.bloomberg.com/...
Next year, demand for DRAM is likely to rise 8.3% and supply climbing 14.1% claims a forecaster. You can't use a memory chip as an accelerometer, for power control or Bluetooth tho. https://www.bloomberg.com/... https://twitter.com/...
'Total processor shipments experienced their largest year-over-year falloff since about 1984. It's a painful hangover following pandemic lockdowns, when the WFH trend spurred demand for PCs and other devices.' https://www.bloomberg.com/...
So a bunch of countries just subsidized more chip supply* at the same time that chip demand is turning south? *(Also possible that taxpayer funds for chips don't increase supply but shift where the supply is made. Plus new capacity takes A WHILE.) https://www.bloomberg.com/...
The global economy appears to be heading into recession, which would aggravate the downturn. There's also a broad push by governments to subsidize new factories, from the US and Europe to China and Japan. There are 24 major fabs under construction this year, well above average.
And must read commentary by @KathrinHille in @FT on China'a SMIC on whether the latest breakthrough on chips with lead to ramped up production allied to country's strategic goals https://www.ft.com/...
“A breakthrough by China's largest chipmaker revealed last month triggered a gasp of surprise among observers outside the industry. ... But how much time and money will SMIC sink into producing its new semiconductor at scale?” https://www.ft.com/...
“The furore over SMIC's progress is quite overblown — they are using extra exposure to make up for the lack of EUV,” says Douglas Fuller, an expert on the Chinese semiconductor industry. “But it is understood that the yield is terrible.” https://www.ft.com/...
“Desktop processor shipments dropped to their lowest level in nearly three decades in the second quarter, according to Mercury Research. Total processor shipments experienced their largest year-over-year falloff since about 1984.” https://www.bloomberg.com/...
Relevance of heterogeneity can't be understated. Semiconductor industry moving further into gluts but still not where shortages were most inflationary. How much did hot 2020 demand for electronics actually set off shortages elsewhere (vs other causes)? https://www.ft.com/... http…
Further fuel for my speculation that the CHIPS and Science Act won't do much to affect the supply of actual chips. Semis have always been a cyclical business. https://www.ft.com/...
TL;dr: Covid hoarding has accentuated the typical boom-bust chip cycle... and supply-chain issues still exist bc Covid still exists. US chipmakers hit by sudden downturn after pandemic boom https://www.ft.com/...
After Congress passed $52 billion CHIPS subsidy: -Micron, which had said it would invest $40b in the U.S. this decade, announced it will cut capital spending -Intel cut its capital spending plans by $4b, but predicts a “growing dividend” for shareholders https://www.ft.com/...
Sounds like the CHIPS Act is turning into a subsidy for semiconductor companies that are missing revenue targets and slashing budgets, rather than investing in new domestic fabs. Intel is still promising shareholders a “strong and growing dividend.” https://www.ft.com/...
I would argue that the focus on buybacks (the symptom) should be reallocated towards whether firms follow through on their capex commitments. Among the objectives of these subsidies is to help crowd-in and secure additional real fixed investment https://twitter.com/...