|That Digitimes report is absurd, IMO. Intel made the decision but was explicitly non-committal in their CC just the day before? The ML analyst seems to me to have nailed the fab situation pretty well in this report copied from the IV thread.|
ML Note on New Fab
New Flash Venture: Expect only minimal impact in 2011-12
Toshiba and SanDisk jointly announced plans for a new fab construction in
Yokkaichi, Japan. This is the third 300mm wafer fab line for the companies under
a new JV entity (50.1/49.9% stake, Toshiba/SanDisk) called New Flash Venture
or Fab 5. That said, it is not a new development as Toshiba has been alluding to a
new shell fab construction since last year, while SanDisk was talking about potential
NAND supply shortage. We have already factored this in to our global NAND supply
model, but still only expect a minimal impact in 2011-12, due to: (1) construction and
initial ramp-up periods (in 2011), and (2) execution risks from using new nodes
(2012). Further, potential budget constraints (Toshiba’s high debt, SanDisk’s
disciplined capex) may delay full utilization of the shell fab even in 2013.
Potential obstacles to the new fab’s full utilization
Toshiba and SanDisk’s second NAND fab is still underutilized – about a half clean
room space is available to add new equipment and tools (incremental capacity:
about 110K wafers per month if US$4bn were spent). Further, Toshiba and
SanDisk need to spend at least US$1.5bn a year to upgrade existing capacities
competitively (Fab 3: 154K wpm; Fab 4: 113K wpm). The third NAND fab will
require about US$9bn to run at a full scale (220K wpm). Net-net, the new fab plan
announced by the companies should be a part of their long-term plans under JV
agreements, and we believe this will not result in a NAND glut, at least in 2011-12.
Samsung’s NAND capex increase won’t be irrationally high
Despite Toshiba camp’s new fab construction, Samsung will not increase its
NAND capex (about US$3.5-4.0bn a year), in our view. Instead, it has recently
addressed new logic fab construction in the US via a NAND shell fab. Samsung’s
currently available clean room space in Korea (line 15) will be used for DRAM.
Thus, the company’s new NAND capacity should come from its underconstruction
shell fab (line 16), which will be operational from late 2Q11.
Enough demand to absorb increase in chip production
Current NAND spot price is much higher than chipmakers’ 3Q projection made six
months ago (16Gb MLC: US$4.5 vs US$3.0-3.5) or our estimate (about US$3.5).
This clearly indicates supply shortage, particularly to meet demand from Apple
and smartphone makers. Although we model about 15% QoQ NAND ASP decline
for the industry in 2H, recent data points (solid demand despite rising macro
concerns, lean inventories in the channels) clearly indicate upside to our 2H
estimates. That said, Toshiba is not our top pick in NAND due to its low-margin
profile for non-NAND products (eg, loss from logic chips and PC). In fact, SanDisk
is our best pick for NAND.