When we play back the videos from 2004, and compare them with the flash situation today, much of the market has changed and much has not developed as some thought it would. However, after all the bravado, new generations of phones and changes in the flash market, the landscape is remarkably unchanged in fundamental ways. Today, the lines of demarcation separating NAND–appropriate applications from NOR-appropriate ones seem to be very much static and stable, with each safe in its own domain, having been defined by some of the events and non-events of the past 4-5 years.

Though NAND made some inroads, it has not yet substantially displaced NOR in phones, as some expected. NAND has, however, caught the major flash GB growth, where it has probably outgrown NOR by 50x in the past 4-5 years. NAND Flash also has taken the phone flash in new directions, such as add-in cards, and has benefited more from advanced phone features like cameras, internet download capability and visually-intense applications. But, phones are still NOR's greatest market opportunity, by a wide margin.

Here's a Quick Summary of Developments over the Past Few Years:
NOR, with XIP and standardized multi-sourcing, is tenaciously holding onto much of the internal flash guts in cell phones, though almost all the flash MB/GB growth inside and outside the phoneset is NAND Flash.

NAND has grown in GB shipments, breadth of applications, GB/application (surprisingly, almost without seeming limit), and price reductions beyond almost anyone's expectations....prices down 10x since 1/07 and 60x since 1/04; GB growth by more than 100x since 1Q04. By contrast, NOR prices/GB have declined about 15%/year for the past five years and GB shipments less than doubling in the same timeframe.

NOR's GB growth today has been a rather pedestrian 20%/year for the past five years, against NAND's 150%/yr. in 2008 alone (slightly smaller than in 2007). For the past several years, NAND has consistently been more profitable than NOR. That will change in the current 4Q08, as NAND makers (like NOR) have too much supply in play against not enough demand. NAND makers' cost- and price-cutting has also run into the reality of running a (supposedly) profitable business.

Industry Investment Levels: NAND has consumed $8-10B/yr in CapEx and $2-3B in R&D/yr, in a market that is about $14B/yr, and virtually profitless in 4Q08. In fact, one might guess that the NAND industry capacity is close to 1M 300mm wafers/month (<~50nm or so processing), and has a Net PPE (approximating net asset value) of in excess of $20B. NAND has been the industry's biggest single sinkhole for CapEx for several years running now.

By comparison, NOR Flash has a pauper's Capex budget and is, for the most part, increasingly relying on partnerships with 'friends,' for new volumes of sheer manufacturing capacity: Spansion with SMIC and Numonyx with Elpida. NOR CapEx in 2008 was probably less than $1B for all suppliers combined.

NOR Fab Strategies: Both Numonyx and Spansion have variants of Fab-Lite fab strategies and are starting to lean heavily on their foundries (Elpida's 300mm line) and SMIC, instead of 'Building Their Own So They Will Come." In addition, neither has the cash or the future wafer requirements to justify building anything but a proving ground for their leading edge parts, and making their existing fabs more cost effective.

Spansion took over ownership of the same fabs they had before divestiture from AMD/Fujitsu, plus added a small 8K 300mm wafers/mo fab in Japan, SP1; Numonyx took over an Intel fab in Israel, plus an unequipped shell from STMicro in Italy.

Migrating their fabs to finer line geometries and more up-to-date capabilities might be the biggest competitive threat they face. Neither company has significant excess cash, and both will increasingly depend on foundries for their growth and process improvements.

NAND Fab Strategies: "Build them and fill them" has been the NAND makers' mantra until very recently. Huge amounts of capacity have been added by existing players (Samsung and Toshiba-SanDisk) as well as relative newcomers to the top ranks, IMFT and Hynix.

Recent NAND fabs have been some of the most expensive and most advanced fabs ever built, expandable to more than 200K 300mm wafers per month, with many running 90K-100K wafers per month now. Today, about $3B is the price tag for a 100K/mo 300mm 45nm fab to build NAND. Today, the industry's processing capacity for NAND Flash, measured in silicon area, is probably 10x what it was on opening day in 2004.

Only recently have some 200mm NAND fabs been taken off line due to their lack of cost competitiveness, so the NAND production mix is increasingly "use-300mm-today-or-don't play."

Process Ground Rules: NAND has forged on to the point where it is today run on about 45nm average production ground rules.

NOR has somewhat closed the lithography gap with NAND where it counts most: in areas of dense, leading-edge products. This should make such comparisons as price/MB, relative cell size and performance more realistic and meaningful..., more "apples to apples."

NOR has been a perpetual technology laggard, but Spansion will have 45nm pre-production by year-end 2008. Both Numonyx and Spansion are running 65nm product today; both will ramp 45nm in 2009. It should be noted that the NAND market DEMANDS leading-edge litho, for simultaneous high-density and low cost, on average, much more than does NOR Flash; in NOR, there is a legitimate and economic reason for some NOR to never shrink below 110nm or 100nm (or to use Multi-level Cell...MLC... technology). With prices/GB declining at 15%/year, both major NOR flash makers should be able to cost-reduce with litho advances or technology fater than prices have been coming down recently...improving or restoring profitability.

NOR Shares: Numonyx and Spansion together account for about 70% NOR market share, and virtually ALL of phone NOR. They are also both now 'indies', having been spun out from Intel-STMicro (Numonyx), and from AMD-Fujitsu (Spansion). Both have been losing money steadily for several years. Despite these losing streaks, however, NOR is still firmly entrenched in the majority of cell phones (>80% of phones having NOR Flash is a good estimate), though NAND inside and as add-in cards, is faster growing.

The nichiness of the NOR Flash market can be appreciated in the fact that though everyone is losing money and the market is nearly flat in GB and down in revenues, there have been few wholesale market retreats by NOR vendors. Several other companies are still in the NOR business: SST and Atmel, NEC, Samsung, Winbond, Macronix, each of whom offers a mix of specialty NOR Flash and standard designs.

In many applications (and company organizations), NOR is seen as a non-volatile alternative for SRAM...quite differently from how NAND is viewed. Introduced more than a decade ago, a thriving 1Mb NOR flash market still persists from the mid-1990s. NOR products are longer lived than NAND could ever imagine.

Market Sizes: The NOR market will be about $6.6B in 2008, vs NAND's ~$14B. (NAND first surpassed NOR in 2005, with $10.6B for NAND to $8.0B for NOR). NOR was $7.7B in 2007; NAND was $14.5B. Almost all of NAND's GB growth since 2005...never under 150%/yr...has been offset by huge price declines, which have been far greater than those of NOR Flash in both dimensions of GB growth and price declines.

New Directions in NOR Flash: Both Numonyx and Spansion are trying to drive NOR or variants into non-traditional memory-needing markets, to try to offset GB growth stagnation in traditional NOR applications. Spansion's EcoRAM is the most noteworthy, being presented as a low-power alternative to DRAM in data servers (link). Spansion announced the industry's first CTF (Charge Trapped Flash) NAND, their ORNAND II in 3Q08, which they claim to be the fastest NAND device available.

Numonyx has a significant effort to develop PCM (Phase Change Memory) for next, or next-next, generation NVM technologies.

Numonyx is seen favoring multichip solutions for phone chip stack replacement, while Spansion is pushing hard to develop a single chip NAND-with-NOR- interface (ORNAND and ORNAND 2).

OneNAND (Samsung) and ORNAND and ORNAND 2 (Spansion) are attempts to drive derivatives of one technology into the others' basic applications space, with some success. Neither is second-sourced, and neither company seems to be looking very hard for one, either. It also seems, as in the case of many integrated functions chips, that neither has a roadmap or product that can keep up with the changing face of phone flash to actually reduce phone chip count laterally (meaning you might be able to sell it into a lower-end phone, but not displace the flash stack in a phone in the same market segment), namely, MB ratio of NOR to NAND, and performance. Most phones still have a multichip flash stack, and an increasing fraction have a slot for NAND Flash add-in card.

Multi-Bit Cells - More Similarities and Differences: Two-bit cells are more prevalent in NAND (maybe 95%), than they are in NOR (60-70%), due mostly to differences in cost-importance in the respective markets, and some technical issues. Both trends seem to have expanded about as far as it makes sense, based on market requirements. No one talks of x3 NOR Flash today, but basic market economic pressures will make x3 NAND Flash a major contributor to GB and revenues in 2009.

Sales levels, Flash Suppliers: Spansion will deliver about $2.45B in flash in 2008 (overwhelmingly but not exclusively NOR flash), and Numonyx will deliver more than about $2B, ranking them, among all flash players, smaller than only Samsung (~$5.5-6B) and Toshiba incl. $4.5B (incl. sales to SanDisk), but comparable to Hynix (<$2B), and IMFT (Intel + Micron), also about $2B this year. NAND is growing way faster in GB shipments, but revenues have eroded much faster, too, by price declines running more than 60%/year almost since market inception.

Samsung in NOR Flash and Differentiated Flash Products: Everyone always wants to know and understand what Samsung is doing, esp. in markets they may have turned up their noses at in earlier days . It appears that Samsung's "NOR offensive" from 2006 and last year has been paused, though they remain the third largest NOR supplier today, with about 10% market share.

Samsung has been able to keep its NAND GMs up with 'fusion' differentiated NAND products, such as OneNAND (above) Flex-i-NAND and MoviNAND, but these seem to be less than about 10% of GB output.