The problem of the missing fab(s) was sorted very quickly with the announcement that Toshiba - with SanDisk, its NAND flash manufacturing buddy - would build two, not one new monster-sized 300mm fabs as part of their plan for world domination in that market. 

Though site selection for the first fab has yet to be decided, and as a result the groundbreaking date, the stakes in the NAND flash market were raised once again and this time we could actually see some real consolidation occur down the road!

The reason for this is fairly simple.

Although Eli Harari, head honcho and leading light guru at the company rambled on and on during SanDisk’s annual analyst meeting earlier this week, every now and again he mentioned the near-term objectives of continuing to partner with Toshiba and the need to feed an almost bottomless pit of capital spending to ramp and build the biggest fabs in the world. All this is being done while continuing to push technology node migrations faster than the technology (lithography) is able to achieve!

This strategy is to allow Toshiba to become the world’s largest supplier of NAND flash, probably in about 2011-12, though Harari didn’t actually give a timeline for this. What he did say was that Toshiba wanted to be Number ONE and aimed to have a 40 percent market share!

This would in actuality be something like 20 percent as SanDisk would control the other 20 percent once Fab 5 was built and ramped to full capacity, hence the timeframes given above – more on that later.

My first reaction on hearing this via the webcast was to ponder who would have to exit the NAND market to kindly accommodate this world-beating pair? Micron? Don’t think so! Samsung (current Number 1)? Hardly!

So that would leave Hynix. Sorry Hynix, but destiny looms!

Of course, a few other scenarios are possible. One is where Hynix, Micron and Samsung’s market share in NAND takes a massive decline as the dynamic duo take the big part of the cake and no company raises the white flag. The other scenario is that it all goes horribly wrong for the pair, in which case you read the future here first!

But let’s step back a minute and actually look at the capital spending plans of SanDisk as revealed in the analyst meeting.

Firstly, I am surprised that SanDisk gave such a forward-looking and highly detailed insight into its spending plans. This gained my respect straight away, as there is a growing trend of announcing as little as possible for as long as possible within the semiconductor industry.

SanDisk actually spent more on capital equipment in 2007 than it had originally projected. This, according to CFO Judy Bruner, was due to changes in the tool layout at Fab 3, which enabled the partners to increase tool density in the cleanroom and boost overall productivity. More space allowed for more tools, which led to more spending.

That meant that instead of spending $1.375 billion in 2007, SanDisk spent $1.857 billion of which $898 million instead of the expected $800 million went to Fab 3.

Remember that these figures do not include Toshiba’s 50 percent spending input, so Fab 3 probably received about $1.6 billion worth of investment in 2007.

I did have a little chuckle over how Bruner then linked the layout improvements to the increased wafer capacity size of Fab 3, saying that Fab 3 was only expected to be able to crank out 135,000wspm but had now reached a ‘true’ full capacity of 150,000wspm (end of 2007). The funny thing is that SanDisk originally put the capacity closer to 110,000wspm when Fab 3 was first announced.

Then, when the first pictures appeared of the near-completed facility, it simply dwarfed the two 200mm fabs next to it and gauging its scale from these, it had to be the biggest fab ever built and in no way a 110,000wspm fab. Now we know that it was really a 150,000wspm fab in the first place!

So, some serious layout changes indeed! I would love to get a technical paper from Toshiba explaining that one!

But this doesn’t explain away all the spending in 2007. SanDisk highlighted that they had also spent $700 million on equipment for the mother of all fabs - Fab 4. This turned out to be significantly greater than initial projections for 2007. SanDisk actually projected only $275 million for Fab 4 in 2007!

Bruner said that this was due to the faster-than-expected ramp of the fab after only coming on stream in the fourth quarter of 2007. So, with yields high and wafers flowing, SanDisk had to match Toshiba’s actual ramp rate and spend nearly double on equipment than originally expected.

In 2008 SanDisk expects to spend a further $375 million on Fab 3 and a further $325 million in 2009. We can assume that some of that money is for tool upgrades etc., but I get the impression that Fab 3 will still see a higher capacity level than the 150,000wspm now claimed!

The majority of the investment for the next two years goes to Fab 4. $1.275 billion will be spent on Fab 4 in 2008 and $1.2 billion in 2009 (don’t forget that this is only 50 percent of the actual spending).

With the expected construction starting on Fab 5 in 2009, SanDisk expects to spend only $100 million on that fab. I would assume that this is for immersion lithography tools as the lead times are 10 months or more, and Toshiba - not SanDisk - is actually funding the construction!

See what I mean about detailed information?

But it goes further as Bruner showed a slide (see below) highlighting capital spending through 2010! Yes - a three-year forward projection of capital spending!

The presentation slide highlights that SanDisk’s capital spending has significantly increased year-on-year since 2005 and that it is expected to peak in 2009 at $3.0 billion. Though the 2010 spending figure is lower at $2.4 billion, spending only falls back to the 2008 expected level.

 In the five-year term, SanDisk would have spent $11.5 billion on capital expenditure (note: this does include final manufacturing).

Those among you who are eagled eyed will have noticed that the SanDisk slide points to a 25 percent capital investment structure for Fab 5, and not the 50 percent structure set with Toshiba for Fab 3 & 4!

It would seem that SanDisk had seen where the capex trend was heading as each node gets more expensive and fab scaling needs to be increased. This all leads to a deeper and deeper bottomless pit.

To get spending under control it has reduced is direct stake in Fab 5, yet will still gain access to 50 percent of the output. The extra 25 percent will be conducted on a ‘preferred foundry basis,’ according to Bruner, with an option to fully invest to the 50 percent level if that ‘preferred’ arrangement doesn’t turn out to be very preferential!

What that does mean is that SanDisk will pay cost plus something for 25 percent of the capacity out of Fab 5, something it hasn’t had to do so far. This is one little point that wasn’t lost on the financial analysts, I must say.

All these detailed capital spending projections highlight that market share gains, regardless of market conditions in the NAND flash space, is the number one game plan.

Looking at it slightly differently - in the way portrayed by SanDisk - the NAND flash market will continue to grow, and will the grow market share it needs to continue to build monster-sized fabs.  Spending on Fab 3 allowed SanDisk to gain a 15 percent market share. Spending on Fab 4 will give it a 17 percent market share and Fab 5 will allow for a 20 percent share of the NAND worldwide market.

When you add in Toshiba’s half (you didn’t forget, did you?) that’s a 40 percent market share.

The countdown has begun!