Is Toshiba Going Fabless?

Could Toshiba be going fabless? It seems unimaginable, but an official Toshiba spokeswoman, Hiroko Mochida, said earlier this week that Toshiba is considering out-sourcing 28nm chips where demand exceeds Toshiba’s capacity.

 

However analysts like Yuichi Ishida, of Mizuho Investors Securities, point out that it’s daft to spend billions to make 28nm chips if you can outsource them. Ishida expects Toshiba to out-source all its 28nm production.

 

If Ishida is right, then Toshiba is going fab-lite after the 40nm generation. And fab-lite is a euphemism for ‘going fabless’.

 

If you’re not going to build a next-generation fab, then you’ll never build another leading-edge fab, and you’re on your way to fablessness.

 

It seems unimaginable that Toshiba, which had the world’s leading CMOS process in the 1980s, should now be considering giving up leading edge manufacturing, but it makes sense.

 

From an annual capex of over $3 billion for the last three years, Toshiba’s capex has been cut to under $1 billion this year. And you can’t build a leading edge fab for $1 billion or even for $2 billion.

 

AMD, Freescale, Philips, Infineon, NEC, Fujitsu, LSI Logic and many others have already taken the fab-lite to fabless route.

 

But no company which is primarily a memory company has ever gone fabless.

 

So Toshiba has to make the choice: Does it stay in memory and build new fabs? Or does it become an SOC player and go fabless?

 

Toshiba’s new CEO, Norio Sasaki, said last June he may give the go-ahead to the building of two 28nm NAND flash plants which had been postponed. He said it would depend on market conditions. And the NAND price is hardening.

 

But after losing $2.8 billion in the NAND market last year, and having to raise $5 billion of fresh funding in the debt and equity markets, Sasaki will be demonstrating real courage if he decides to build the fabs.

 

Sometimes, when you see the choices they have to make, you simply don’t envy these CEOs.

Tags: freescale, fujitsu, hiroko, ishida, lsi logic

Related Tech News

6 Comments

  1. David Manners
    September 18, 2009 10:57

    Ash, I think you may be right, but it’s something the damn-fool accountants will never understand.

  2. Ash
    September 18, 2009 10:10

    Thanks David. Although Process is just another differentiator the fact is that many IDMs have set their corporate strategies up around this very differentiator. I believe that IDMs may be misled in their thinking that’s its easy to shift to fab lite then fabless. Make no mistake, this is a change in corporate culture. Take away the innovation in process technology and you have to change your whole strategy to one of innovation in design (or IP) instead. Not an easy task I think.

  3. David Manners
    September 15, 2009 13:49

    It’s a good point Ash. If you give up process capability as a competitive differentiator you’re left with design capability and IP as the only differentiators. Well people like Qualcomm, Xilinx, Nvdia and Altera have done OK with this strategy becasue they’ve fought to protect their IP and design skills and have never had fabs. But companies going the other way – from being fabbed to going fab-lite en route to becoming fabless – may have a difficult time unless they already have a very strong IP positions or some unique ability in design. The enemy, as always, is commoditisation, and the only way to avoid commoditisation is differentiation. Process is just another differentiator.

  4. David Manners
    September 15, 2009 11:22

    It’s a very key point, Bob, I think a lot of people, particualrly in automotive, think the same. It’s another reason why fab-lite going on fabless may be a strategy which ends in tears.

  5. Bob Moreton
    September 15, 2009 11:19

    Speaking as an engineer I tend to avoid sourcing from fabless or fab-lite companies wherever possible just in case of supply problems further down the line.
    I doubt that I am alone in this approach.

  6. Ash
    September 15, 2009 11:46

    I was under the impression that companies ought to be keeping their core value-add in house and outsourcing those components that are commoditised. In other words shouldn’t semi vendors be investing in next gen fabs and outsourcing (selling off) their older tech fabs to the likes of TSMC etc … ? It would be interesting to compare the fixed costs of running existing older tech fabs versus the cost of building and maintaining a new next gen fab…

Share your knowledge - Leave a comment