Electronics Weekly Electronics Design & Components Tech News
“Fabless ain’t all it’s cracked up to be”, said Malcolm Penn, CEO of Future Horizons, presenting his evidence for that statement in the following bullet points:
26 of the fabless top 40 are over 20 years old.
The average age of the top ten fabless companies is 22.6 years
There’s only one fabless company in the top ten
There’s only one company in the fabless top 40 which is under 10 years old
So all the fabless companies in the fabless top 40 are old companies.
So beware the primrose path to fab-lite being urged on the industry by the analysts, the MBAs, the accountants and all assorted bean-counters – this has not proved to be the fast-track route to growth.
Qualcomm, Broadcom, Nvidia, Marvell, MediaTek, Xilinx, LSI Logic, Altera, Avago, Himax, Novatek, CSR, Qlogic, Conexant, Realtek, PMC-Sierra, Atheros, Megachips, Silicon Labs, Zoran, SMSC, DSP Group, AMCC, Semtech, SST, VIA, Etron, Zarlink, Silicon Image, SiRF, Lattice, Actel, Vitesse, ISSI, Sitronix, Wolfson, Power Integrations, Sigma Designs, Sunplus, Cirrtus Logic
David
I am interested in 40 fabless companies.
Could you send me what are the top 40 companies?
Thank you.
By interested fabless’ person
Yes Ian I met that chap – Hirokazu Hashimoto, CEO of Advanced Process Semiconductor Foundry Planning Company – on a boat on the Danube during one of Malcolm’s conferences in Budapest. Hashimoto told me: “Japan has 11 integrated device manufacturers at the moment but their fab capacity is not big and they can’t compete with TSMC which has 60,000 wafer-per-month fabs.” He said he wanted to raise an initial $1bn for a 10,000 wafer-per-month fab and then another $5bn to extend it to 60,000 wafers-per-month. It seemed to me a really smart way to go. I expect you’re absolutely right about the reasons why it didn’t work, but now that the Japanese companies are having to go to Taiwan, China and (maybe) Abu Dhabi for fab, I wonder if they’ll regret they were so intransigent with eachother. My understanding is you are absolutely right about the attitude of the Japanese government – they were all for it. Maybe it’s a model for Europe!
They could have — for example, ending up with a “Japan Foundry Consortium” — but they didn’t.
Personally I was surprised that this didn’t happen given the strong Japanese preference (consumer and business) for Japanese-sourced products. If I was “Japanese Consumer Electronics Inc.” I wouldn’t feel comfortable having my entire business — a large part of the GDP, which is why the government would be worried — dependent on an offshore foundry which I had no control over or alternative to.
My suspicion is that even if the Japanese government tried a lot of arm-twisting behind the scenes to make such a consortiom happen, there was just too much incompatibility and lack of agreement between the big Japanese IDMs to make it work.
Sharing a multi-billion-dollar fab sounds fine until you try and make it happen…
That’s interesting Ian. I think Malcolm’s take on this is that half a dozen companies of Fujitsu’s size could get together and share the costs of a fab over which they’d have control.
For 40nm and beyond it’s not really a case of IDMs going “asset lite” to make the balance sheet and ROI look better.
It’s simply that for each new process generation the cost of building a high-volume fab goes up, the turnover required to pay for this also goes up, and the number of IDMs (or foundries) big enough to pick up the tab goes down.
I work for Fujitsu which is certainly *not* an IDM run by bean-counters who don’t understand the technology and the issues of going fab-lite, but it’s public knowledge that we’re partnering with TSMC for 40nm and beyond — even though we’ve developed our own processes, it’s just too expensive for a “small” IDM (only a few billion dollars turnover 🙂 to build a large production fab.
Here’s a quote from an industry report:
At 22nm, the only companies globally that will have wafer manufacturing capacity for
nonmemory products will be the following:
Lets look at it from antoher perspective: IDM’s that move towards being fabless or asset lite.
Malcom has already commented on this in an earlier blog. In these cases, going asset lite is really an exercise in making the balance sheet and ROI look better.
The semi supply chain has been dis-aggregating for years and it will not stop. You have to be big and have scale to be an IDM. No doubt the ranks will thin over the next decade and beyond.
The next big thing in the industry will be the outsourcing of the logistics function.
Since when has a league table of turnover been a good measure of company success?
I would have thougth things like return on investment, dividends and future prospects were better measures. A niche semi will never get in the top 10 but that does not mean it is a bad company.
Is Rolls Royce a really bad company. I do not belive it has ever been a top 10 car manuacturer!
What is not mentioned here is how many of those “old” companies have acquired younger fabless over the years to help them get into or maintain their standard in the top 40! Sticking it out on your own for 20+ years is not the only way to succeed.
For some reason I’m reminded of elderly people selling their houses so that they can benefit in their twilight years from the capital they have built up. Not that the analogy is at all applicable here, you understand.
David,
Not sure what to say… “lies, damn lies and statistics”
Most fabless companies start at ZERO revenue, so lets assume $20M rev for their 2nd year of operation and grow this at 50% CAGR. After 10 years they will have about $500M revenue putting them barely in the top 20 fabless companies for revenue, and ALL they had to do is grow revenue 50%pa. EVERY year.
To get into the top 10 semiconductor they need to be 10 times this revenue so even if they continue at 50% CAGR it would take a startup 15 years to break into the top 10 IC companies.
The top 10 fabless companies have combined revenue of about $30B (or average of $3B) achieved in an average of 22years, This ONLY requires 30% CAGR (with no bad years, average across the group) WOW!
(BTW the average looks a lot better if you remove the dying IDM’s (AMD, LSI and Avago) from the fabless list.
One final point, that single company that is less than 10 years old achieved over 50% CAGR (this is unbelievably fantastic growth, for the years 2001-2010)