IBM Paying Globalfoundries $1.5B to Take Unit in Retreat from Chips | Hacker News Reader

IBM Paying Globalfoundries $1.5B to Take Unit in Retreat from Chips

bloomberg.com

74 pointscolmvp44 comments

I think this story is a really good object lesson for the typical HN reader that there are mind-bogglingly huge businesses out there and they don't operate like startups.

IBM employees 61x the number of people as Facebook and makes almost 13x as much revenue.

If you can't figure out why IBM works and what it does, you need to keep studying how businesses work because you'll only ever understand small startups and be missing out on a big picture.

Even if all you ever want to work in are small startups, knowing how big businesses think can be really important to getting your company sold to one of them.

> IBM employs 61x the number of people as Facebook and makes almost 13x as much revenue.

The interesting thing to me is the ratio between those two numbers: IBM employs 4.7 people to get the same revenue Facebook gets with 1 (an even better comparison, of course, would be between payrolls and revenues, but I couldn't quickly find the payroll statistics for the two).

Why is this interesting vs not a particularly suitable metric for comparing them?

This may answer the natural question of "why not just shut it down, rather than pay to have it taken over":

    > In a 10-year partnership, Globalfoundries will supply IBM with
      Power processors in exchange for access to IBM’s intellectual property

I suppose it's one way to cut losses while not sacrificing short-term access to the technology; after all, 1+ billion is "only" one year of losses, according to the article.

An intriguing move in any case. I wonder if anyone can point to previous examples of a company paying to have a division acquired?

It's very common for companies selling divisions to sign pre-purchase agreements. This article's title and tone are at odds with how normal this particular deal is.

Thanks - that puts this in a new light. Based on what you are saying, this isn't purely "IBM Paying Someone $$$ to take off money losing division from their books" as it is, "IBM purchasing Processors, Chips, from Global Foundry, cost of such products being the net-worth of their manufacturing/engineering division + $1.5B"

> I wonder if anyone can point to previous examples of a company paying to have a division acquired?

Siemens paid BenQ for absorbing their mobile phone business (Siemens Mobile became BenQ Mobile, using the BenQ-Siemens brand).

One German news article [1] says that Siemens subscribed for new BenQ shares worth 50 million euros, while BenQ didn't have to pay anything (final paragraph).

Another article [2] states that Siemens agreed to pay 400 million euros to BenQ by installments, but held back the last installment of 117 million euros following BenQ's shutdown of BenQ Mobile in Germany (second paragraph).

[1] http://heise.de/-107908 or https://translate.google.com/translate?sl=auto&tl=en&u=http%...

[2] http://heise.de/-131172 or https://translate.google.com/translate?sl=auto&tl=en&u=http%...

In most of the civilised world there are significant costs to just shutting down a business especially redundancy costs for staff. This is in addition to considering outstanding debt in the business being sold and makes zero or negative purchase prices sensible in many scenarios.

I don't have a list but it doesn't strike me as at all strange or odd even where there is no ongoing commitment or contract between the companies as in this case.

IBM had a unique process call silicon-on-insulator that gives you something like an extra step in moore's law with regards to power and speed(but not cost). They also had their EDRAM technology , which is much more efficient memory type for cache.

This enabled them to build a huge(100MB) and fast(3TB/sec) L3 cache in their processors. This is something that could give a huge boost for many kinds of applications, for example databases d\and machine learning.

While intel has some edram capability, i haven't seem them release an integrated edram with the processor, not sure why.

And let's not forget, IBM now licenses the design for it's beast of a processor, the power8 ,in a modular format, enabling chip companies to easily add accelerators and innovate.

So it would be very interesting what intel's competitors will build with all those capabilities. I'm getting the popcorn.

I do believe SOI has some cost disadvantage, and this contributed to its unnpopularity. The POWER8 chips are huge and probably can't be made cheaply, or to consume sane amounts of power. IBM would have to take a very thin cut to compete with Intel.

The cheapest node today and for a long time is 28nm. But there are many applications where people are willing to pay more for performance, like processors.

Second there's a debate whether SOI is more expensive once it reaches volume. The claim vary between 10% more expensive , to cheaper , to something in the middle. We'll see.

Intel has included a 128MB eDRAM chip on-package (though not on-die) with some Haswell SKUs as a L4 cache for the processor and graphics systems.

It seems like this could pose some problem for the DOD, since IBM was the only trusted foundry with a cutting-edge process for logic. Presumably they don't want the government of Abu Dhabi fabbing chips for them.

Do you have a source for this? Afair, it was always Intel that was touting "made in US" on its marketing slides..

The majority of Intel's fabs are in the U.S., but they don't seem to be part of the DoD Trusted Foundry program. I'm not sure if there are barriers to them being certified, or if they just aren't interested.

The current list: http://www.dmea.osd.mil/otherdocs/AccreditedSuppliers.pdf

I believe that list is for fabs who can supply the DoD with custom/proprietary chips (i.e. with business models similar to TSMC or Samsung but for defense related products). Intel has not been in that line of business so there is no point in joining the program.

They did recently announce two fab customers, an FPGA manufacturer and then something else that is also low volume and obscure. So maybe the DoD can make it worth their while.

That one FPGA manufacturer, by the way, WAS started by people associated with the DOD intelligence apparatus. I can't remember the name of the company though, but they're making 22nm FPGAs on Intel's tri-gate process that they introduced with Haswell.

As of mid-2013 they had five fab customers [1]: Achronix, Tabula, Netronome, Microsemi, and Altera.

[1] http://www.oregonlive.com/silicon-forest/index.ssf/2013/07/i...

Intel generally does not fab other companies chip designs, with only a few exceptions.

Desperate to meet that EPS number for the 2015 roadmap, apparently.

This whole thing reeks of financial engineering; playing games shifting numbers from one bucket to another. After all, if ibm still needs chips -- and they're paying $1.5B, plus a 10 year contract -- how does really make sense?

Perhaps IBM figured that they don't need that many chips to warrant its own fab, which are expensive to maintain/upgrade to keep up with cutting edge process technologies.

I can believe that it made more financial sense for someone else to actually make the chips it wants (like nVidia, Apple and almost everyone else besides Intel and Samsung) than running its own fab in the long run due to economies of scale involved in fabs.

You might enjoy Robert Cringely's recent book (essay?) on the decline and slow-motion failure of IBM - http://www.cringely.com/2014/06/04/decline-fall-ibm/

Look at HP, which has taken a divestment / restructuring charge basically every year for a long time, each time labeling it a one-off thing.

Financial "engineering" is rampant all across listed American companies.

I was reminded of the "Real men have fabs" line and did a cursory search. Which hit a 1994 piece debating the issue, noting the downsides:

http://www.businessweek.com/stories/1994-04-10/real-men-have...

Research, chip capacity, other divestments. What is IBM these days? At a certain point, the name should just be retired and committed to the annals of history. At least Sun and SGI are (mostly) not brought up in any fashion related to their prior glory days.

I think they're betting on software and mostly on "cognitive computing" and watson. And they're focusing on that, leaving aside businesses that won't really give them a competitive advantage, like their fab.

If they can achieve what they promise, they'll regain their place in the history books.

I think you're greatly underestimating how large and successful IBM is. IBM is enormous in enterprise IT software and services. You just don't hear about it on places like HN where the focus is on consumer/startup stuff.

* Thinkpad and personal computer division sold to Lenovo * HDD division sold to Hitachi (now Western Digital) * Server division to be sold to Lenovo * Cell Processor production halted

In the meantime the IBM strategy seems to focus more and more on online services as the latest batch of acquisition seems to be mostly software and online services, for a complete disinvestment in hardware:

* Kenexa - software solution for HR * SoftLayer Technologies - hosting/cloud * Lighthouse Security Group - security around cloud hosting solution

The only significant hardware purchase seems to be of Texas Memory Systems, a SSD manufacturer. Which makes sense for server oriented business.

And yet they still do ~$100 billion in business per year. About 12.5 times the amount Facebook does, a company that can claim something like 1 out of every 6 people on the planet as a user.

They also employ 431,000 people, or about 61 times the number of people Facebook keeps alive in the global economy.

Several of the sold divisions were in fact money makers. They were sold because Big Blue could make more on services.

Maybe a third or less of the server division was sold. IBM retained Power servers, z mainframes (which aren't the same as Power), and storage.

Wooo. Maybe this will help GF provide AMD with process tech that isnt super old.

Agreed. AMD desperately needs to adopt cutting edge processes, even if it costs them 50 percent more per chip than it would if it would wait a year or longer.

It's bad enough that it doesn't seem to be able to outcompete Intel in CPU design anymore, but it's making the problem worse by not adopting a modern process, too (as modern as it can be, since Intel is already ahead of other foundries - although the gap will shrink a lot with the next-gen FinFET processes). That especially made it impossible for AMD to enter mobile or even ultrabooks, with AMD's old processes and higher power consumption.

Fortunately, 1) GF is apparently licensing Samsung's 14nm FinFET process, and 2) there's a tiny rumor that AMD will be using it by the end of next year, which could be very early in the process' life.

AMD is preparing to announce new CPU micro-architectures for both x86 and ARMv8 in a couple of years. If it actually puts them on a cutting edge process, and with a big push on the branding/PR side, the company might start to improve its chips' image (and sales).

IBM has sold off their PC business, they sold off their mainframe/server business; now they're selling off their chipmaking business.

What, exactly, does IBM think that its core business is? Making Jeopardy-playing AIs?

They still make mainframes. They sold off everything hardware wise except for that. Per the article they signed an 10 year agreement with GF to supply POWER chips.

IBM is a software, service, SAAS, PAAS, and consulting company based on their website.

IAAS too (IBM SoftLayer).

PCs and servers have never been a core IBM product. Their core product is the same as it's always been: enterprise consulting, and mainframes. Enterprise consulting is the biggest part now, although mainframes have long been effectively subsumed into that anyway (renting a mainframe is just part of a solution package).

"Enterprise solutions" (ie. consulting and expensive SaaS), as it has been for decades.

Things with better margins than hardware.

IBM has been exiting low margin businesses for years now so the chip announcement isn't that surprising. I don't see how cloud can be high margin because there is some serious competition in that market.

IBM has also been pushing hard into the information security market with the recent purchases of ISS (X-Force) and Q1 Labs (QRadar SIEM), both of which are insanely high-regarded products in the security space.

I spent 9 years at IBM and learned that, indeed, IBM does listen to its customers and try to give them what they want. But I also learned an unwritten rule: Customers should not call on a Wednesday because that ruins two weekends.

At one point, Gerstner said "IBM is the most inwardly directed, process oriented, arrogant company" he ever saw. Well, likely it still is.

Once I got an offer from IBM but turned it down, but with the offer came some advice: "You might think that IBM is an electronics company or a computer company, but you'd be wrong. IBM is a marketing company. IBM would get into the grocery business tomorrow if they saw a good business opportunity.

You might think that in IBM research comes up with new ideas, manufacturing turns them into products, and marketing sells them. Then you would have everything exactly backwards. Instead, marketing sees what they can sell at a profit and has manufacturing make it, and manufacturing goes to research if they need help.

What a product sells for is what the customer is willing to pay, and that has nothing to do with what it cost to make the product. In good cases, the profit margin is large, but at times it may even be negative."

Once I met with some guys who were doing some data analysis for the CEO and top level managers. Their explanation was that long the top managers would meet each year, announce their great results, predict even better results for the next year, and then beat those predictions. Then, for no obvious reason, at the annual meetings the top managers were not meeting their predictions, and that went on and on, for years. The only explanation was that God had ceased to smile on IBM."

Sure, IBM could have done Google. IBM Research was long just awash in people who could have done Google as a fast toy off the backs of their hands. But had they done so, manufacturing and marketing would have ignored the results. Marketing just would not have seen any big, traditional IBM customers in bank and insurance company back office paper pushing asking for a Google. Instead, a Google-like product in Research would have had to have been championed by the CEO or just left to die -- no one else with any product authority would ever touch any such thing.

Google? Well, that's about the Internet. Okay, early on the Internet was ARPANet. Later it was NSFNet, and then, or about then, it was run for NSF by IBM. So, right, in the early days of the Internet, IBM literally was running the whole thing. Or they had the ball in their hands and an open field to the goal line but all by themselves fumbled the ball, dropped the ball, tripped over the ball, fell on the ball, lost the ball, and ended up face down in the mud. Bill Gates, Larry Ellison, Page and Brin, Zuck, Bezos, etc. picked up the ball.

Reply on news.ycombinator.com