Why AMD is set to go fabless

2 Apr 2008 | 09:50 BST

By Charlie Demerjian

AMD IS GOING TO go fabless. The chip maker is likely to split into a fabless semi outfit and a foundry. The interesting parts are in the details, and there are a lot of potential ups and downs.

According to our research, the impetus for this move comes from two sides: the desperate need for money and the oodles of capacity they firm has. Fab 36 is running at a decent enough clip, but 38 next to it sitting almost idle. AMD could equip it, fire up the lines and make chips, but for what buyers? With Luther Park options coming up soon, they will have even more capacity than before, and a huge capital outlay to make the fab. They don't need the capacity and don't have the capital.

AMD could simply abandon the Luther Park plans, but there is a billion dollar subsidy attached to it, and that is a lot of money. Even if they only make the fab and sell it, the potential to pocket the cash remains.

So, what do you do? Spin off the fabs to a separate company and sell that to pay off your debt. Silly as it sounds, they did much the same with flash and Spansion. AMD has two state-of-the-art-ish fabs in Dresden and a billion dollar off coupon for upstate NY. Together, they are worth a lot of money, and selling them would pay off the debt AMD has amassed.

That is exactly what they are going to do. The real question is to whom, and that is easy - Abu Dhabi. The emirate already owns almost 10 per cent of AMD, and is awash with money. AMD may not be an AAA investment, but unlike most companies, it has something to offer any country that buys it, technology.

There are few players in the world that can make 45nm parts, and fewer still that are for sale. A buy like this could be the catalyst for the founding of a national technology program, or simply a few billion dollars for bragging rights. Either way, think ulterior motives that are not immediately monetary.

AMD would split into American Micro Devices, the design firm likely run by Dirk, and Arab Micro Devices, coined by Theo and run by Hector. The influx of cash would be sufficient to pay off the amassed debts and bankroll some of the future plans. Lets call the child firms AMD Design and AMD Foundry.

Simple as this sounds though, there are some snags. First is the Intel license, it stipulates that there is a 20 per cent cap on non-AMD produced wafers, and Intel probably isn't in a mood to loosen those rules. That means that AMD Design would have to own at least 51 per cent of AMD Foundry until the contract is renegotiated sometime around 2010.

Because of this, AMD would only be able to cash in on half the value of the fabs immediately. This may seem like an impediment, but that 49 per cent is still potentially worth billions. The selling price would likely fall into the 'enough' category. In 2010, the negotiations begin anew and any new deal would probably not have that clause, allowing the other half to be sold off.

Other interesting details are what happens to the fab partners, namely Chartered and IBM. They both make wafers, and neither is looking particularly healthy. One source suggested that Foundry could subsume the businesses of IBM and Chartered, making the resultant company a huge player in the field.

This makes a lot of sense, both from the size perspective, and from the IP perspective. AMD is good at SOI wafers but has very little experience with bulk of late. About 5 per cent of the world's wafer need is SOI, and Foundry would have three mega-fabs pumping out SOI, more than anyone will ever want.

If they suck in IBM and Chartered, they get bulk IP, and the ability to do more mainstream wafers as well. Basically, they will have enough product breadth to go around and make a viable company.

Other sources also mentioned something that is sure to come up in the conspiracy side of things. Intel has a lot of investments in Israel, and by some accounts is facing boycotts by countries that do not like the Israelis. On the ground, this doesn't seem to have slowed down any purchases, and is mainly rhetoric.

With Foundry being owned by in large part by Arabic interests, it sets the conspiracy mill churning at a mind-numbing rate. It is unlikey to affect more than a handful of sales, but the war of words is going to be astounding, count on it.

At the end of the day, you will have two companies, and one will own the majority of the second. With luck, Design will take the cash injection and run with it, the albatross removed from around its neck. Foundry could become a major player in the semiconductor world with five large and state of the art fabs, along with a bunch of tier one customers from the start.

If all goes wrong, you will have two companies that ride each other into oblivion. Initially, if Design can't make designs people want, Foundry will have nothing to make, and will have found, pun intended, the weak spot of this plan.

Whatever happens after the deal is done, we are very confident that this deal will happen. The company will be split, the business will be divided as specified, and the cash infusion will alleviate the most immediate concern of the company.

From that point, it is simply up to them to do the work, design the parts, and fab them. The slate is clean, AMD Design and AMD Foundry will simply have to execute from that point on, or it was all for naught. µ

See also
AMD's Sanders: real men don't need fabs

© 2007 Incisive Media Investments Ltd. 2007