AMD completes spin-off of its assembly and test operations

A few months ago, AMD announced that it was spinning off its chip assembly, test, mark, and pack (ATMP) operations into a joint venture with Nantong Fujitsu Microelectronics (NFME). The companies jointly announced the successful closure of that deal today. AMD will receive $371 million from NFME as part of the bargain. In exchange, NFME will get an 85% share of AMD's ATMP operations in Penang, Malaysia and Suzhou, China. AMD will retain a 15% stake in those businesses.

A worker tests chips at AMD's Suzhou facility

AMD CEO Lisa Su says the deal "marks another step in building a more focused AMD as we complete our transition to a fabless business model, enhance our supply chain operations, and further strengthen our financial position." After taxes and other expenses, the company will get a $320 million shot of cold, hard capital. It also expects the move will "significantly reduce" capital expenditures. AMD will continue to use the joint venture's ATMP services going forward.

Comments closed
    • Kretschmer
    • 4 years ago

    Soon AMD will be comprised entirely of executives and janitors. “Sell everything that isn’t bolted down” rarely results in a successful corporate transformation.

    • ronch
    • 4 years ago

    What chips are those shown in the photo? Opterons? The dies seem big!

      • JumpingJack
      • 4 years ago

      yep, looks like it… probably the entire batch they will sell all year.

    • kuttan
    • 4 years ago

    AMD will receive $320 million ?? So finally we may see net-profit for AMD in Q2 2016 😛

      • Khali
      • 4 years ago

      I wouldn’t count on it. One of those “one time charges” is likely to pop up and eat the whole amount. That, or part of it will be given out as bonuses to the already over paid management team.

      • NoOne ButMe
      • 4 years ago

      We will see a GAAP net-profit. Their non-GAAP net-profit is still unknown. AMD has claimed this move will help lower their SG&A/R&D (I gather this as AMD has said 1,700 people will move to the new company). So it might improve their non-GAAP profits enough for them to be in the black.

    • puppetworx
    • 4 years ago

    Did these guys deal with the interposer on Fiji or was that outsourced?

      • NoOne ButMe
      • 4 years ago

      Fiji is done by Amkor I believe.

    • blastdoor
    • 4 years ago

    I wonder when the asset purge will end…. maybe when AMD is left holding the x86 license and patents on x86-64 (etc), at which point they’ll put the license through the shredder and start charging Intel a fee for x86-64?

      • JumpingJack
      • 4 years ago

      AMD has various components of the IP around the 64-bit extensions, but the majority of the x86 IP belongs to Intel.

      • Deanjo
      • 4 years ago

      [quote<] start charging Intel a fee for x86-64?[/quote<] Those AMD patents expire in a matter of months (in 2017).

        • maxxcool
        • 4 years ago

        Oh that’s right they do don’t they.. Aught to be interesting.

          • Deanjo
          • 4 years ago

          Ya, their x86-64 patents are pretty much worthless at this point for generating revenue, some have already expired and the rest soon to follow. The only real instruction set patent that has any length of life on it is SSE 4a which intel doesn’t even bother with supporting anyways.

        • BurntMyBacon
        • 4 years ago

        Didn’t know that. When do Intel’s x86 patents expire? Or are they covered under some different arrangement?

          • Deanjo
          • 4 years ago

          intel’s actual 32-bit x86 patents expired years ago. The patents that they are licensing out to AMD are the extended instruction sets such as SSE2/3/4 (SSE2 expires soon as well), AVX etc. Once the current cross patent deal with intel expires (also in a few years), AMD doesn’t have anything to offer intel with regards to licensing of instruction sets and will likely have to start paying intel a healthy sum to be able to continue offering support for those newer instruction sets.

            • maxxcool
            • 4 years ago

            Ouch. queue the buy-out rumors or all in on the ARM wagon rumors …

            • Deanjo
            • 4 years ago

            Oh I imagine that the recent rumours of AMD trying to license out graphics IP to intel is an effort to keep licensing costs balanced out.

            • DrDominodog51
            • 4 years ago

            So anyone could make a PIII (MMX, FPU, SSE1) level x86 compatible cpu at this point?

            • Deanjo
            • 4 years ago

            Yes

    • NoOne ButMe
    • 4 years ago

    And there goes any real packaging advantage AMD had.

    For those who don’t know what I mean by packaging: [url<]https://en.wikipedia.org/wiki/Integrated_circuit_packaging[/url<]

      • chuckula
      • 4 years ago

      Nonsense, [url=http://www.legitreviews.com/images/reviews/1741/amd-fx-metal-tin.jpg<]AMD's metal tins still look great.[/url<]

        • Neutronbeam
        • 4 years ago

        Those tins ARE great. I use mine to hold my spare Intel CPUs and Nvidia cards. 🙂

          • jihadjoe
          • 4 years ago

          Quake tin is best.

            • meerkt
            • 4 years ago

            ref: [url<]http://media.shpock.com/Quake-III-Arena-TIN-BOX-Team-Arena-PC.jpg[/url<]

            • Spunjji
            • 4 years ago

            Quad Package?

    • Mr Bill
    • 4 years ago

    How does one transition to a fabless business model and still design cutting edge computational devices? When AMD had Jim Keller to design Zen for example, do you need to be very aware of the fab process or do you just leave implementation up to the fab?

      • NoOne ButMe
      • 4 years ago

      Like how Qualcomm, Apple, Nvidia and a few others do?

      You design for the process. Look at the RV770/870 stories that Anandtech did a while back.

      • DPete27
      • 4 years ago

      ARM processor architectures are pretty successful….they’re a fabless company.

        • chuckula
        • 4 years ago

        ARM has never sold a single processor.
        It’s not their business model to do so.

        AMD couldn’t adopt ARM’s business model even if it wanted to since AMD doesn’t have its own unencumbered processor architecture to go out and license.

          • DPete27
          • 4 years ago

          I’m aware. I used ARM because they may be one of the most extreme examples since they haven’t sold a single processor. Yet they still have the ability to sell you all the blueprints you’d need to fab one. Point is, you don’t need your own in house fab to design computing chips.

          I’m a structural engineer, my company doesn’t have it’s own fabrication shop, and yet I sit at my desk every day and design bridges. Heck, many engineering consultants don’t even employ their own crews to build/install their products.

            • smilingcrow
            • 4 years ago

            The question is does having the process design team working directly solely with one architecture design team lead to ‘significant’ advantages versus a generic process aimed at multiple architecture clients and designs?
            Intel would have us believe that is true but who knows!

            • DPete27
            • 4 years ago

            Keep in mind, up until recently, Intel has been a process node step ahead of everyone else. They can pound their chests all they want about process and architecture teams working closely, but it’s not that hard to come out ahead when you can cram more transistors onto your die than the next guy can. (not to mention R&D budget)

Pin It on Pinterest

Share This