Intel pushes deeper into HPC market through InfiniBand acquisition

Intel is well known for its consumer PC chips and for its business solutions, which defined the Wintel era. Intel also has a strong networking portfolio. Long a dominant player in the adapter/LOM space, Intel has been aggressively expanding its position to adjacent technologies. Last year, Intel acquired merchant silicon vendor Fulcrum Microsystems and today announced that it is acquiring QLogic’s InfiniBand business. The acquisition is for $125M in cash; QLogic originally acquired the InfiniBand technology as a $60M purchase of SilverStorm in 2006 [Update: and $109M on PathScale also in 2006]. Mellanox is the market leader in InfiniBand, and it added Ethernet to its portfolio last year.

Despite Cisco’s efforts to eliminate the InfiniBand market – it bought Topspin in 2005 for $250M and subsequently ended support of those solutions – the IB market is still doing well in the HPC marketplace. Intel has been making a push into the HPC market and InfiniBand solutions are an important piece to broaden Intel’s ability to meet customer requirements in this space.

This acquisition is not negative on QLogic’s position in the marketplace. From the product set that QLogic offers, FC and Ethernet have obvious overlaps, while InfiniBand has a different sales, marketing and partner engagement. QLogic can stay focused on its core storage networking strength rather than being distracted by a battle with Mellanox in the HPC market. It looks like a good move for both QLogic and Intel.

, , ,

  • greg schulz

    And what ever happend Stu to Cisco’s purchase of InfiniBand vendor Topspin which has been MIA now for some time, and with Mellanox having gobbled up Voltaire, to say the IBA market is has been consolidating would be an understatement…
     
    Good move for the Q to offload and get some cash to invest elsewhere.
     
    Not a bad move by Intel, at least they did not pay too much for the technology/business line.
     
    gs @storageio

  • http://jtarrio.tumblr.com jtarrio

    QLogic did not only spend $60M in SilverStorm, they also spent $109M in PathScale (
    http://www.pathscale.com/node/100), so selling the technology years later after investing millions in R&D, marketing and sales for $125M doesn’t seem like a huge success.

    Regarding Cisco, they didn’t buy Topspin to try to kill the Infiniband market. What happened is that they failed at their first vision of the “unified fabric” (http://www.cisco.com/en/US/prod/collateral/ps6418/ps6423/ps6429/prod_white_paper0900aecd80337bb8.html) and had to look somewhere else (FCoE) instead. So the products died a natural death

  • http://blogstu.wordpress.com stu

    Thanks, I forgot about PathScale (updated in the article) – my comment was not that QLogic made a huge return on the investment – it had a solid revenue stream for 5 years – but that selling IB off was a proper strategic move since there are not as many synergies for QLogic as they may have hoped.
    As for Cisco and Topspin – Topspin was a very exciting company, selling product into IBM and others. Cisco has utilized the Vframe management, but InfiniBand was never strategic to Cisco; Ethernet was always the “unified” layer of choice.
    Intel and Cisco have plenty of ties (Cisco UCS uses Intel amongst other connections); it will be interesting to see how the Intel acquisitions play into the competitive dynamics.

  • http://blogstu.wordpress.com stu

    Greg – the Topspin products are all EOL at this point. I’m sure some of the engineering talent has filtered into other parts of Cisco, but the IB ecosystem is quite small (but still gets plenty of discussion at the SC/HPC conferences) as you’ve said.