The Wall Street Journal on July 25, in an article titled “AMD Deal Raises Stakes in Chip Game”, made the following two important points. The first reflects the perceived advantages of integrating graphics and central processing. The second reflects the perceived advantages of open-ness – since there is another very large graphics chip provider, Nvidia, that will remain independent in the market.
- “Executives at AMD, of Sunnyvale, Calif., say buying ATI will help it match Intel's capability, and allow chips that occupy neighboring sockets to be merged into single products that handle computing, graphics and other functions.
“With this transaction, we will move from being neighbors to being a family,” said Hector Ruiz, AMD's chief executive, during a conference call explaining the transaction. “As technology advances, integration is not only inevitable but also advantageous.”
- [Nvidia CEO Jen-Hsun Huang] said AMD needed to show it will continue to fully cooperate with Nvidia for his company to keep making products that support AMD microprocessors. “If they work with us, we would be delighted to work with them,” Mr. Huang said.
Tom McCoy, AMD's executive vice president of legal affairs and chief administrative officer, said the company plans to be evenhanded in supplying technical information to Nvidia and ATI, and he stressed its customers still have a choice. “We are going to let the market decide what it wants,” Mr. McCoy said.”
The New AMD sounds like it aims to be more 'integrated' but also 'evenhanded' in its dealings with non-integrated partners. What will this take? Could Microsoft achieve that goal in the software world and if not, why not? For that matter, was Toyota able to do it?
It's an immediate and tangible manifestation of what Sallet and I see as the foundational question for the 'broadband value chain' today: What is at stake (and what are the potential commercial advantages) when a firm decides between a more “open” or more “closed” business strategy? What is at stake (and who benefits) when governments act in ways that promote open-ness or closure? What intellectual disciplines should be brought to bear to aid in analysis? What market diagnosis should be conducted to grasp more readily the impact on other firms – both vertically and horizontally – and on public policy objectives?
From an engineering standpoint integration has many advantages. And if we are thinking about it through that lens, why stop at the GPU/CPU/Chipset ? Why not integrate the radio, the network, the screen, the storage, the search function, the content creation…. Most people recoil at that idea. Why? Not because it wouldn't have engineering advantages – it would. But probably because it seems like a bad idea from an economic perspective – it's a bad business model. And it's probably because it sounds like a bad idea from an innovation perspective. And it's probably a bad idea from a public policy standpoint.
The outstanding question then remains. Where does integration make sense? What's the algorithm for a company to move the line from one place to another? What should policy makers permit, demand, facilitate, or block?
At that high level, the question reminds me very much of the 'make or buy' argument that fell into place around transaction cost economics in the 1980s. That question, which arose from basic understandings about the theory of the firm, at least had a shared analytic framework for people to talk about what made sense. It was not algorithm as 'template' that made corporate strategy simple and/or gave policy makers a clear formula for determining when anti-trust and other lines had been crossed. But it was a shared language in which people could argue.
Many people today use the terms 'open' and 'closed' to grasp for a similarly shared language. But they are too vague and too laden with emotional baggage to serve that purpose. Don't underestimate the emotional baggage. One of my colleagues at GBN wrote a book a few years ago in which he argued that the basic formula for prosperity in the future was 'open = good, closed = bad'. I know that sounds like 1999 hype and it was, but the underlying emotional valence is still remarkably strong (at least where I live in the SF Bay Area).
The point of this post was simply to set up a particular version of the problem. Later, I'm going to raise some ideas about how to think about solutions at least in terms of concepts. One comes from the theory of trade integration and the trade-offs between regional and global trade open-ness. Another comes from an experimental view of how the broadband value chain is, and is likely to, evolve.