IBM, as usually seems to happen in formal gatherings of the great and the good of the industry, got its way at this week’s meeting in Washington of the Semiconductor Industry Association, and members agreed unanimously not only to go ahead with the proposed Sematech non-profit joint venture company to develop and prove new chipmaking technologies, but that the new venture would manufacture only in low volumes. The received wisdom in the semiconductor industry is that the only way to prove a new manufacturing technology successfully is to apply it first to large-volume manufacture of an architecturally simple part like a memory chip, and Intel argued vigorously that Sematech should do just that. IBM was opposed, some observers say, because as one of the biggest US users of chips, it might be expected to buy a large part of the output, whether or not it came up to its own exacting standards. Sematech is no defined as an operation that will manufacture hindreds rather than thousands of wafers a day. Much more thorny is the problem of who will pay for it, and following the agreement, the US government will be asked to meet half the expected $300m to $400m annual cost.