The main ingredient of a semiconductor is silicon, but it might as well be pyrite, or fool's gold. That is because consistently making money out of chips is notoriously difficult. Cyclical demand means that profits are volatile, and new kinds of chips quickly become commoditised. The business is also highly capital-intensive: a new fabrication plant, or fab, costs $3-5 billion, and new facilities must be built every few years as technology advances.

Accordingly, many Western technology firms, such as Philips, Hewlett-Packard, Motorola and Siemens, long ago spun off their chipmaking units in order to focus on the final products, rather than the bits inside them. Japan's huge electronics conglomerates have largely resisted this "fab lite" strategy. This;now seems to be changing, though the companies' willingness to let go fully is still in doubt.

On October 18th Sony said it would put its processor-chip division into a joint venture with Toshiba, which will also buy Sony's chipmaking facilities. Sony will no longer have to make huge investments in chip technology, and will still be sure of a supply of processors for its PlayStation 3 games consoles and other products. Its chip division lost $90 million last year, and the company has been getting rid of non-core businesses. Last month it floated its financial-services arm, raising nearly $3 billion. For its part Toshiba, one of the world Vs biggest chipmakers, will gain economies of scale.

Sanyo, another Japanese electronics firm, had hoped to do something similar. But its plan to sell its semiconductor unit for nearly $ 1 billion to Advantage Partners, a private-equity fund, fell through on October 16th. Sanyo is owned by a number of investors, including Goldman Sachs, which are doing their best to revive the struggling company by slimming it down to focus on its solar-panel and battery businesses. But the banks financing the purchase of Sanyo's chip unit balked at the price and at Advantage's plan to retain the existing management. Sanyo's announcement that it would keep the unit sent its share price plummeting.

Meanwhile, NEC, which in 2002 turned its chip business into a separate, publicly listed subsidiary, NEC Electronics, has spurned an offer from Perry Capital, a New York fund. Perry is willing to pay $1.3 billion to raise its stake from roughly 5% to 30%, on the condition that NEC relinquishes control of the chipmaker's board. (That works out at a premium of 60% over the average share price in the past three months.) Now Perry is quietly trying to convince other shareholders of the merits of its offer.

Taken together, this action (and inaction) adds up to a test of the willingness of managers at Japan's electronics firms to take rational but uncomfortable decisions. Spin-offs make sense because there are too many firms doing the same thing on too small a scale, and the need to finance new faces is a drag on the firms' main businesses. In Japan, however, corporate pride often trumps economic logic. Electronics giants are used to being diversified and vertically integrated: they regard selling a subsidiary as akin to amputating an arm. Still, some now see the need for surgery.