Technology: High stakes required to rake in the chips: Intel is facing expensive choices as it battles to stay ahead in the microchip industry

'IT TOOK us 22 years to reach dollars 1bn ( pounds 640m) a quarter,' said Andrew Grove, chief executive of chip maker Intel, announcing the company's latest quarterly results last week. It then took less than three years to reach sales of dollars 2bn a quarter, he added.

This is strong evidence of the explosion in demand in the PC industry - but the bumper profits obscure a number of difficult decisions ahead.

Intel, one of the most profitable companies in the world, is finding that survival in the fast- moving PC industry is expensive. This year, R&D expenditure will top dollars 1bn and capital expenditure will cost a further dollars 1.6bn.

Each new generation of chips is exponentially more expensive than the last to develop and build, according to Dave House, Intel's senior vice-president. A plant to manufacture the company's 386 processor cost dollars 100m; a plant to build the 486 processor cost dollars 400m; and one to build the company's new Pentium processor - its first significant new chip since 1989 - will cost dollars 1bn.

Although Intel benefits from some hefty mark-ups - the 486 sells for more than dollars 300, but costs less than dollars 20 to make - costs are tightly controlled. Like everyone at Intel, including Mr Grove, Mr House has a small 'open plan' cubicle rather than an office.

Pentium has forced Mr House, as the company's corporate strategist, to make some difficult decisions. First, as reported here in February, Intel has had to break with the past in its choice of a name for it. The courts have ruled that a number cannot be a unique trademark, so goodbye 586 - successor to the company's 286, 386 and 486 chips - and hello Pentium, which although less familiar to customers can at least be copyrighted.

The strategic thinking behind Pentium is complex. First, the chip is designed to eat into the market for the 'RISC' chips used in high-end workstations as well as to power ordinary PCs. To do this, the new chip - the most powerful microprocessor in the world - performs extraordinary internal convolutions, including processing instructions in parallel, while retaining compatibility with previous generations of PC chips.

The result of all this is increased speed, but how much depends on the way that programs are written. Programs designed to take full advantage of Pentium will run twice as fast; others may simply achieve increases of tens of percentage points. To maximise revenues, Intel needs to perform a balancing act.

Prices are still under wraps, but industry analysts are predicting that users wanting Pentium speed will have to pay between pounds 1,000 and pounds 2,000 extra initially. Many will choose not to. Not a problem, says Mr House. The 486 'workhorse' chip is going to be around for a long time yet, and the company is still increasing production levels: last quarter's output was more than four million.

Dataquest, the US firm of analysts, predicts Intel will sell two million Pentium chips versus 36 million 486 chips in 1994. But therein lies the difficulty: as with the earlier 386, Intel now shares the 486 market - and therefore has to settle for lower margins.

Rival AMD's shares soared recently, when a California court overturned a previous ruling that had denied the company access to the Intel-developed 'code' inside each chip. By contrast, Intel's shares dropped almost 15 per cent on the news.

The immediate challenge facing Intel is to wring as much money as it can from Pentium sales while holding 486 prices as high as it possibly can in the face of competition. Longer term, the strategy is to wean users away from buying new PCs when they want more power to buying a new chip to fit into their old PC instead.

Various upgrade options are already available for 486 PCs, including an 'Overdrive' processor to speed machines by up to 70 per cent. The future will see much more of this. Sources close to Intel have mentioned a possible deal with Dixons, which would see microprocessors sold on the high street.

The concept is potentially a source of great tension between Intel and PC manufacturers - in theory,more upgrades could mean fewer PC sales. But this will not happen, according to Mr House. 'We did a bunch of market research as to what would happen - and it didn't result in the acquisition of any fewer machines.'

Businesses appear to depreciate PCs over four to five years, he says, and Intel sees the market for upgrades being an additional, rather than alternative expenditure.

Most manufacturers fight shy of product proliferation, but not Intel, which now has a range of 15 different microprocessors.

Mr House is also sanguine about the theoretical limits of chip technology.

While he accepts there are genuine limits, he insists they will not be reached in the next 10 years.

(Photograph omitted)