How money and markets are shaking up semiconductors

http://www.usatoday.com/story/tech/2013/08/05/semiconductor-industry-minyanville/2618625/

For every dollar they earned in the first half of 2013, Intel and Taiwan Semiconductor Co. spent $1.40 in capital investments. That's a lopsided number, and a sign of how competitive the semiconductor industry has become. At one time, TSMC dominated the independent foundry business, churning out everything from graphics cards to smartphone chipsets. In recent years it has become a three-horse race, with Samsung Electronics and Globalfoundries carving out large portions of the market, and Intel threatening to open things up even further with a move into mobile processors.
Both TSMC and Intel hope that, by throwing enough money at the problem, they can distance themselves from the competition. For Intel, this might be true. "Chipzilla" has a fat wallet, and a position in the PC industry that guarantees a decent return on its investment. TSMC has neither advantage, and may struggle to get ahead in the rapidly evolving mobile industry.
For both companies, costs are rising.
Gartner estimates this growth at 7%-10% annually, and predicts that by 2016, a capex budget of $8-$10 billion will be de rigueur for chipmakers like TSMC. Intel is naturally happy about this, and former CEO Paul Otellini speculated last year that new technologies -- such as 450mm wafers, which are expected to be more cost-effective than today's industry-standard 300mm, and EUV, or "extreme ultraviolet lithography," an up-and-coming technology that Intel hopes will allow it to further miniaturize its processors -- would cut the competition by half. "We've got [a] transition to 450mm [wafers] at some point; we've got a transition to EUV at some point," he said. "Both are going to be expensive, and are going to require scale."
Scale that Intel – which generated $19 billion in cash last year, and holds net working capital of $17 billion – can easily afford. Despite high capital expenditures and soft earnings in the first two quarters, the company still generated free cash flow of more than $4 billion. TSMC, on the other hand, had to tap debt markets to the tune of $3 billion as free cash flow turned negative. With working capital of less than $4 billion, and plans to maintain its current level of capital spending in 2014, the manufacturer is going to accumulate debt quickly. Samsung and Globalfoundries are spending less, but they have deep resources to draw upon, and aren't going anywhere.
TSMC's investments need to pay off soon, and in a big way – but this is a certainty that only Intel can count on. With its virtual monopoly in personal computers, and control over both chip design and production, Intel can force the adoption of new manufacturing technologies. Customers generally want these improvements, and are willing to pay for them. New processors typically account for one-third or one-fourth of the price of a new machine. Notwithstanding the dismal PC market, prices have held steady for both computers and processors, and with sales declines leveling off in the US – a leading indicator – it looks as though Intel's investments will pay off, regardless of its success in smartphones and tablets.
Taiwan Semiconductor faces a tougher road in the consumer device industry. Here, processors generally account for 2%-3% of the final product's retail price, and large improvements often go unnoticed by customers. There's small demand for a high-performance chip, and with long battery life already the norm, competition has focused on price. IDC estimates that smartphone prices have fallen 17% since the beginning of last year, and that's a potential problem for TSMC, which makes $9 on a high-end smartphone but only $4 on a low-end unit. To be practical, new factories need to lower costs – and Nvidia doesn't see this happening. Last year, the long-time TSMC customer complained that the new few generations of fabs will do little to cut costs.
That could end up being a moot point, if supply issues prevent customers from taking advantage of a new technology in the first place. Last year, shortages forced Qualcomm to second source its chips – to have some of them made by an alternative supplier, in this case Samsung. In the second quarter conference call, TSMC CEO Morris Chang was candid about his approach to customers. "We don't always put in the amount of capacity that a customer requests…. We have not and will not always follow the customer's estimates for their capacity need."
This obviates any technological advantage Taiwan Semiconductor might have, by forcing customers to design chips that can be moved between suppliers. TSMC expects to have a 20nm fab ready for mass production next year, but if Samsung and Globalfoundries are still at 28nm – in this industry, smaller is better – then Qualcomm has good reasons to stick with the older technology. TSMC might have solved this problem in the near-term by signing a deal with Apple, but without a change in philosophy, the new client probably won't end up any happier than the previous ones.
It's easy to underestimate the value of a strong balance sheet, or the security of a mature market. Intel can probably buy its way into the future, while Taiwan Semiconductor probably can't. On the other hand, neither one is likely to reinvent the industry through massive capex budgets. As smartphones and tablets continue to fall in price, it will become harder for TSMC CEO Chang's "grand alliance" – the ecosystem of chip designers and independent foundries – to push the envelope on performance, and just as difficult for Intel to compete with them on price. Change is the rule in any market; but it may be that the more the semiconductor industry changes, the more it stays the same.