Intel is on life support. Can anything save it?
Summary
- Only drastic action can revive America’s chipmaking champion.
SINCE ITS founding in 1968 Intel has been synonymous with shrinkage. In its first four decades this was high praise. Every two years or so the American chip pioneer came out with new transistors half the size of earlier ones, a regularity that came to be known as Moore’s law, after one of the company’s founders. Twice as many chips thus fit onto roughly the same silicon wafer—and could be sold profitably for roughly the same price. That allowed Intel to corner the market for memory chips and then, when “memories" became commoditised in the 1980s, for the microprocessors which powered the subsequent PC revolution.
Nowadays when “Intel" and “shrinking" are uttered in the same breath it is no longer a compliment. After two sets of disastrous quarterly earnings the company’s market value has shrivelled to $84bn, less than the value of its plants and equipment, from over $210bn in January. The artificial-intelligence (AI) boom’s voracious appetite for silicon has propelled other chip firms, but Intel’s shares have seldom been this cheap since the late 1990s. It may be about to be booted out of the Dow Jones Industrial Average, possibly in favour of Nvidia, the champion of AI chips.
In early August Pat Gelsinger, Intel’s chief executive, announced that its 125,000-strong workforce would shrink, too, by more than 15,000. So will its annual capital spending, from over $25bn to just $20bn, and its yearly dividend, from $3bn to nothing. “Our costs are too high, our margins are too low," Mr Gelsinger wrote in a letter to employees. Its share price contracted by a third in the following days. Industrial-policy fans in the Biden administration, eager to revive domestic chipmaking, would be excused for worrying about their national champion.
Intel’s future now hinges on a new turnaround plan, which Mr Gelsinger is reportedly due to present to the board in the coming days. Intel-watchers expect some combination of more lay-offs, the sale of one or two peripheral businesses and maybe the shelving of plans for a $32bn factory in Germany. That would be too cautious. To save Intel, more radical change is needed.
Mr Gelsinger, who was Intel’s technology chief before later running a cloud-software firm, is not wholly to blame. He was brought back in 2021 to get the company on track after a string of beancounter-bosses failed to spot big changes reshaping the industry. In the late 2000s fat profits from the PC business blinded Intel to surging demand for mobile-phone chips. It stuck to making its own processors even as rivals went “fabless", outsourcing production to “foundries" such as TSMC of Taiwan. Repeated manufacturing slip-ups delayed the launch of new central processing units (CPUs) and let AMD, a fabless rival working with TSMC, steal market share. And lately Intel utterly missed the rise of specialist AI chips, which before a recent stockmarket wobble briefly turned Nvidia into a $3trn behemoth.
Mr Gelsinger’s original turnaround idea centred on segmenting Intel into a fabless design studio and Intel Foundry Services (IFS). Unshackled from in-house production, the design unit could pick the best foundry for its purposes. Stripped of a captive client, IFS would win business on its own merits. All very sensible. Except that it took for granted Intel’s ability to keep making money from CPUs and to reclaim stewardship of Moore’s law, which it forfeited to TSMC owing to those production blunders. This has proved, in the words of one exasperated analyst, “delusional".
Intel’s revenues sank from $79bn in 2021 to $55bn in the 12 months to June, as cyclical demand for CPUs cooled. At IFS, promising advances have not stopped customers from harbouring doubts about its manufacturing chops. Reuters recently reported that Broadcom, a $700bn chip-designer, has tested Intel’s TSMC-beating process and concluded it is not yet ready for large-scale production. IFS’s one big contract is with its sister unit—which, in a show of sibling cruelty, tasks TSMC with its high-end wares.
Mr Gelsinger insists the new production process will be ready by next year. Intel’s high-end chips would then, he says, come home and outside clients would honour $15bn-worth of loose purchase commitments that IFS claims to have bagged. Those revenues, plus the cost savings, asset sales, $8.5bn in grants and up to $11bn in loans from America’s government, would bring in the money for IFS to jump to the cutting edge and pull in orders. One day, when it can stand on its own, IFS may be spun out.
A less Pat response
This underplays Intel’s dire state. Free cashflow (left over after any capital spending), once a healthy $10bn-plus a year, turned deeply negative in 2022 and has worsened since. Deals with a pair of private-asset managers to co-invest $10bn-15bn apiece in two chip factories helped, but only a bit. A more decisive leader might quit chasing AI chips, where Intel is hopelessly behind, and flog the entire design unit to a fabless rival like Broadcom or Qualcomm. This would recapitalise the foundry without the need to offload assets bit by bit or put German noses out of joint.
Fabless chipmakers have an interest in seeing Intel’s foundry succeed. Without it, they would rely even more on TSMC for the most advanced chips. Letting Intel fail would be both commercially foolish and risky, given the possibility of China invading Taiwan. There is a precedent for chipping in. In 2012 ASML raised money from three big clients—Intel, TSMC and Samsung—for a new generation of its chipmaking tools. Intel paid $3.1bn for a 15% stake, most of which it sold by 2018. Today this would be worth around $50bn, equivalent to three-fifths of Intel. Potential backers, notably tech giants such as Amazon, Google and Microsoft, have deep pockets and want to design more of their own AI chips. Intel has expertise and talented engineers. Even if it does not offer an ASML-like return on investment, it may not be a bad bet.
© 2024, The Economist Newspaper Ltd. All rights reserved.
From The Economist, published under licence. The original content can be found on www.economist.com
