Semiconductor companies face a long road back to growth
Even with government money soon headed their way, top-tier semiconductor companies will still have to deal with a range of challenges, from falling PC sales to material shortages.
With the passage of the CHIPS and Science Act, semiconductor companies got some much-needed good news after weathering the pressures brought on by chip shortages the past couple of years.
That relief might be short-lived.
Pouring cold water on the lukewarm optimism chipmakers have with the CHIPS Act money are the slower PC-buying market following the pandemic and soaring inflation causing IT shops to hit the pause button on purchases.
"The PC industry went through a period of irrational exuberance during the pandemic," said Glenn O'Donnell, vice president and research director at Forrester Research. "IT now has all the PCs they need, so naturally you see trouble coming for not just chipmakers, but carrying over to OEMs like Dell, HP and Lenovo."
The PC industry went through a period of irrational exuberance during the pandemic. IT now has all the PCs they need, so naturally you see trouble coming for not just chipmakers, but carrying over to OEMs like Dell, HP and Lenovo.
Glenn O'DonnellVice president and research director, Forrester Research
Rising costs caused by inflation are slowing not only PC sales, but servers as well. Companies that typically replace servers every five years might extend the life of their existing servers another year or two as a cost-cutting measure, O'Donnell added.
Inflation's rippling effect has also reached cloud providers like Microsoft, which is starting to milk its Azure assets longer and slow-walk the expansion of its data centers, O'Donnell said. This has meant that hyperscalers, who as a group buy a significant number of chips, have delayed their purchase of servers, which again contributes to the eroding chip sales of Intel, AMD and Nvidia.
Chip supply chain slows
Top-tier semiconductor companies -- most notably, Intel and Nvidia -- reported down revenues for the quarter ending in June and/or forecast a continuation of flat to down revenues over the next two or three quarters.
Yet another factor that figures to hold chip sales down in the coming quarters is the slowdown in delivery of components that chipmakers need to deliver finished products -- not just shortages on vital items item such as substrates, but other materials stretching from precious earth metals to aluminum. The latter material largely comes from war-torn Ukraine, where the flow of supply has slowed. This slowdown has reportedly caused a significant rise in the cost of that material, which is being passed on to chip manufacturers.
Some analysts see the lack of demand for PCs as one problem that will disappear sooner rather than later. They note that this down period is part of the normal selling cycle.
"Vendors sold more and more PCs, resulting in outsized growth even during the chip shortages, resulting in huge backlogs," said Dan Newman, analyst at Futurum Research and CEO of Broadsuite Media Group. "You knew the merry-go-round had to stop at some point, but you are starting to see a move back to normalization."
While falling PC sales affects all the top-tier chipmakers, each company has other problems unique to its situation. In Intel's case, problems that started before the pandemic included not having the proper manufacturing equipment, which inhibited delivery of innovative technologies. This also caused the company to miss multiple delivery dates on strategically important processors.
The latest pushback occurred two months ago when, for the second time, Intel pushed back delivery of its next Xeon processor, code-named Sapphire Rapids. Intel had previously announced that Nvidia would use Sapphire Rapids, not AMD's Genoa chip, in its DGX H100 system.
Intel expanded its D1X factory in Oregon earlier this year to provide 270,000 square feet of space for chip development.
Chip delays continue
Intel initially planned to deliver Sapphire Rapids -- a chip the company hoped would make it more competitive with Genoa -- in the fourth quarter of 2021. Intel told TechTarget Editorial it expects to deliver the chip sometime in the second half of 2023.
With each delay, the company falls further behind the technology curve compared with archrival AMD, which has been steadily eating away at Intel's market share.
"To fix most of its problems, Intel has to do three things: execute, execute and execute," said Frank Dzubeck, president of Communications Network Architects. "Every delivery date has to be met, every chip plant opening date has to be met, and research has to be functioning on all cylinders."
Intel's move to correct a series of management missteps was to bring back longtime Intel executive Pat Gelsinger in February 2021. Gelsinger has been aggressive on multiple fronts: investing $100 billion to build new chip fabrication plants over the next three years in the U.S. and Europe, acquiring key chip manufacturers, and reorganizing the company to establish new divisions dedicated to data center and AI technologies.
Still, the company has missed delivery dates on critical products and not offered any word on fixes applied to its manufacturing problems. Some analysts believe the manufacturing problems won't be resolved until the company builds the new fabrication plants starting in 2023 and continuing into 2025.
"The major challenge Gelsinger and crew have is to put better management in place to manage manufacturing," said Jack Gold, analyst and president of J.Gold Associates. "They have great designers, but you must have the ability to manufacture what you design."
Arm momentum continues
One other challenge Intel has is a competitive one with Arm Holdings. The U.K.-based manufacturer of lower-cost chips has gained impressive momentum over the past few years. Its competitive strength is that its chips require much less power than Intel's -- an attractive asset to owners of larger data centers concerned with energy costs and sustainability issues.
"Arm used to be seen as kind of a toy, looking down on it because it was mostly running mobile phones, not data centers," O'Donnell said. "That view held up for many years, but it's not so true anymore. You can crank it up now to outperform Intel at many functions."
The good news for Intel regarding its competition with Arm is that it's difficult to switch a large data center over from Intel chips to Arm chips. The process can be time-consuming and potentially expensive because it introduces compatibility issues with a data center's existing server-based applications.
"Enterprise data centers have had applications optimized for x86 hardware for years," Gold said. "To change over to Arm is not easy because the compatibility issues are not very transparent. Are busy IT people going to take the time to see if their mission-critical applications run efficiently on Arm? Probably not."
Nvidia's business has also been affected by the plummeting sales of PCs, particularly in the gaming market, where users heavily depend on the company's GPUs to drive performance. In the recent quarter, the company reported sales to that market had dropped by 33% year over year. Revenues of its data center products were up 61% year over year, but those sales dramatically slowed to 1% from this year's first quarter to the second, led by a decline in sales of its GPUs.
"Nvidia has had blowout numbers for so many quarters in a row, so there is no reason for people to lose their minds over this downturn," Newman said. "You will see this pullback among consumers and businesses for a period of time before we kick into the next cycle -- likely next year."
One advantage Nvidia has over Intel for a quicker recovery is the advances it has made by aggressively investing in AI-based technologies, which can be seen in its Omniverse offering, and tying those technologies to its GPUs. The company is also the first to incorporate its GPUs with its own CPU, which results in much faster performance.
AMD had a decidedly more upbeat quarter, reporting its eighth straight quarter of revenue gains, led by strong gains in its data center and embedded product businesses. The company expects to see continued growth in its data center business through the end of the year with the arrival of its 5 nanometer chip technology, according to AMD CEO Lisa Su.
Intel said it expects to deliver its first 7 nanometer processor, code-named Meteor Lake, in 2023.
Su also said at the most recent financial analyst meeting that she holds a "more conservative outlook on the PC business in the coming quarter," which could negatively affect overall revenues.
Microsoft underlined Su's comments last month in its quarterly revenue report, saying production slowdowns and a deteriorating PC market caused a $300 million hit to its Windows OEM business, which supplies Windows to all PC manufacturers.
"AMD had decent growth in its PC business, but that was from a much smaller number than Intel's," Newman said.
As Editor at Large with TechTarget's News Group, Ed Scannell is responsible for writing and reporting breaking news, news analysis and features focused on technology issues and trends affecting corporate IT professionals.
