In base terms, Intel’s data centric chip sales sucked in the third quarter after going gangbusters in the first half of 2020.
Government and enterprises in 3Q didn’t buy as many servers with Intel silicon inside after apparently snatching up too many in the first half, analysts said.
Intel CEO Bob Swan made it clear on Thursday that 2020 flip-flopped the silicon maker’s normal pattern of slower data center sales in the first half followed by an improvement in the last half.
But 2020 has been anything but ordinary.
Intel’s data center group gained 23% in revenues in the first half, then dived 7% in 3Q over a year ago, hitting $5.91 billion in revenue. The decline matters -- a lot -- because data center chips were nearly one third of Intel’s revenues for the quarter, which hit $18.33 billion, down 4% from a year ago. Net income for 3Q was $4.3 billion, down 29%.
Investors freaked out Friday after the earnings report, dropping the share price 10% to $48.91. By mid-day Monday, shares were down another 3% to $46.49 and closed Monday at $46.72.
To its credit, Intel saw 15% cloud data centric growth in 3Q, but its enterprise and government market fell 47% after two quarters of 30% or greater growth.
For Intel investors, this is a worrying time. Stay in the stock or get out? Intel shares have grown steadily since the March low that hit all stocks, but that’s obviously not enough for some investors.
For buyers of servers, like IT folks in government and companies, Intel share price won’t matter as much as what Dell EMC or HPE charge for servers and the features they include. Still, industry analysts recognize that IT managers care what happens to a longtime supplier like Intel, which has been a rock for decades that every server maker will continue to support for a long time.
“Enterprise and government care about stable supply and reasonable costs,” said Kevin Krewell, an analyst at Tirias Research in an email comment to Fierce Electronics. “While Intel’s process issues have impacted new product ramps, overall product supply is stable. Intel has the capital to invest in building sufficient fabs to meet demand. Competition from AMD helps to keep costs in check.”
Swan made a point last week of saying that it will be ramping up factory production, noting that despite COVID-19 and trade disputes globally, 2020 will actually be a stronger year than 2019 “with a better product portfolio going into next year.”
After industry concern over Intel’s ability to ramp 10nm chip production, Intel also noted a 10nm chip fab in Arizona is operational and the company expects to ship 30% more 10nm volumes in 2020 than Intel expect in January. Demand for 10nm has been higher than previously expected, the company said.
Long-term take from industry analysts
Intel’s 3Q results don’t take the long view, at least according to three industry analysts contacted by Fierce Electronics.
“It would be foolish to count Intel out over the long term,” said Jack Gold, an analyst at J. Gold Associates. “They’ve had issues like these in the past and roared back.”
He acknowledged that Friday’s downward market reaction to the 3Q report shows that “Intel seems to have slipped.” Competitors AMD and Nvidia have seen sales improve in the past one to two years, starting from a much smaller base. Both AMD and Nvidia have out-marketed Intel, he added.
“Intel has some real issues, like not being able to get to a 7nm chip process like their competitors, which is a big challenge,” Gold added. “A lot of what happened with data center revenues [ in 3Q] is that customers pre-bought over the previous two quarters as a result of pandemic and shutdowns and then backed off when they had too much capacity or shifted more to cloud rather than buying on premises.”
“Cloud sales are picking up for Intel, but also AMD and Nvidia,” Gold added. “Couple that with the major cloud providers like AWS and Google Cloud Platform making some of their own chips to try to get power down and prices per CPU lower. This is a dynamic that hasn’t happened at Intel in the past several years.”
Some technology analysts, if not investors, are taking the long view with Intel technology and what it means to IT buyers of servers and not only to Wall Street. These analysts have noticed Intel is undergoing a wholesale re-think of how computer systems are designed with a system-centric approach, instead of a more singular focus on whether a CPU is designed with a 7 nm or 10 nm chip process.
“There is far too much [investor] focus on chip process technology and supply chain and speculation that Intel will go fabless due to their slip with the 7nm chip process,” said Leonard Lee, an analyst at NeXt Curve. “CIOs don’t buy chips, they buy systems.”
As Intel continues to focus on device and server manufacturers, Lee said AMD has a limited portfolio that is centered on the CPU. “I think AMD recognizes that in order to play in the data center market, you need hardware acceleration,” which is why a rumor emerged that AMD wants to buy Xilinx and its FPGA business. AMD announces third quarter earnings on Tuesday.
“CPU is not everything,” Lee said. “I think Bob Swan put it well, that you need an XPU portfolio. That’s why Nvidia hopes its Arm transaction is approved. It’s all about the system, not a device.”
Bob Swan offers update on Intel’s XPU concept
Whether investors listen or not to Bob Swan, Intel first laid out its XPU concept more than a year ago, where XPU stands for “Other Processing Units” that include a range of specialized processors such as GPUs and FPGAs. The more mainstream term now gaining vogue is "heterogenous computing."
In comments about XPUs last Thursday, Swan said, according to a Seeking Alpha transcript:
“CPUs are foundational to our business, but we are also adding a range of other processing engines or XPUs to our portfolio. We’ve made great strides in graphics and we are now scaling our graphics architecture from integrated to discrete levels of performance. Our first discrete GPU DG1 is shipping now…”
“Beyond the CPU and GPU, our customers tell us that they want a diverse range of AI solutions to fit every power level and performance needs from the intelligent edge to the data center. For the most demanding AI workloads, our customers are looking for purpose-built XPUs that leverage a standard-based programming environment. With that in mind, we acquired Habana Labs almost a year ago…Habana’s inference card is now in volume production and shipping to customers.”
“Additionally, we have been working closely with the ecosystem on the open standards oneAPI effort as part of the XPU transformation. With oneAPI we are creating an open unified software architecture that can support the variety of XPUs that our customers demand. We’ve made tremendous progress with developers and released spec 1.0 of oneAPI in the third quarter and are on track to have the gold release of oneAPI software in the fourth quarter this year.”