Intel's Results and Guidance
After the bell on Thursday, Intel reported Q2 revenue of $19.73 billion (up 20% annually), GAAP EPS of $1.19 and non-GAAP EPS of $1.23. Those number respectively topped consensus analyst estimates of $18.54 billion, $1.05 and $1.11. Strong server CPU, notebook and flash memory demand more than offset weak demand for desktop CPUs and chips used within manufacturing, retail and automotive end-markets.
However, Intel's Q3 guidance was more mixed. Revenue guidance of $18.2 billion is above a $17.92 billion consensus, but still implies a 5% annual decline. Though Intel expects to continue seeing strong consumer notebook and cloud and telco server demand, weak enterprise/government server and commercial PC sales (particularly for desktops) are expected to be headwinds.
And with Intel forecasting a greater mix of chips relying on its 10-nanometer (10nm) manufacturing process node will weigh on its gross margin, it's guiding for Q3 non-GAAP EPS of $1.10, below a consensus of $1.14.
Intel also guided for full-year revenue of $75 billion (up 4%) and non-GAAP EPS of $4.85. While those numbers are respectively above consensus estimates of $73.92 billion and $4.81, they also imply Q4 revenue of $17.2 billion (down 15%) and EPS of $1.07, below consensus estimates of $17.67 billion and $1.12.
A Big Manufacturing Process Delay
Intel's stock fell 10.6% in after-hours trading to $54.00. While Intel's guidance may have a little to do with its post-earnings tumble, the bigger factor by far was the disclosures made about its next-gen, 7nm process node -- disclosures that helped AMD and Taiwan Semiconductor's (TSM) shares respectively rise 8% and 4.7%.
Citing a recently-uncovered "defect mode" that led to manufacturing yield degradation, Intel said that yields for its 7nm process -- seen as competitive with TSMC's 5nm process, which recently entered volume production -- are "trending approximately twelve months behind the company's internal target." And as a result, the launch schedule for Intel's 7nm CPUs "is shifting approximately six months relative to prior expectations."
Separately, Intel said that its Alder Lake 10nm PC CPU line -- it will feature its first 10nm desktop CPUs, along with its third-gen, 10nm, notebook CPUs -- will ship in the second half of 2021. The company still plans to ship its first 10nm server CPU platform, known as Ice Lake, by the end of 2020, and plans to ship a successor 10nm platform known as Sapphire Rapids in the second half of 2021.
On Intel's earnings call, CEO Bob Swan said that the first 7nm product to see production shipments will be a PC CPU arriving in late 2022 or early 2023, and that Intel's first 7nm server CPU will launch in the first half of 2023. And in comments that suggest Intel will be rolling out another 10nm PC CPU line in the second half of 2022, Swan said that Intel is "focused on maintaining an annual cadence of significant product improvements independent of our process roadmap, including the holiday refresh window of 2022."
Swan did also say that Intel's first server GPU -- it's codenamed Ponte Vecchio, and had been slated to use Intel's 7nm process and ship in late 2021 -- would ship either in late 2021 or early 2022. But to make this happen, Ponte Vecchio, which like many of AMD's CPUs uses a "chiplet" architecture in which a number of chips are placed within a larger chip package, will now rely on both "external and internal process technologies."
And later, when asked if Intel would be willing to outsource more of its manufacturing work to foundries (such as TSMC) in light of its 7nm setback, Swan didn't rule out the possibility.
"Our priorities in the ideal world [are] leadership products on our process technology so we capture the economic benefits of [having internal manufacturing operations]," Swan said. "But the focus will be leadership products, so to the extent that we need to use somebody else's process technology, and we call those contingency plans, we will be prepared to do that."
Implications for AMD, TSMC and Others
Even if Intel doesn't see any additional setbacks, its 7nm delay, together with the delays it has seen in launching 10nm desktop and server CPUs, could create a pretty big opportunity for AMD.
At its March analyst day, AMD disclosed that its first CPUs to use TSMC's 5nm process and a CPU core microarchitecture known as Zen 4, will arrive by 2022. And given AMD's current launch cadence for new CPU lines -- its first PC and server CPUs based on its Zen 3 microarchitecture are due in late 2020, less than 18 months after it launched Zen 2-based CPUs -- an early-2022 launch looks more likely than a late-2022 launch.
As a result, AMD, whose use of TSMC's 7nm process -- seen as competitive with Intel's 10nm process -- for its current Zen 2 CPUs and upcoming Zen 3 CPUs already gives it a manufacturing technology edge in the desktop and server CPU markets, could ramp shipments for its Zen 4/5nm CPUs about a year or so before Intel does so for its first 7nm CPUs.
Moreover, if Intel's 7nm delays also push back launch dates for CPU lines due further down the road, that bolsters AMD's competitive position beyond 2022.
Intel's 7nm delays might also benefit some of Intel's non-CPU rivals, such as Nvidia (NVDA) , Xilinx (XLNX) and Broadcom (AVGO) , although much depends on how willing and effective Intel is at turning to foundries to avoid being at a manufacturing disadvantage. Intel's plans to partly rely on a foundry to manufacture some of the chips used by Ponte Vecchio (likely the ones responsible for logic functions) point to a willingness to lean on foundries to bring advanced non-CPU products to market.
TSMC, meanwhile, naturally benefits when AMD, Nvidia, Xilinx and Broadcom sell more chips, since all of these companies rely heavily on TSMC for manufacturing. And while it's possible that Intel could also turn to Samsung as it relies on foundries to manufacture some new products, TSMC looks a little better-positioned right now to grab those orders, given its current manufacturing technology lead and Intel's history of relying on TSMC for select products.
TSMC's next-gen 3nm process node, it's worth noting, is slated to enter volume production in the second half of 2022. That means it should be arriving around the same time that Intel begins ramping production for a process node that's viewed as competitive with TSMC's current state of the art.