Intel (INTC 5.41%) recently posted its fourth-quarter earnings, and the chipmaker's stock price initially rallied after the results. Its revenue dipped 1% year over year to $26.0 billion, but it beat estimates by $2.5 billion. Its adjusted earnings stayed flat at $1.52 per share, but it also cleared expectations by $0.41.
For the first quarter, Intel expects its adjusted revenue to decline 12% year over year to $17.5 billion, which surpasses analysts' expectations, and for its adjusted EPS to drop 24%, which matches expectations.
Those numbers seemed to calm investors, but Intel's stock price still fell after incoming CEO Pat Gelsinger declared he was "confident that the majority of our 2023 products will be manufactured internally" during the conference call. However, Gelsinger said Intel would still "likely" expand its usage of "external foundries for certain technologies and products."
Image source: Intel.
In other words, Intel won't outsource the production of its main CPUs to external foundries like Taiwan Semiconductor Manufacturing (TSM 4.37%) anytime soon, but it could let TSMC or other foundries manufacture its discrete GPUs and other chips. That's a bold statement -- but can Intel catch up to TSMC and its big customers like Advanced Micro Devices (AMD 3.46%)?
What happened to Intel?
Intel still manufactures its own chips with its in-house foundry. This makes it different from "fabless" chipmakers like AMD, NVIDIA, and Qualcomm, which outsource the capital-intensive manufacturing process to major foundries like TSMC, GlobalFoundries, or Samsung.
The number of chip foundries across the world has steadily declined over the past few decades, due to the high costs and technical expertise required to produce smaller and more powerful chips. TSMC, the world's largest contract chipmaker, is now the leader in this "process race," which is measured in ever-shrinking nanometers.
Intel once kept pace with TSMC in the process race with a two year "tick-tock" cycle for chip upgrades. In each "tick" year, Intel would introduce a smaller node. In each "tock" year, it would upgrade the chip's architecture, but the size would remain the same.
But under former CEO Brian Krzanich, who led Intel from 2013 to 2018, that two-year cycle expanded. Intel's foundry struggled to jump from 14nm to 10nm chips, and that failure caused a CPU shortage that allowed AMD to pull PC makers away from Intel. Under Bob Swan, who succeeded Krzanich after his abrupt resignation, Intel struggled to resolve that shortage, and fumbled its transition from 10nm to 7nm chips.
As a result, Intel's foundry has now fallen a full generation behind TSMC, which continues to produce AMD's latest CPUs. In December, Swan said Intel could possibly outsource its chips to TSMC, and would make a final decision in January.
Can Intel catch up to TSMC?
Intel has repeatedly argued that its current-generation 10nm chips are comparable to TSMC's 7nm chips in terms of transistor density. But that point could soon be moot as TSMC advances to even smaller nodes.
Image source: Getty images.
TSMC started mass producing 5nm chips last year, and it plans to mass produce 3nm chips in 2022. To maintain that lead, it will spend $25 billion to $28 billion on capex this year, up from $17.2 billion in 2020.
Under Swan, Intel actually reduced its capex -- from $16.2 billion in 2019 to $14.3 billion in 2020 -- even as it struggled with R&D and manufacturing issues. Swan, who has a financial background rather than an engineering one, also prioritized cost-cutting measures, buybacks, and divestments instead of resolving Intel's most pressing issues.
Gelsinger -- who was once Intel's chief technology officer and the architect of the original Intel 80486 processor -- is certainly better qualified to fix Intel's manufacturing issues than Swan. But that turnaround will require an enormous boost to its capex this year, and Intel's management didn't provide any specific capex guidance during its conference call.
Instead, Gelsinger expressed his confidence in Intel maintaining its in-house manufacturing, and said he was "pleased with the progress made on the health and recovery of the 7-nanometer program."
Gelsinger didn't disclose which of the company's non-core chips would be outsourced, but they'll likely include its new discrete GPUs. Other potential candidates include Altera's FPGAs (field-programmable gate arrays), Mobileye's automotive chips, and its Internet of Things chips.
Does Intel's strategy make sense?
Gelsinger likely believes outsourcing the production of its x86 CPUs to TSMC would make it tougher to differentiate its chips from AMD's, since they'll be manufactured on the same process. It could also permanently prevent Intel from moving ahead of AMD as it did throughout AMD's disastrous "Bulldozer" era, which preceded its current-gen "Zen" chips.
That commitment is admirable, but Intel must increase its capex significantly this year if it wants to catch up to TSMC. Therefore, investors shouldn't be surprised if Intel's earnings -- which are already expected to drop 10% this year -- decline more than expected. In short, investors should expect a lot of short-term pain as Intel sacrifices its profits to fix its business -- but those sacrifices could pay off over the long run.
