Text | Beyond the Layout, Written by | Admin
On August 2, 2024, Intel's stock price plummeted by 26% overnight, falling back to levels seen a decade ago and marking its worst single-day drop in many years.
Three months later, an even more significant turning point arrived. On November 8, 2024, Nvidia officially replaced Intel as a component stock of the Dow Jones Industrial Average.
A single index adjustment is also a pronouncement of an era. GPUs replaced CPUs, and Nvidia replaced Intel. This industry honor that Intel had upheld for 25 years ultimately changed hands, becoming the most striking symbol of its decline.
On April 23, 2026, Intel's stock price surged nearly 20% in after-hours trading following the release of its earnings report, marking a strong rebound. Less than two years had passed in between.
From being declared obsolete by the times to rebounding against the odds, what exactly did it go through?
Behind this lies a dramatic and turbulent business history.
I. Those Years of Being Kicked Out
Intel's decline is a textbook example of path dependence.
In 1999, Intel's market capitalization peaked at over $500 billion, making it the world's most valuable semiconductor company. Back then, one in every two PCs worldwide ran an Intel chip, and "Intel Inside" stickers were plastered all over the industry.
It's not just a chip company; it's more like the infrastructure of the computer age.
Then, the mobile internet arrived.
Intel, relying on its PC dominance, missed the smartphone race. The ARM architecture swept the mobile market, dominating Apple, Qualcomm, and MediaTek. Intel was not the winner in any mobile chip market.
Despite being almost completely wiped out in the mobile era, it still manages to maintain a respectable position by relying on the PC and server markets.
Following the mobile internet era, AI has experienced explosive growth.
Since 2020, training large models has become the most critical computational need. This requires the parallel computing power of GPUs, with tens of thousands of cores running simultaneously to repeatedly feed data to a model, tune parameters, and iteratively optimize it.
Nvidia's GPUs were designed for this purpose, while Intel's CPUs were relegated to a supporting role.
2024 will be Intel's worst year yet.
The company reported a net loss of $1.6 billion in the second quarter and announced layoffs of more than 15,000 employees, representing 15% of its total workforce. The second-quarter earnings report was released after the market closed on August 1st of that year, and the stock price plummeted by approximately 26% the following day, wiping out about $32 billion in market capitalization.
The cumulative decline for the year was close to 60%.
With this record in hand, Intel CEO Pat Kissinger was "forced to retire" on December 1, 2024.
Two, three CEOs in five years, and still couldn't save it.
Kissinger is no ordinary professional manager.
He joined Intel at the age of 18 and worked there for 30 years, rising to the position of Chief Technology Officer. He was widely recognized as Intel's "son." In 2009, he left to become CEO of VMware, where he made the company very successful.
In 2021, Intel's board of directors brought him back, hoping the veteran could salvage the situation.
His plan is ambitious, with a $200 billion investment in planning and building wafer fabs to transform Intel from a design company into a manufacturing company, competing with TSMC for the foundry market, and making Intel both a chip designer and a chip manufacturer.
Four years later, Intel's market value had evaporated by $150 billion during his tenure.
Kissinger's problem wasn't a lack of effort. His gamble, the wafer fab, required time, money, yield rates, and customer trust. These aren't things that can be built in two or three years.
But the market can't wait.
Before Kissinger, there was Robert Swan, a CEO with a finance background who was skilled at managing money, but the chip industry requires more than just financial discipline. Going back further, there was Brian Krzanich, during whose tenure Intel repeatedly lagged behind TSMC in process technology, allowing AMD to overtake it.
Five years, three CEOs, a complete cycle, but Intel's core dilemma remains unchanged: in the era when it most needed to bet on AI computing power, it bet on the wrong direction.
In March 2025, Chen Liwu, the ninth CEO, walked into Intel's headquarters.
III. The Ninth CEO's Three Trump Cards
Chen Liwu was not trained by Intel.
Born in Malaysia, he immigrated to the United States at a young age and worked in the semiconductor EDA field for decades. During his tenure as CEO of Cadence Design Systems, the company's revenue more than doubled, and its stock price rose by over 3200%.
He had no emotional baggage with Intel, nor did he have a Kissin-style revenge complex.
The first thing he did after arriving was to streamline the company.
In the past few years, Intel has gone too far down the path of doing everything. Kissinger's strategy was to do chip design and wafer foundry at the same time. The former requires product innovation, and the latter requires manufacturing precision. Trying to balance both leads to dispersed resources and a bloated organization.
Chen Liwu's assessment was to first stop the bleeding. He cut R&D and marketing management expenses by 8% and made it clear that further reductions would continue in 2026. This was no small move; the money saved in the first quarter was directly reflected in the profit statement, with non-GAAP net profit increasing by 156% year-on-year.
The second thing is to redefine Intel's core battleground.
Chen Liwu placed his bets on two areas: data center Xeon processors and foundry business.
The former is Intel's deepest moat; in the global server market, Xeon remains the mainstream CPU, and no company can shake this ecosystem in the short term. The latter is a heavy asset left by Kissinger, but Chen Liwu did not abandon it; instead, he spent time improving the yield rate.
In this quarter, the yield of the 18A process exceeded expectations, the proportion of Intel's 3 process EUV wafers increased, and the foundry business revenue reached $5.4 billion, a year-on-year increase of 16% and a quarter-on-quarter increase of 20%, beginning to show an independent growth curve.
The third thing is to proactively accept collaborations that others deem impossible.
Nvidia's investment, Google's order, and Musk's invitation—these three events, occurring within a year, all point to the same conclusion: Chen Liwu chose to open the door to allies when Intel was at its weakest.
This is a stark contrast to the closed strategy of the Kissinger era. In the past, Intel tried to do everything itself, resulting in failure in every area. Chen Liwu's logic is to identify the few things that are truly valuable to Intel, and then let others validate that value.
These three key strengths underpinned the better-than-expected performance in the first quarter of fiscal year 2026. Total revenue reached $13.6 billion, a 7% year-over-year increase; non-GAAP net income was $1.5 billion, a significant 156% year-over-year increase; and gross margin (non-GAAP) was 41%, showing steady improvement.
Fourth, its new friends are all former rivals.
In September 2025, the entire industry was shaken.
Nvidia announced that it will acquire $5 billion worth of shares in Intel at a price of $23.28 per share.
The two companies have been battling it out for thirty years. Intel's CPUs and Nvidia's GPUs have long vied for dominance in the computer market. Now, Jensen Huang has chosen to invest in his old rival and has also signed a joint development agreement.
Jensen Huang's logic is clear: Nvidia handles training, while Intel handles scheduling. The more powerful the GPU, the more powerful a CPU is needed to manage it. The two are naturally complementary, and Nvidia's investment in Intel is laying the foundation for its own ecosystem.
In December 2025, the U.S. FTC approved the deal.
Just six months after the news of Nvidia's investment broke, Google followed suit.
On April 9th of this year, Google announced an expanded collaboration with Intel, committing to deploy multiple generations of Xeon 6 processors in its AI data centers for AI inference and general-purpose workloads, and jointly developing custom ASICs. Intel's stock price rose 4.7% that day.
The significance of Google's order goes far beyond the purchase itself. Just a few years ago, Google was one of the world's most proactive companies in developing its own chips; the TPU was a product of Google's effort to wrest control of computing power from Nvidia. Now, while aggressively pursuing TPUs, Google is also clearly betting on Intel's Xeon CPUs.
In AI inference and agent workloads, both CPU and GPU are indispensable, and Google does not intend to gamble on one side of this issue.
Musk's kick was an even more powerful assist.
On April 7th of this year, Intel announced its participation in Elon Musk's Terafab project, joining forces with SpaceX, xAI, and Tesla to build a super chip factory with an annual computing power of 1 terawatt. Musk revealed that Terafab will use Intel's 14A process, with Tesla responsible for building the pilot production line and SpaceX responsible for mass production.
What Musk did was integrate Intel's manufacturing capabilities into his empire, which spans the three industries of space, automobiles, and AI.
For Intel, Terafab is much more than just an order. The computing power needs of Musk's ecosystem will grow along with xAI, Starlink, and autonomous driving, making Intel's position in this supply chain increasingly difficult to replace.
Three companies, three motivations, all leading to the same conclusion: Intel has suddenly become indispensable at this juncture.
V. The Era of Reasoning: The CPU Returns
Chen Liwu made a statement during the Q1 earnings call that is worth pondering. He said that as the focus of AI workloads shifts from training to inference, the ratio between CPUs and GPUs is changing from 1:8 to 1:1, driving a surge in CPU demand.
The first half of AI development involved training, stacking GPUs, feeding data, and iterating model parameters repeatedly. The CPU played a marginal role in this stage, primarily managing memory and I/O, nothing more. This was Intel's most painful period in recent years, and Nvidia's most glorious era.
The second half of AI's development lies in inference and agents. When models are deployed in enterprises, embedded in products, and invoked hundreds of millions of times, the landscape changes completely. Each intelligent agent, to complete a task, must invoke multiple tools, switch between multiple contexts, and coordinate multiple sub-tasks.
Who does this scheduling work? The CPU.
At the recent Google Cloud Next conference, Alphabet CEO Sundar Pichai stated that Google processes 16 billion tokens per minute. Behind every tool call and every context switch, a CPU is involved in processing these 16 billion tokens. (Further reading: Google doesn't want to beat Nvidia )
In the Agent era, both GPUs and CPUs are needed. GPUs are responsible for inference and execution, while CPUs are responsible for task orchestration.
This demand is also reflected in the financial reports. Intel's DCAI (Data Center and AI) revenue reached $5.1 billion in the first quarter, a surge of 22% year-over-year. The report disclosed that ASIC revenue increased by more than 30% sequentially and nearly doubled year-over-year. Xeon 6 secured orders from both Google and Nvidia, with demand across the entire product line far exceeding supply.
Intel failed in the AI training era, an era that required the parallel computing power of GPUs, while CPUs took a backseat. It took Intel nearly five years to find its place in the AI era: in inference, in scheduling, and in the central hub of agent orchestration.
6. Out of the ICU, but not yet discharged.
But this doesn't mean Intel has suddenly risen again; the underlying risks remain.
Under GAAP, Intel reported a net loss of $3.7 billion in the first quarter, primarily due to $4.07 billion in restructuring charges and goodwill impairment related to Mobileye. While its foundry business grew by 16%, it will take time to transition from loss to profitability. The 18A process yield "exceeded expectations," but there is still a long way to go before it reaches mature mass production.
Chen Liwu is still cleaning up the mess left by Kissinger.
Another key question is whether the demand for CPUs in the inference era is structural or cyclical? There's no definitive answer yet. As AI's agent capabilities continue to advance, chip architecture may change further. AMD is racing, Arm is racing, and Nvidia itself is also making moves in the CPU direction.
Intel has gained a window of opportunity, but that window won't stay open forever.
In the financial report, Chen Liwu said: "Today's Intel is completely different from when I first joined a little over a year ago. We have returned to our original mission, driven by data, maintained a sense of crisis, and focused on engineering."
This statement doesn't sound like it came from someone who has won a battle; it sounds more like it came from someone who has just climbed out of the deepest trough and is starting to regain their footing.
Intel struggled for nearly five years, went through three CEOs, laid off more than 15,000 employees, lost nearly $200 billion in market value, was removed from the Dow Jones Industrial Average, and was suppressed by competitors from all sides. At its lowest point, its stock price fell back to what it was ten years ago.
Then, the era of AI inference quietly began. Its rebound was not entirely due to its own efforts, but also to the arrival of a shift in the industry cycle.
Fortune changes every day, but many companies don't live to see that day.
A note outside the main page:
Every major technological wave in history creates both winners and victims.
Sometimes, the same technological wave can knock you down first, and then lift you up again.
Intel failed in the training era of AI; whether it can succeed in the inference era of AI remains to be seen.
But that nearly 20% increase at least illustrates one thing:
The market believes it still has a chance to win.
