How far is Intel from a $1 trillion market cap?

Author: Su Yang, Tencent Technology

“I will correct the mistakes of the past.”

14 months ago, Chen Liwu said this at Intel Vision 2025, which was his debut as Intel’s CEO—at the time, he was 65.

Over the past 14 months, Chen Liwu has completely “recreated” Intel and received “high scores” from the capital markets in response.

As of June 2026, Intel’s share price has climbed from around $20.7 in the early days of Chen’s tenure to above $132, with a cumulative gain of more than 530%. The company’s market capitalization also rose sharply—from less than $100 billion in mid-2025 to now more than $660 billion.

I summarize his “correction” tactics as: layoffs, equity restructuring, betting on the 18A process, and so on.

Intel legend Andy Grove once said a timeless truth: “Only the paranoid can survive.” This nearly 60-year-old giant has long practiced that philosophy.

But today, Intel is starting to shift its gaze outward.

If you scroll through Chen Liwu’s Twitter, you’ll find the style is strikingly consistent—he has become a “group-photo fanatic.” The content is all photos of him with partners. This reflects that under Chen Liwu’s leadership, Intel is walking a completely different path from the past.

“Group-Photo Fanatic” Chen Liwu and His Circle

Called by many the “godfather of China’s semiconductor investment,” Chen Liwu keeps using his semiconductor industry resources to look for collaboration opportunities for the company. His Twitter posts are the best proof.

An insider said that at his busiest, Chen Liwu would have four dinners in a day (business banquets), starting as early as 4 p.m., describing the work state of this 66-year-old man.

“7106.5”—7 a.m. to 10 p.m., 6.5 workdays a week. Another insider revealed, “I can’t remember the exact timing, but anyway he’s really pushing it—that matches our 996.”

Not long ago, he made a public appearance for the first time on an external video podcast. He said that over the next 5 to 10 years, he would deliver 10x investment returns for Intel shareholders. If you rewind the clock to 14 months ago, that statement would have sounded incredible. And in 2024 and the beginning of 2025, the answer was absolutely impossible. But seen today, the 10x return goal has already been achieved to nearly 60%.

It may be for this reason that Chen Liwu decided to say something on the podcast camera—talking about what he has done during his tenure as Intel CEO, or rather, the results. This may also be related to his low-profile, restrained Asian background—without showing some achievements, it would be hard to say much.

Layoffs and Cost Reduction & Efficiency Improvement

Among the many “correction” measures, layoffs are the most direct and quickest way to improve the company’s financial situation.

According to Intel’s latest disclosed data, its headcount reached its lowest level since 2012. From 120,400 employees on a consolidated reporting basis at the end of Q3 2024, it was cut to 83,200 in Q1 2026, a net reduction of 37,200 people—about a 30.9% decrease.

In Q2 2025, layoffs led to $1.9 billion in restructuring expenses. Under GAAP, the company reported a loss of $2.9 billion for the quarter. By Q1 this year, the loss under the same GAAP basis was $3.7 billion, but most of that came from an impairment of Mobileye goodwill. This means that the one-time pain from layoffs and the negative impact of labor costs on performance are being gradually eliminated.

Looking at non-GAAP measures, Intel’s profit in Q1 reached $1.5 billion—layoffs’ effect was immediate. Rough estimates suggest that cutting headcount by 30% can save nearly $10 billion in costs each year.

I know many people at Intel. Because they work at foreign companies, their English names are all Mary and Lily, and they all have the letter “y” in them. But later, many of them left in this round of layoffs. When discussing this topic with peers, I always say, “The ‘y’ I knew is gone.”

The impact of layoffs and downsizing can be felt directly across many businesses.

In October 2025, Intel invited many Chinese colleagues to visit its Arizona Fab 52 wafer fab to witness a milestone in 18A process mass production. But before the event even began, a responsible staff member suddenly “resigned.”

Because of the rapid contraction of scale, many business operations may face similar shocks. For example, once someone in a certain position takes leave, the role may not operate smoothly, and cross-regional secondment of manpower is not out of the question.

In terms of organizational structure, semiconductor companies seem to like flat management.

The book The NVIDIA Way disclosed that nearly 40 NVIDIA executives report directly to Jensen Huang. Chen Liwu at Intel is the same—but he is even more “radical” than Huang: he requires all engineers to report to him. It must be said that the 66-year-old Chen Liwu is also a master of time management with abundant energy.

“From day one, I decided to have all engineering teams report directly to me,” Chen Liwu said in a podcast interview. “I’m too used to the culture of startups—the kind of lightning-fast pace to move forward.”

A Favorable External Environment

Organizational restructuring has created a good internal environment, but the external environment is also crucial.

“The most surprising thing is that in my professional background and training, I never learned how to deal with this—one morning, Trump asked me to resign, citing a conflict of interest,” Chen Liwu said in an interview, referring to the way Trump “challenged” him from afar.

The incident started with a post by Trump.

On August 7, 2025, Trump published an extremely short, forceful post on his social platform—using the distinctive all-caps emphasis style: “The CEO of Intel has a high degree of conflict of interest and must resign immediately. There is no other solution to this problem. Thank you all for your attention to this issue!”

Many people thought the “conflict of interest” was related to Chen Liwu’s identity as the “godfather of China’s semiconductor investment.” But that was just a surface-level excuse. What truly triggered the nerves of Trump and the U.S. government was the rumor that Intel was “pausing expansion of its foundry business.”

In late July 2025, Reuters and Fortune magazine successively disclosed an internal letter written by Chen Liwu. Especially in Fortune, it quoted Chen Liwu’s words very directly: “There are no more blank checks.”

Fortune quoted Chen Liwu’s original words in its coverage: “There are no more blank checks”

According to the report, in his internal letter Chen Liwu emphasized that every investment must meet economic returns. In the future, Intel would break the traditional logic of “building factories with huge spending regardless of whether there are orders,” and would further slow down construction of its $28 billion fabs in Ohio.

You have to understand: just a year earlier (November 26, 2024), Intel had secured nearly $8 billion in manufacturing subsidy commitment from the U.S. government via the CHIPS Act. Now you suddenly say you won’t expand—fab construction is paused on the spot. That’s like putting “MAGA” on a fire. How would the dream of “bringing advanced manufacturing back to the United States” be realized?

Later, Intel’s statement also explained the issue, to a greater or lesser extent.

At the time, Intel publicly said the company was making major investments in the United States, aligning with Trump’s “America First” agenda. Clearly, the purpose of this statement was to put out the fire—to tell the U.S. government that Intel would not stray from the “MAGA” narrative framework.

But because of the topic and the traffic effect, more people still chose to believe that Trump’s “anger” came from the China-U.S. narrative.

So in the podcast, Chen Liwu said he was willing to accept Intel because it is an iconic company, crucial to the semiconductor ecosystem and also crucial to the United States.

Regarding Trump’s challenge, Chen Liwu recalled: “He listened to my explanation—I was born in Malaysia, grew up in Singapore, went to MIT, and I’ve been living in the United States ever since. I’ve never lived anywhere outside the U.S. I told him all of that. He seemed to listen very seriously, and then he gave me this opportunity.”

Chen Liwu’s recollection was consistent with Intel’s statement at the time, both emphasizing “America First.” And it was those two messages that helped Intel secure a relatively more relaxed external environment.

In fact, in creating the external environment, nearly the entire Silicon Valley tech circle had already “bowed their heads in submission.”

Comparing Trump’s two inauguration ceremonies, you’ll find his second term is practically a spectacle—Silicon Valley big shots like “First Brother” Elon Musk, Tim Cook, Jeff Bezos, and others were almost all there. Even Zuckerberg—whom Trump had criticized for a long time—appeared on site. But in his first term, there were no public records of these giants showing up.

Mixed Reform and Asset Liquidation

During Chen Liwu’s tenure, he brought in a large amount of external capital for Intel, pushing this old giant to complete the largest-scale equity restructuring in its history. And because the U.S. government participated, this capital operation was also called the U.S. version of “mixed reform.”

According to public data, starting from the second half of 2025, Intel raised a total of $15.9 billion in cash directly from the U.S. government and global tech giants through targeted share issuances and the introduction of external strategic consortiums.

First, leveraging the CHIPS Act, the U.S. government converted subsidies into equity. Valued at $8.9 billion (including a $3.2 billion “secure enclave” project award), it obtained 433.3 million shares of Intel common stock, representing a stake of nearly 10%, becoming Intel’s largest shareholder. Then AI chip giant NVIDIA followed with an investment of $5 billion (subscribing for about 4% equity). SoftBank Group also invested $2.0 billion (subscribing for about 2% equity), becoming the fifth-largest shareholder.

Through the equity restructuring, Intel obtained a total of $15.9 billion in cash.

To further replenish ammunition and implement a “lighten the load” strategy, Intel carried out intensive disposal of assets during the same period. By divesting non-core subsidiaries, it recovered $5.2 billion in cash. After these two rounds of core funding converged, Intel obtained a total of $21.1 billion.

Before this, Intel was trapped in the most severe liquidity crisis in nearly 30 years—huge losses, debt piling up, and severe cash drain.

In full-year 2024, under GAAP, Intel recorded a net loss of $18.756 billion. Among them, the wafer-fab foundry business—which had been heavily expected to perform—continued to show operating losses of between $2 billion and $3 billion in each quarter from 2024 through early 2025.

At the same time, Intel’s annual capital expenditures at that time (building new wafer fabs, purchasing ASML lithography machines) were between $18 billion and $25 billion, while total liabilities rose to $914.53 billion. This heavy-assets “burning money” meant that the $8.2 billion cash flow from operating activities for the full year was simply not enough to cover the tens of billions in capex needed for fab construction. After deducting high interest and debt principal, its leveraged free cash flow was negative $115.75 billion—an extreme case of not living within its means.

The $21.1 billion obtained through mixed reform and asset liquidation could be described as Intel’s “life-saving money”—interest-free and not required to be repaid. Its scale was close to 84% of SpaceX’s first-ever $25 billion bond issuance after listing.

When I discussed the recent “storage price surge” topic with an insider, the storage business Intel sold off came up.

From 2021 to March 2025, Intel sold its NAND flash and solid-state drive business (including its Dalian, China wafer fab) in two phases to SK Hynix for nearly $9 billion.

From today’s perspective, if Intel hadn’t sold the storage business back then, how much room would its market cap have had?

There is no “what if” in reality.

After all, at that time, Intel was stuck in a quagmire. Without such asset dispositions, Intel might have been gone.

18A, Waiting for the Wind

Lei Jun said that “when the wind blows, even pigs can fly.” And right now, Intel is actually on top of the Agent wind.

So many people jokingly categorize Intel as a “small-cap bull stock” to emphasize its potential in the AI market.

The key to Intel being positioned for the Agent wind lies in the value of CPUs to Agents. When users use Agents, they essentially use the Agent to invoke various tools to handle tasks—such as calling the browser’s API to look up web pages, calling the operating system’s file system to read and write documents, and calling the database to access data.

Every time an external tool is called, the CPU must perform millisecond-level system interrupt handling and context switching. This is because the core control and scheduling of everything—execution at the bottom layer of all external APIs, the encapsulation of network protocol stacks, and memory bus management—are all in the hands of the CPU.

That’s why over the past few months, semiconductor pioneers such as Chen Liwu and Lisa Su have kept emphasizing the value of CPUs and continually updating the demand mix of CPU:GPU.

Chen Liwu’s most recent time talking about this was at JPMorgan’s 54th Annual Global Technology, Media and Communications Conference. He said, “The CPU-to-GPU configuration ratio is shifting from 1:8 toward 1:1, and may even surpass.”

As Intel’s most core cash cow, CPU-related business has averaged contributing revenue of more than $10 billion over the past eight quarters. In Q1 2026, Client Computing Group and Data Center and AI Group combined for cumulative revenue of $12.8 billion, accounting for more than 94% of total revenue.

Agent is inherently favorable for CPUs, and Intel is also trying to expand its own “edge AI” narrative—putting large models and Agents into NAS and into cars.

At this year’s Chain Expo, Intel showcased the concept of an “AI home brain.” By deploying small models locally, it can schedule and control IoT devices. It packs five small models—LLM (brain reasoning), ASR (speech recognition), TTS (speech synthesis), etc.—with a total parameter count of 140 million into one device, enabling the so-called “AI home brain” to understand language, see images, and understand intent.

The “edge AI” performance described above for Core Ultra is inseparable from support from the 18A process.

As Intel’s most critical generation of advanced process technology, the internal validation of 18A is the decisive factor for whether Intel’s foundry business can pull itself out of the quagmire. The application performance on Core Ultra—whether good or bad—directly determines future external customers’ choices.

An Intel insider also emphasized that the benefit of 18A for local model deployment on the third-generation Core Ultra platform is reflected in power consumption. “The biggest benefit of the node process currently used on the CPU tile for 18A is energy efficiency. It also allows AIPC to deliver relatively strong battery life when running local models.”

Additionally, because Core Ultra uses a discrete memory architecture, and the 18A process supports LPDDR5X-9600 MT/s physical bus bandwidth, with sufficient bandwidth it offers more memory capacity flexibility compared with a unified memory architecture (where memory is directly packaged on the chip substrate). In edge form factors such as Mini PCs and AI NAS, memory can be expanded to 96GB or even 128GB and above.

As for specific model sizes, Intel says it supports deployment of mainstream large models ranging from 4B to 35B parameters.

“35B is not the ceiling. With more memory, you can match larger models. But when balancing cost and capability, 35B is the sweet spot for locally deployed models,” the insider emphasized.

Returning to the 18A process node, Intel does not disclose precise yield data. In the most recent earnings call, Chen Liwu said, “On the 18A process, the good news is that things are much better than when I just took over. I can see a 7% monthly improvement in yield. Our progress is already ahead of our original (2026) year-end target.”

So we can only find supporting answers indirectly from some data.

In Q4 2025, Intel’s gross margin guidance for Q1 2026 was 34.5%, but the actual figure was 41.0%, exceeding expectations by 650 basis points. Considering that 18A began ramping up on Core Ultra in Q1 this year, this to a certain extent also reflects the pace of 18A yield ramp-up.

At the beginning of the year, the U.S. investment bank KeyBanc explicitly pointed out in its tracking report that Intel’s 18A yield had already successfully broken through 60%. Recently, GF Securities’ overseas technology analyst Jeff Pu even tweeted that 18A yield is around 80%.

In early June, Intel also released the Xeon 6+ server CPU based on the 18A process. That means that what previously had only one customer (Core Ultra) now has a new one. Jeff Pu also emphasized that starting from Q3, the main production capacity of 18A will officially tilt toward the server-side Xeon 6+ chip.

With higher yields and more internal order validation, the 18A process node is waiting for a stronger and stronger wind.

At Citi’s 2025 Global TMT Conference in the second half of last year, Intel CFO David Zinsner emphasized that 18A is a long-term node, and there will be more customers in the future. As TSMC increases foundry fees for 7nm and below advanced processes by 5%-10%, Intel is expected to benefit from this adjustment—provided Intel can also secure sufficient capacity.

At the end of last year, we estimated 18A’s capacity: Intel’s Fab 52 is expected to have a monthly capacity of 20,000 wafers.

On a 12-inch wafer, in theory you can cut 640 CPU tiles of 100 square millimeters. Based on the “around 80%” yield disclosed by Jeff Pu, with 20,000 wafers of capacity, even with full-year utilization, you can only produce 123 million CPU tiles. If monthly capacity is increased to 30,000 wafers, it would only be close to meeting Intel’s own “capacity freedom.”

In his tracking report, Jeff Pu disclosed Intel’s goal to double production capacity by the end of 2028 (compared with the end of 2026).

“We have clearly reduced spending on factory space and are allocating more dollars to equipment procurement. To address the current supply shortages, our equipment spending in 2026 has seen a significant increase compared with 2025. These expenditures will strongly support demand in 2027 and beyond,” David Zinsner said during the Q4 2025 earnings call.

Having insufficient capacity, forcing additional equipment procurement—this looks like a boomerang sent back to the internal letter Chen Liwu wrote in 2025: “There are no more blank checks.” It also serves as an add-on lesson to break the old practice of “spending huge sums to build fabs regardless of whether there are orders.”

But we can’t look at the problem in isolation from the specific historical conditions.

At that time, given Intel’s financial and technical situation, the only priority was how to survive—not the kind of “if we must have both” scenario. But now, with 18A waiting for the wind, as long as it clears the capacity hurdle, Intel has a chance to sprint toward a $1 trillion market cap—close to the long-term goal of nearly 10x shareholder returns.