news

Header Image | Movie "God of Gamblers"

Produced by Huxiu Technology Group

Author | Wandushan

Editor | Miao Zhengqing

In February 2021, Pat Gelsinger was appointed as Intel's eighth CEO. Just like when Jobs took over Apple again, Gelsinger also prepared a "return speech."

The reason why it is called a return is that he was the first CTO (Chief Technology Officer) of Intel. That year, he just turned 26 years old.

The CEO, who was an engineer, demonstrated excellent speaking skills on the spot and passionately introduced his grand blueprint to Intel employees: We must not only maintain our leading position in chip design, but also regain the chip manufacturing business, and even make chips for other companies in the future.

In the following three years, Intel restarted its foundry business, proposed the IDM 2.0 model, acquired Tower Semiconductor... and various reorganizations and integrations between business departments were also carried out in parallel.

Although during this process he was repeatedly warned by Wall Street analysts that the company's financial burden was already showing signs of being irreversible.

But Kissinger is still willing to bet everything. In his view, chip manufacturing will help Intel regain its former "technological confidence."

However, investors seem to have lost their patience.

On August 2, Eastern Time, Intel's stock price plummeted 26.06% after the opening, hitting the largest drop since 1982. After the market closed, S&P directly included Intel on the "negative credit watch list."

The critical blow suffered by Intel came from a financial report it released. In the second quarter of 2024, the company's revenue was US$12.8 billion, a year-on-year decrease of 1%; its net loss was US$1.6 billion, while its net profit in the same period last year was US$1.5 billion.

Along with the financial report, there is also a $10 billion "cost reduction plan":Intel will lay off approximately 15,000 employees worldwide and suspend dividends to shareholders.

Fate seemed to have played a huge joke on this blue giant.

Since the explosion of generative AI last year, Intel has been the first to promote edge AI andEdge computingIt is a chip company that first proposed the product category of "AIPC".

But just as its peers were making a fortune from AI, this "blue giant" had already stepped off the edge of the cliff.

Kissinger, did you really lose the bet?

01 Technology geeks’ “setting things right”

If we look back at Intel’s development history over the past decade, we may understand why Kissinger wants to develop the chip foundry (manufacturing) business with an “All In” attitude.

In 2006, Jobs approached Intel CEO Paul Otellini and asked him if he would be willing to supply chips for a top-secret project for Apple. After a brief assessment of market demand and costs, he declined the offer.

Later, Otellini revealed the name of the "top-secret project" in an interview with Time magazine - iPhone.

Many people attribute Intel’s decline to the difficulty in producing 10nm process technology, but in fact, missing out on the iPhone was Intel’s biggest strategic mistake since entering the 21st century.

To some extent, this event even directly changed the world line, because from then on, a large amount of semiconductor production capacity began to flow into Taiwan, and American chip foundry companies represented by Intel were reduced to the role of followers a few years later.

After Otellini retired, his successor, Krzanich, pushed the company to mobile business. After spending $1.2 billion to restructure the department, he was fired by the board of directors due to a "sex scandal". Bob Swan, who succeeded Krzanich, was once named "America's Best Financial Officer of the Year", but he knew nothing about technology. He once said in a conference call after the release of the financial report, "We may have to let TSMC help us produce chips."

When Kissinger took over Intel, the chip giant had lost all its former glory, and its once proud "engineering culture" was riddled with holes.

Therefore, the first thing he thought of was to reshape Intel's technology-driven corporate culture, which, at the execution level, meant rebuilding the chip foundry business.

To this end, Kissinger also proposed a concept called IDM 2.0.

The so-called IDM refers to a vertically integrated semiconductor model that takes care of everything from chip design to manufacturing to packaging and testing. Corresponding to it is the Foundry model (wafer foundry) represented by TSMC and the Fabless model (no wafer factory, only IC design) represented by AMD.

The more advanced IDM 2.0 is to insist on producing chips by itself while also providing foundry services to third-party chip design companies, and at the same time handing over some process chips to other foundries to supplement its own processes. Under this assumption, this "hybrid model" is more flexible in capacity allocation and production line adjustment, and is very suitable for diversified product development.

Another strategy proposed by Kissinger is "Five Processes in Four Years", which means gradually conquering five process nodes, namely Intel 7, Intel 4, Intel 3, Intel 20A and Intel 18A, from 2021 to 2025. If all goes well, theoretically Intel can catch up with TSMC next year, or even slightly lead it.

Of course, these radical ideas require real investment.

When chip manufacturing was at 55/65 nanometers, the investment in a 12-inch wafer fab was approximately US$2-3 billion. However, when chip manufacturing entered the 3-nanometer and 2-nanometer era, this investment became an astronomical figure.

Take the two wafer fabs Intel is currently building in Ohio as an example.When breaking ground in 2022, Intel said the project would cost $20 billion, and just a week ago, Intel said it would need an additional $8 billion in investment for the project.

According to Intel's plan, the land they reserved in Ohio can be used to build another six advanced process wafer fabs.

02 The “2030” is far away

Currently, Intel’s newly expanded production capacity will not be put into production until next year at the earliest. Before that, how should Intel explain to investors the losses caused by the huge investment?

Kissinger has mentioned a key time node many times in public: 2030.

According to his vision and estimation, Intel will produce a chip that can integrate one trillion transistors in six years. In comparison, the most powerful consumer graphics card currently available on the market, the RTX 4090, has 76.3 billion transistors.

At the same time, Kissinger emphasized that the foundry business would also break even that year, implying that as long as we can hold on until 2030, things will get better.

Of course, Kissinger also understood that it was not enough to just draw big promises, and no one could accept the situation of continuous losses. Therefore, this "tech geek" came up with three "conspiracies" to solve the problems brought by the foundry business.

The first step is to hold a foundry service conference (Foundry Direct Connect) to demonstrate its technological strength to the outside world and give investors a reassurance.

Objectively speaking, although Intel lags far behind TSMC in chip foundry market share, the technical strength it demonstrated at the foundry services conference is outstanding.

For example, the RibbonFET gate surrounds the field effect transistor and the PowerVia back power supply network. Simply put, the former can further reduce the size of transistors on the chip; the latter transplants the power and signal lines of the chip to the back of the wafer, thereby reducing power consumption.

Perhaps it was because Kissinger's reforms had an immediate effect, or perhaps it was because Intel still had a strong technical foundation. In any case, after the Chip Service Foundry Conference, the industry began to have confidence in Intel.

However, it is not enough to just show off your muscles, you must also do something with the financial statements.

Starting from the first quarter of 2023, Intel adjusted its depreciation standards and adjusted the depreciation period of some assets from 5 years to 8 years, which caused a strange phenomenon:

The amount of fixed assets, which consists of property, plant and equipment, increased, while depreciation expenses decreased during the same period.

Taking Q1 2021 as an example, Intel's fixed assets were US$57.33 billion and depreciation expenses were US$2.454 billion during the period. By Q4 2023, Intel's fixed assets had climbed to US$99.9 billion, but depreciation expenses had dropped to US$2.1 billion.

It can also be seen from here how exaggerated Intel's investment in recent years is. Its fixed assets have increased by 74% in three years.

It is also worth mentioning that Intel has also made new disclosure standards for foundry business revenue since the second quarter of this year. Internal offset items are now also included in the revenue of the wafer foundry business to avoid the losses generated by the wafer foundry being shared by other business units.

In addition to the above adjustments, Intel also split some business units so that they can calculate independently.AutopilotThe Mobileye department, the FPGA chip business department Altera, and the most important IFS department (wafer foundry business).

In general, against the backdrop of surging investment year by year, Kissinger has tried almost every method he could think of. But just as when one problem is solved, another one pops up. When the wafer foundry business began to stabilize, the company's "main business" was challenged again.

03 To prepare wedding clothes for others?

“AI will fundamentally change, reshape and reconstruct the PC experience through close collaboration between the cloud and PC, unleashing people’s productivity and creativity. We are moving towards a new era of AI PC.”

In September 2023, at the Intel On Technology Conference, Kissinger made such a judgment. This was the first time that the concept of "AIPC" appeared in the industry. Driven by Intel, a number of PC OEM manufacturers in the industry followed suit. In less than a year, AIPC flourished.

So, as the initiator, has Intel reaped the benefits of AIPC?

At least at this stage, it seems that there is no such thing. The second quarter financial report of 2024 shows that during the period, Intel CCG (Client Computing Group, which can be understood as PC chip division) achieved revenue of US$7.41 billion, a year-on-year increase of 9.3%, which is the only growth among the company's businesses.

However, the current profitability of the CCG department may not be optimistic. Intel Chief Financial Officer David Zinsner said in a conference call after the release of the financial report: "The gross profit margin in the second quarter was negatively affected by the accelerated growth of AIPC products."

In the future, the R&D costs of AIPC chips may increase further. According to Intel's plan, the product department is accelerating the development of new AIPC chips on the 18A process node.

It should be noted that Intel 18A is a process node that just completed the first batch of tape-outs last year. Developing consumer-grade products based on this node will inevitably significantly increase R&D expenses. Kissinger also bluntly stated, "This will have an impact on current profits, but will help future development."

While Intel is bearing high R&D costs, competitors in the industry are also exerting high pressure.

In May this year, at the Microsoft Build Developer Conference, Microsoft and Qualcomm demonstrated the "Copilot+PCs" combination. The first batch of products released by the host manufacturer were all equipped with Qualcomm Snapdragon X Elite/X Plus chips.

This chip, which was released in October last year, is built with TSMC's 4nm process and is equipped with a new customized Oryon CPU. In the 4+4+4 triple cluster, the main frequency of all 12 cores can reach 3.4GHz. Its NPU has a computing power of 45TOPS, and the integrated Adreno GPU can also provide 4.6TOPS computing power.

In comparison, the total computing power of the Intel Core Ultra currently commonly used on AIPC is about 34TOPS.

At the recently concluded ChinaJoy, AMD also showed its fangs. Intel's most direct competitor announced that the first batch of PCs equipped with Ryzen AI 300 series processors have begun to be launched. This series of chips is equipped with the XDNA AI NPU, which is known as the "strongest NPU on mobile terminals" and has an NPU computing power of up to 50 TOPS.

Although Intel still holds the top spot in the PC sector, companies such as Qualcomm and AMD have also demonstrated their strength to threaten Intel's industry position.

On the other hand, the situation of the DCAI (Data Center and AI, which can be understood as server chips) business unit is also not optimistic. The second quarter financial report of 2024 showed that the department achieved revenue of US$3.05 billion, a year-on-year decrease of 3%.

A horizontal comparison can be made: in the first quarter of fiscal year 2025 (ending April 28 of this year), Nvidia's data center business revenue reached US$22.6 billion, a year-on-year increase of 427%; AMD's data center business revenue in the second quarter reached US$2.8 billion, a year-on-year increase of 115%.

Nvidia is leading the pack, and AMD is catching up. It remains to be seen whether Intel can carve out a path forward under the leadership of Kissinger.

04 Conclusion

In addition to being a "technocrat", Kissinger is also a devout Christian. He once wrote a book called "The Wisdom of Balance", sharing how he balances "work, family, and relationship with God."

In the book, he compared himself to "an acrobat who juggles and catches three tennis balls with both hands." He required himself to be able to perform the three scenes with ease and not let any of them "fall."

Yet there is little balance at Intel today.

On the one hand, this chip giant still has considerable technical strength, whether it is the upcoming Meteor Lake and Gaudi 3 chips, or the "four years and five nodes" plan, which is enough to prove this.

On the other hand, after exhausting investors’ patience, it is still unknown when the technological revolution driven by these massive amounts of capital will be converted into tangible returns.

For Kissinger, he cannot give up either chip foundry or AI chip, because the former represents the foundation of Intel and the latter represents the future of Intel.

Intel is now like a chariot with the accelerator pedal pushed to the floor, but will it reach the finish line or break down along the way? Kissinger, the driver, has yet to give an answer.