Intel, how did it miss the mobile, AI eras?

Intel, the undisputed giant of the global semiconductor industry, is in trouble.

Over the past year, Intel's stock price has been decimated, making it one of the worst-performing tech stocks in the S&P 500. The company's latest quarterly earnings release also disappointed investors, and expectations were lukewarm, leading to an evaporation of more than $30 billion in market value.

Not long ago, the market even rumored that Qualcomm is interested in acquiring Intel. Although there is no nail in the coffin yet, Qualcomm just put forward a proposal, but it also shows the dilemma of Intel. At that time, Qualcomm shares had a total market capitalization of $188.2 billion; Intel shares had a total market capitalization of $93.388 billion.

Looking back at the past two decades, Intel can be said to have missed a series of key opportunities, including ignoring the explosion of the cell phone market, slow to act on the application of EUV technology, and the early cancellation of the general-purpose GPU program, resulting in today's failure to keep up with the artificial intelligence boom.

How did Intel, once a miracle of engineering and technology, fall step by step to its present situation? Its failure, and can bring people what inspiration?

01
anachronisms

The first key mistake, missing the wave of the mobile computing revolution, was Intel's most dramatic setback.

In 2007, Apple released the iPhone, and smartphones quickly became the global computing platform in the years that followed. However, despite its dominance in the PC era, Intel stuck to its x86 architecture and failed to capitalize on the opportunity to provide chips for mobile devices. Back then, Intel's CEO allegedly turned down Apple's order for chips for the first-generation iPhone because Apple's bid was below Intel's budgeted cost.

Subsequently, competitors based on the ARM architecture, such as Qualcomm and MediaTek, quickly dominated the mobile space due to their power efficiency and low cost. Intel's attempt to enter the mobile market with the introduction of the Atom processor in 2008 was largely unsuccessful due to poor energy efficiency and performance.

The company outright canceled its smartphone business in 2016.

The miscalculation of the mobile space not only cost Intel billions of dollars, but also paved the way for the rise of companies like Qualcomm. And the ARM architecture, once limited to mobile devices, is now even spreading to Intel's traditional laptop and server markets.

Qualcomm's Rise to the Top in Mobile ｜Image credit: Qualcomm

The mobile market blunder is just the beginning.

IntelThe second error of thefailures in technology decision-making.(not only ...) but alsoIt happened in one of its proudest manufacturing areas..

In the use of extreme ultraviolet lithography (EUV) technology, an advanced process, Intel hesitates, preferring to continue to use its own original technology, so that its process technology in the competition with rivals TSMC and Samsung gradually lag behind.

Ironically, this expensive chip manufacturing process was initially funded by Intel itself, theBecause Intel acquired some shares of ASML in 2012, but the Intel CEO at the time was reluctant to use its just-launched EUV device.

Since Intel co-founder Moore proposed 'Moore's Law', Intel has been pushing for smaller, more efficient processes, and in the post-leadership era, this push has encountered unprecedented delays. In the transition to smaller processes, the manufacturing process advantages of the past appear to be stretched. 10nm process was originally scheduled to be launched in 2016, but experienced multiple delays and did not come out until 2019. The subsequent push to 7nm has also experienced delays.

At the same time, TSMC and Samsung from ASML large number of procurement EUV equipment, and constantly reduce the size of the chip's process, improve the efficiency and performance of the chip, to seize the market of advanced technology, AMD also TSMC's manufacturing process with the help of the launch of the chip with superior performance. For Intel, the cost of delay is painful, the technology advantage gradually disappeared, and let competitors gradually eat up the market share.

Also.Failures in manufacturing also existed in decisions regarding the foundry business.

In the past, Intel has always insisted on its IDM model, which was its strength in the glory days by controlling design and manufacturing at the same time. However, with the rapid changes in the market environment, this insistence, on the contrary, has become the company's shackles. Compared with TSMC, Samsung and other competitors who focus on foundry manufacturing, Intel's model limits its flexibility in the supply chain and makes it difficult to adapt to changes in market demand.

With an eye on TSMC making a fortune with its foundry model, Intel has finally decided to replicate that model in recent years with the launch of Intel Foundry Services, but that decision was as much about missing out on the mobile market as it was about missing out on theIt came too late, not only did it not make any money, but the foundry business even lost billions of dollars last year.In the first half of this year, only 1% of the revenue of the foundry business came from external customers.

A view of Intel's foundry｜Image credit: Intel

Even Apple has chosen to abandon Intel just as it continues to decline.

In 2005, Apple CEO Steve Jobs publicly embraced then-Intel CEO Paul O'Denning at the Worldwide Developers Conference, announcing that Mac computers would be moving to Intel chips across the board. At that moment, the Intel leader was in a state of euphoria.

But in 2020, Apple announced that it was abandoning Intel processors in favor of its own M-series chips based on the ARM architecture. Intel has since lost a large chunk of the high-end market for personal computers, and has further weakened the influence of the x86 architecture.

Steve Jobs unveils a Mac with Intel chips in 2005｜Image Credit: Apple

Intel's third mistake(math.) genusIt's the abandonment on GPU graphics that has resulted in the company being completely dumbed down at a time when AI is exploding and not eating any of the dividends of AI at all.

Intel's early attempts to enter the GPU market through the Larrabee project were canceled in 2010, as was Intel's entire GPU program. The consequences of this decision became official a decade later, with Intel missing out on the GPU and AI markets.

This is now very upsetting to current Intel CEO Pat Kirschinger, who worked at Intel for 30 years from 1979 to 2009, and was "sidelined" from the company the year before the GPU program was canceled. In his opinion, if Larrabee was launched back then, Intel would not have to spend a lot of money on acquiring some AI companies, and Intel would not be so far behind in the AI arms race today.

Alleged prototype of Intel's Larrabee GPU, photographed by a collector on Ebay in France | Image credit: leodanmarjod

Why cancel some projects and programs? In addition to judgment about strategic technologies and issues of competence, another problem for large companies is that theProfessional managers at the helm of companies are so focused on short-term returns for financial metrics that they spend hundreds of billions of dollars on dividends and buybacks rather than investing in R&D, and, in order to boost profits and cut costs, often axed programs that weren't making a lot of money at the time, or laid off employees.

Over these two decades, Intel has made several large-scale layoffs, including in 2006, 2016, and 2022, with thousands of layoffs as 'part of a cost-savings program'. In the early layoffs, many key engineers involved in next-generation process and architecture development were laid off, once considered by the industry to have caused long-term damage to Intel, particularly to its ability to innovate and to be overwhelmed in the face of technological change.

Meanwhile, while Intel continues to miss opportunities, the competition never stops; AMD is making a comeback in the CPU market, TSMC is producing chips for top companies in and outside of the industry with its pure-play foundry model, and NVIDIA is dominating the GPU and AI space.

Intel has acquired a number of AI companies, such as Nervana and DeepLabs, but many of these acquisitions have not yielded the desired results. For example, Nervana's products were canceled only a few years after the acquisition.

02
How do I get back to heaven?

At the moment, Intel can be said to be in dire straits, the market from time to time rumors that Intel may be acquired, in September this year, Qualcomm is rumored to be a major potential buyer.

Although Intel has not confirmed the plans, the news of a possible takeover of this once dominant market giant reveals the fall from its peak to its struggles.

Intel's current CEO, Pat Kilsinger, returns to the company in 2021 with a strategy of investing heavily in manufacturing capabilities and refocusing on technological innovation.To that end, Intel is investing $100 billion in new factories in the U.S. and promising to open up foundry services to outsiders.

Intel has also recently taken an important step - separating the foundry into a separate entity. The reason for the separation, apart from the poor business performance and the aim to improve operational flexibility as well as facilitating access to potential external funding, and more importantly, probably trying to convince potential external customers that Intel is willing to come up with cutting-edge processes to be able to make chips for 'outsiders'.

In the past, Intel reserved only state-of-the-art node technology for its own processors, but Pat Kirschinger thinks this was a mistake. He is very concerned about Intel's foundry business and currently plans to continue investing even though he is losing a lot of money.

Pat Kirschinger, current CEO of Intel｜Image Credit: Intel

Intel has also begun to ramp up its investment in AI and next-generation chip design, including the launch of the AI chip Gaudi3 in an attempt to compete with NVIDIA and AMD in the AI market, as well as the development of new AI PC chips such as Lunar Lake, Arrow Lake, and Panther Lake, with the goal of shipping more than 100 million units cumulatively by the end of 2025.

In addition, in September, Intel announced an expanded partnership with Amazon to customize AI chips for Amazon's cloud computing division.

Intel wants to customize AI chips for Amazon｜Image Credit: Intel

However, for Intel can be like TSMC to manufacture chips for external customers, the current market is still questionable. Citibank analyst Christopher Danely had sarcastically said: "I'm five feet six inches tall and over 50 years old, even if politicians all over the world want me to play in the NBA, it's almost impossible to realize.

Intel's revenues and net income have declined significantly in recent years. full-year revenues for 2021, 2022, and 2023 are $79 billion, $63.1 billion, and $54.2 billion, respectively, while net income has fallen from triple digits to double digits.

In August of this year, after the second-quarter earnings still disappointed the market, Intel announced the start of a comprehensive cost-cutting program, including the elimination of 15,000 jobs, accounting for 15% of the total number of employees, and plans to control non-GAAP research and development and administrative expenditures and so on in 2025 to 17.5 billion U.S. dollars, reducing capital expenditures by more than 20%.

The one person still willing to give Intel the most support right now is the U.S. government.

Intel is a direct beneficiary of the U.S. Chip Act, with funding from the policy providing a temporary cushion for the company to continue its technological innovations and factory expansions, theThis includes up to $8.5 billion in grants and $11 billion in loans, up to $3 billion in funding, etc..

However, policy funding support alone may not yet be enough to regain the giant's lost market share; funding from policy is limited and not forever.

As for Qualcomm's potential acquisition, which could theoretically spawn a bigger chip giant, there are other challenges, even if it doesn't face an antitrust crackdown from U.S. regulators. Because Qualcomm doesn't have manufacturing experience of its own, its current cash reserves may struggle to swallow Intel. There is also potential resistance from geopolitics.

In any case, Intel is at a crossroads, the fatigue is obvious, how to turn the situation around, and even regain the past glory, the current CEO can be said to be under great pressure.

Facebook (now Meta) had a thumbs-up button at the front door of its Menlo Park workspace for many years, and in the same spot was a sign for Sun Microsystems, which used to rule Silicon Valley until it was finally acquired by Oracle.

Even the most powerful giants will be overtaken by smaller but faster rivals when they encounter a technological paradigm shift they are unable to capitalize on. Fortunately, at least Intel has realized this matter.

This is a Geek Park original article, reproduced please contact Geek Jun WeChat geekparkGO