At the end of 2024, Intel's annual YTD share price plummeted by nearly 60%. Not only has Intel been replaced by Nvidia and has been removed from the Dow Jones Index, but its market value has also fallen below 100 billion US dollars. At the same time, Intel is also facing a series of negative news. Qualcomm was once reported to be seeking to acquire Intel.
There were also rumours to sell factories, lay off employees, sell subsidiaries, and was even jointly sued by many shareholders for management's potential "false financial statements". In the following, we shall delve into the history of Intel, and the downfall of the once industry leader.
History
Intel has dominated the PC market for more than 30 consecutive years from 1991. At its peak, Intel had a market share of over 80%.
In the past, Intel had three major moats. The IDM model, the design manufacturer that holds x86 patents, and a brilliant sales strategy. Under these three moats, Intel's sales have soared under these three moats, escorting it to the position of industry leader for many years. According to multiple data rankings, Intel has entered the top ten in the global semiconductor market since 1987. From 1991 to 2021, it only fell to second place three times in the past thirty years. At other times, it has been ranked first. At its peak, Intel's market cap has exceeded US$500 billion.
Even after the technology bubble burst in 2000, Intel's market cap has slowly recovered to the level of US$300 billion. But after 2021, Intel's stock price plummeted. What caused the decline of the former "monopoly" with such a brilliant record? In fact, Intel's problems have been buried for more than ten years.
Intel's Strategic Failures
Intel’s decline can be integrated into three main issues: missed the mobile phase, flopped in GPUs &AI, and loss of CPU dominance.
In 2006, with the popularization of 3G networks, using mobile phones to surf the Internet became a fashion. The mobile Internet era was about to begin. However, Intel missed this wave. The main reasons were two: wrong partnerships and sold the wrong business.
Back in 2006, Steve Jobs was planning on the first-generation iPhone and hoped that Intel could provide it with processors. However, Paul Otellini, the CEO of Intel at the time, refused to provide CPUs for Apple. The reason being that Intel thought that the profits of the business were too low. This may seem reasonable at that time, but Intel miscalculated one factor - the future shipment volume of the mobile phone market will be much larger than that of the PC market.
Years later, with the release of the original iPhone and the wave of mobile internet, mobile phones gradually replaced computers and became the largest traffic portal on the Internet. At this time, Intel finally realized its mistakes and hoped to return to the mobile market. With regards, they found two giants at the time as partners: one being Microsoft and the other being Nokia. However, as according to history, this can be said to be a "loser alliance" combination. And in the end, Intel officially withdrew from the mobile phone market in 2016. The previous investment of US$10 billion in the field of mobile Internet was in vain. Intel launched its first independent graphics card i740 in 1998, but only a year later it announced its decision to integrate the i740 iteration i752 on the main version and canceled the independent graphics card.
Later, in 2010, Intel itself admitted that the power density of the x86 core was indeed not as good as other architectures of GPUs. But in the next ten years, no improvements have been shown and Intel has been absent from the independent graphics card market. Intel missed the mining boom and missed AI. The company terminated multiple AI projects due to budget constraints that could have been influential.
Perhaps is the guilt of missing such a big opportunity, after former CEO Gelsinger’s return to Intel in 2021, he proposed to return to the graphics card market. A year later, Intel Arc series graphics cards arrived as promised, but the game market pattern at this time has been determined. The late-arriving Arc graphics cards still failed to shake the status of AMD and Nvidia, the two red and green manufacturers, due to insufficient game driver adaptation and lack of ecology.
Looking back at Intel's entire graphics card development history, we found that Intel has never paid enough attention to this business. In fact, as early as 2005, Otellini proposed to acquire Nvidia, but the board of directors believed that the acquisition was too expensive and therefore strongly resisted. In addition to missing the graphics card market, Intel's other pain is missing out on AI. In 2017, OpenAI, which was founded just over a year ago, began discussing cooperation with Intel. Intel could acquire 15% of OpenAI's shares for only $1 billion in cash. If Intel was willing to provide hardware equipment for OpenAI at cost price, it could also obtain an additional 15% of the shares, which is a total of 30%.
As expected, I believe everyone can guess that in 2018, Intel once again rejected the opportunity to get rich. The then CEO Bob Swan believes that it is too early for AI to enter the market, and the investment will definitely lose money. This reason seems a bit untenable. If Intel really does not value AI, then what was the purpose of it acquiring Nervana Systems, a startup that makes AI chips back in 2016?
People familiar with the matter added that another reason why the deal failed was because Intel's data center division was unwilling to open up the door to cut prices, which would affect future profits. Intel originally missed out on the 30% stake in OpenAI that is now worth $47.1 billion. Now that Intel also intends to develop in the direction of AI, what did they do after acquiring Nervana? On August 20, 2019, Nervana released NNP-T and Nervana NNP-I. However, less than half a year later, Intel switched. In December 2019, Intel acquired another AI chip company, Habana Labs. The AI chip they produced is called Gaudi. NNP-T and NNP-I have been hidden in the snow. Perhaps the reason for giving up Nervana is the same as giving up i752 and Larrabee graphics cards.
Missing the growth point not once but twice caused Intel to be widened by its opponents. But the most fatal thing is that now even the CPU business that it was the proudest of in the past has begun to lose. In Intel's hypothesis, the chip design is improved year by year, and the chip manufacturing is improved like a pendulum. This plan is also named "Pendulum Plan", but in 2016, it happened to be the production process. Generally speaking, the smaller the nm reached the better the functionality is the chip. However, in the development years, Intel has been stuck at the 14nm node. Since Intel’s IDM model only makes chips for itself, when the old equipment is no longer useful, spreading the equipment cost evenly over the annual chips will lead to lower profit margins. If the R&D expenses can be reduced by improving the old equipment, the profit margin for each chip will increase. However, the result of betting on DUV (deep ultraviolet lithography machines) is that from 2015 to 2021, Intel has been stuck at 14nm while TSMC has already reached 5nm.
In 2022, Intel's revenue share in the data center dropped by 10% and was divided up by other competitors led by AMD. According to Mercury Research, by the second quarter of 2024, AMD's revenue share in the server market was 33.7%, and the average selling price was higher than Intel. It controlled the high-end market just when Intel had internal problems.
What makes it worse is that as NVIDIA GPU becomes the main equipment of AI data center, NVIDIA begins to use Grace on some products, for example bundle sales of CPUs and GPUs, such as the latest GB200. At the same time, we cannot ignore the change in the server paradigm in the AI era, which has caused the server CPU market to begin to shrink. From a CPU-centered server to an accelerator- centered server, whether it is GPU or ASIC, this paradigm shift has led to the CPU's share in the entire server processor market to continue to shrink.
Why did Intel Decline?
Speaking of Intel’s ultimate downfall, perhaps we can take a look at Intel’s board of directors. According to the information disclosed on Intel’s official website, the board of directors previously had 11 members when Pat Gelsinger left. When we look at these 11 directors, surprisingly we can find only 2 of them have a proficiency in semiconductors. Moreover, when Stacy J. Smith joined in March 2024 to replace the outgoing director Li-Bu Chen, there was only one industry veteran in the entire board of directors.
According to industry insiders, leading the operation of a semiconductor company requires very deep industry experience and accumulation. The underlying reason is because semiconductors are not like AI. AI is a relatively new industry, but semiconductors are not. There is a lot of experience that you will not understand without going through it. Thus, it is foreseeable that the board of directors lacks proficiency because people don't know what to talk about at all. According to senior chip field analyst Doug O'Laughlin's description of Intel’s board of directors being on autopilot for a long time is, to use a more elegant word, a runaway horse.
Conclusion
Looking back at Intel's various failures, there are a few words that comes to my mind: profit margin, conservatism, and arrogance. The Wall Street Journal also made an analogy between Boeing and Intel and commented that the corporate cultures of the two companies have gradually developed to prioritize financial performance over engineering excellence.
Before 2018, Intel’s manufacturing process had always been in an absolutely leading position. As a result, its design department actually relied more or less on the advantages of its own chip manufacturing department, thus making it not radical or innovative enough.
But where is Intel's future? Although the market is quite pessimistic now, and I do believe that the board of directors of this company lacks ability, but there might be still exit opportunities out there, perhaps being acquired by Qualcomm or TSMC?