After reasonably determining the forecasts for the past five years, this article discusses what the two companies might face in the next five years.

Editor’s note: This article is from WeChat public account “Semiconductor Industry Watch” (ID: Icbank), author EnerTuition. The original title “Intel has come to Sanchakou.”

Source: This article from the semiconductor industry to observe public number (ID: icbank) translated from “Seeking Alpha”, author: EnerTuition.

Since we published an article about the fate of Intel and TSMC, it has been more than five years. In that article, we predict that TSMC will replace Intel as the largest semiconductor company within five years. This took a few more months than we expected, but since the end of September, TSMC’s market capitalization has surpassed Intel. As of last Friday’s close, TSMC’s market value was $247 billion, easily surpassing Intel’s $228 billion market capitalization. We also predict that Intel is unlikely to succeed in its fab business and is as expected. After reasonably determining the forecasts for the past five years, this article discusses what the two companies might face in the next five years.

To understand this paper, readers need to understand Moore’s Law and Moore’s second law. The well-known Moore’s law states that in leading-edge integrated circuits, the number of transistors doubles approximately every two years. The less well-known Moore’s Second Law points out that the cost of capital for a leading semiconductor factory will grow exponentially over time.

Historically, from Robert Noyce to Brian Krzanich, Intel’s CEOs are convinced of the integrated device manufacturer IDM model. This model of integrating semiconductor design and manufacturing components has served many industry players since the 1990s until the foundry model began to gain and engulf IDM in the first decade of the 21st century. As a result, the competitive environment for each new process node has become thinner (as shown in the following figure, note that Intel’s recent 10nm process is roughly equivalent to the industry’s 7nm process).

The market value of TSMC exceeds Intel, and the former overlord Intel went to Sanchakou

Figure: LargeParticipants on the main process nodes since 2000 (source: author data)

The history of the semiconductor industry tells us that once a company moves away from the cutting-edge technology competition and turns to the foundry model, it will not return to the fab business. The foundry model meets Moore’s second law by aggregating the fab requirements of these fabless players.

Although chip foundry companies have made great progress, Intel management does not seem to realize the cruel reality of Moore’s second law, even if Jerry Sanders, who owns a fab, leaves AMD and AMD to survive. It has been the case after having to strip its fab for many years. A series of Intel’s ceos are very complacent, mainly because they have a wealth of x86 resources, which may give them a false sense of security.

It is certain that Intel has made some half-hearted efforts in building the foundry business at the same time, but failed to make significant progress. The critical time has passed in a half-way way. Intel is now at a stage where building on current in-house OEM models is becoming less and less viable, and x86’s rich may no longer support cutting-edge positions.

Consider the revised capital expenditure budget announced by TSMC in the recent third quarter earnings conference call – $14 billion to $15 billion in 2019, and a similar amount could be achieved by 2020. TSMC’s capital expenditures seem to have reached the same level as Intel, although its revenue is only about half of Intel’s.

In 2019, TSMC expects revenue of approximately $35 billion, while Intel expects revenue of $69 billion. This kind of income is a bit misleading because Intel’s revenue comes from end customers, and TSMC sells to semiconductor companies, which resell these products to end customers, with a markup of about 100% (or 50% of sales). After this difference is adjusted, the comparable income is much more balanced. With roughly the same income, TSMC now finds itself an undisputed leader in process technology. In addition, market dynamics indicate that TSMC will continue to grow and Intel may stagnate or contract.

In order to catch up with TSMC in this environment, especially from the disaster of 10nm process, Intel must now increase the proportion of capital expenditure to revenue. The problem is that Intel’s revenue opportunities are shrinking because of AMD’s market share growth. On the other hand, TSMC’s revenue is growing, which makes TSMC capable of increasing capital expenditures when necessary. Intel’s profit margin will decline due to the impact of AMD and the increase in capital expenditures in small-volume sales. At the same time, TSMC will maintain or even expand its profit margin due to lack of competition in the high-end market. These dynamics mean that Intel’s cash flow and profitability will get worse, which in turn will reduce Intel’s error rate at the leading edge.

From the perspective of investors, Intel shouldIt is gradually becoming two distinct companies:

1. A cutting-edge semiconductor design company focused on CPU and GPU. This part of the company, despite its recent poor performance in AMD, is a strong organization. In both areas, Intel has a high chance of regaining its performance leadership.

2. A recently poorly performing fab is facing a poor economies of scale controlled by Moore’s second law and may be falling behind.

Bundling the future of these two companies will almost certainly lead to Intel losing its product leadership for a long time. With these developments in mind, it is doubtful that Intel can now fight this battle of leadership with its own strength.

Interestingly, Samsung, which is behind TSMC in the 7nm process, is facing a similar situation. If there is a strategic need, and we believe that there is enough demand, then these companies have the opportunity to integrate their power to create a viable foundry product to replace TSMC. We estimate that Intel has about two years to divest its fab business and merge with Samsung’s fab business, giving it a great opportunity to stay ahead. It may not matter what action Intel or Samsung will take after the two-year window period, because TSMC may gain irreversible multi-year leadership. Most likely, competition will be the leader in semiconductor fab business.

Prognosis

In our view, Intel’s management has been pursuing Moore’s Law for the past decade, but has not considered Moore’s second law. Unfortunately, this is a key failure of the terminal type. As a result, Intel lost its leadership position in TSMC and the company is unlikely to return to its leading position in the industry.

CEO Bob Swan will face the decision of the split company. His success as CEO and the survival of Intel’s product leadership depend on the choices he will make. What is needed now is to take the lead, as Andy Grove showed in Intel’s competition in the memory business in 1985 – a major move that has allowed the company to focus on the most promising future growth areas. If Mr. Swan can’t accept this challenge, we will find that Intel will forever lose its position in the semiconductor industry. The difficulty of Swan’s mission is due to Intel’s huge organizational inertia.

Despite the possibility of divesting the fab business and integrating it with the Samsung fab business, and possibly achieving it, Intel’s organizational inertia makes the timing of these initiatives questionable. With this reality in mind, we find that Intel is now likely to have lost its leadership in the process forever and will suffer huge losses.