Intel – Reality

Ben Thompson posted a very strong critique of Intel yesterday. We agree with most of it, except for one specific argument. And that difference leads us to very different conclusions.

More specifically, he asserts the following:

Here is the fundamental problem facing Intel, and by extension, U.S. dreams of controlling leading edge capacity: there is no reason for Intel Foundry to exist. Apple, Nvidia, AMD, and other leading edge fabless chip companies rely on TSMC, and why wouldn’t they? 

We fundamentally disagree with this assessment. We understand the logic, but it we see it as overly weighting the near term and loses sight of the very long time lines involved.

Put simply, the world absolutely needs an alternative to TSMC. Even setting aside US National Security concerns, there are strong commercial reasons why the semiconductor industry badly needs Intel to succeed. Today, there is some semblance of competition at the leading of semiconductors, Samsung is hanging in there, but their hold looks increasingly tenuous, and their commitment ambivalent. Without a viable Intel, we will have a true TSMC monopoly in a few years. As much as TSMC is the paragon of valuing customers, any monopoly is likely lead to higher prices and slower innovation.

Already today, to develop a chip on TSMC’s leading process is at least $50 million upfront, and usually much more. This puts leading edge chip development beyond the reach of any but the largest companies. TSMC today has a fairly flexible approach to working with start-ups, but a monopoly TSMC would have ample motivation to let that flexibility wither away. This would spell the end of chip design start-ups in the US.

And this is not just a problem for start-ups. The largest chip companies do not want to see a monopoly TSMC. TSMC without any competitors or fear of competitors will have little motivation to advance its process technology at the rate it has over the past ten years. That risks leading to stagnation in the industry and an even further slowing of Moore’s Law. The problem is the big companies’ short-term commercial imperatives override their long-term strategic needs. They all want a TSMC alternative, but they are all unlikely to do much about it.

The market needs competition, and Intel is the only viable path to that. We say this not as a plea, but as a commercial fact. There is demand for an alternative, if Intel can survive long enough to get its manufacturing processes working again, they will have demand for those services.

So then the question becomes how does Intel reach that? Until two weeks ago, the plan was five nodes in four years, get the products working, and then fill the fabs with foundry customers. Now everyone seems in a big rush to find an alternative. What has changed?

The obvious answer is Intel reported one quarter of bad numbers. Did anyone think Intel was going to get through this transition without a few bad quarters? We have warned for months that Intel’s products are weak competitively and that they would lose more share, and there are many people out there who cover the company much more closely than we do.

Of course, the problem is bigger than that. Intel’s Q2 numbers were bad enough, and they seemed so surprised by them, that the company lost their credibility with the Street. Every investor seems to be asking what is the next shoe to drop? For starters, they are going to have more bad quarters. Then we are going to find out that 18A is not as good as TSMC’s current process. Again, this was all clear months ago. But 18A does not need to beat TSMC, it just has to be good enough to hold off share loss to AMD in PC and data center CPUs. None of this good, but is it bad enough to cause the company to radically alter course?

We would argue it is not. The current path is painful and it is going to get worse, but there is no ready alternative. Splitting the company between Product and Foundry will end up destroying value, the two sides are still too closely intertwined. As Thompson points out rightly, this is the heart of the reason Intel is in this position, but a cold-turkey approach will prove too damaging a shock to the system.

Take Altera as an example. Intel announced they would spin off that acquisition in February. Altera has only been part of Intel for ten years, it still has many of its internal corporate structures in place. Nonetheless, we have some insight into that transition process and the sheer amount of work required to create new accounting, HR, ERP, CRM and all the other systems is exhausting and expensive. Splitting Foundry and Product, which have been tied at the hip for 50 years will take far longer to peel apart. Years during which the whole organization will be in turmoil, beset by distractions from auditors and consultants, all of which risks delaying 14A.

At this point, there is a risk of over-reaction to near term conditions leading to actions that harm the long-term viability of Intel. We understand that no one trusts management. We sympathize with concerns about all the things that can still go wrong. But it is important to remember we have to measure results with timeframes relevant to semiconductors. The planning cycles for $20 billion and $30 billion fab investments have to look five to ten years ahead. The decision to split the company, coming so soon after a single bad quarter, risks over-weighting the short-term at the expense of the requisite long-term perspective.

We reiterate our stance that the best thing Intel can do now is sell a ~20% stake to a consortium of leading chip design companies – Qualcomm, Broadcom but also Amazon and Microsoft – all those companies who will really need an alternative to TSMC in five years. Then the company has to get honest with itself about its position in the market, its AI roadmap and timeframe needed to get 14A to commercial volumes. This will all take time, but there is a viable path here, and we do not think any of the other options currently on the table offer that path.