A probable break up. A less needed bailout. And the only sustainable path forward.
TL:DR
● Intel's board will accelerate the company break up now that Mr. Gelsinger has been jettisoned.
● Much of the public discussion is focused on Intel's Products Group and IFS, but the valuation and opportunities in those divisions are well understood. It is the sprawling R&D group and its work in several critical technologies where there is more opportunity to unlock value.
● IFS won't survive without contract obligations from the Products Group. Those obligations will misalign the interests of both organizations over the next few years.
● A US Government bailout (or nationalization) makes less sense if TSMC's Arizona facility is successful. Any bailout for IFS should require a plan to win in the marketplace.
● The long-term solution is to seed the market with fabrication technology that will change the industry structure. It has all the makings for a new Manhattan project.
Our instincts were to take a pass. There were dozens of well-thought-out essays on Intel's future in the two weeks following Mr. Gelsinger's "retirement" announcement. It was so plainly evident that the board forced him out. Everyone with an opinion had already piled on. There was not much further to add.
We reconsidered after meeting with our retainer clients during quarterly debriefs. Many of our clients have been diligently following the Intel situation unfold. Despite the coverage, there was still a lot of thoughtful dialogue. So, even though this post feels a bit late to the discourse, we thought publishing some of that discussion in the open would be worthwhile.
This post won't recount the history that got Intel to this point. There is enough blame to go around. Yes, Intel should have stopped dividend payments much sooner. Yes, the "non-technical" board made many critical mistakes. Yes, Mr. Gelsinger had most definitely over-promised and underdelivered on his 5N4Y (five nodes in four years) strategy. Sure, maybe the strategy was a bit arrogant. It is indeed a word that had come to define Intel during its decline. But no, the turnaround plan wasn't crazy. Intel still has a lot of very competent technologists, and we would never bet against those assets in any company.
Let's face it, Intel isn't the first company to collapse under the weight of its bureaucracy.
Carving out the Nuggets
So, what's next? Let’s dispense with the story telling. All the articles on what Intel should do next are interesting. But that isn’t reality. The board will speed up the process of breaking the company apart. The x86 processor still dominates the PC and data center server market. Even though it is losing market share, there is still a lot of value left to be extracted. These divisions within the Products Group could be spun out or acquired. Any competent analyst could draw up the valuation on the back of an envelope, and we are sure that shrewd PE firms and M&A directors are working through the investment opportunities.
Smaller units like Altera (the FPGA group) are already in the bidding process. Intel will most certainly divest its remaining interests in Mobileye.
Intel Foundry Service (IFS) will probably be spun out into a separate company. As many have already pointed out (see here for one review), this is the same path AMD took when it spun out its fabs into GlobalFoundries. Notably, a few years later, IBM also sold its chip manufacturing business to GlobalFoundries. Neither transaction went smoothly. The deals came with contract obligations that nearly bankrupted the companies. There was acrimony. There were broken promises and renegotiations. There were lots of lawyers and lawsuits.
Intel's offshoots will probably experience similar contentious relationships. But these misaligned interests are unavoidable. Like GlobalFoundries, IFS won't survive if its primary customer (Intel Products) walks away and decides to use TSMC. Fabs are expensive, and you have to run many wafers through them to keep costs manageable.
The spinouts could be more granular and perhaps unlock more value, especially in some of the smaller divisions within the Products Group (like networking). For example, Apple carved out Intel's smartphone modem business in 2019, along with 1000+ patents (many of which came from InterDigital). Apple will finally introduce its internally designed modem in 2025. It has been a hard road, but over the long term, that investment may pay back handsomely for Apple as it moves to further vertically integrate its smartphone supply chain.
It's also possible to do the same for the R&D group (or groups). Intel's R&D groups, spread over several boxes in the organizational chart, are jewels within the morass of its corporate bureaucracy. As Intel grew into a corporate giant, it attracted the best scientists and engaged in cutting-edge development in lots of different technology areas. There were probably too many. Intel worked on everything from basic transistor device design to quantum computing, silicon photonics, off-chip interconnects, sensors, RF, and power.
Department heads could always rationalize these activities in some way or another. Managers learned quickly that some of these R&D projects fit nicely in the white space on the upper far-right end of the PowerPoint slide showing roadmaps for extending Moore's Law. If that didn't make sense, Intel's "system foundry" business model was broad enough to include almost any function under its umbrella. Multi-functional systems captured the vision of the US Defense Department, which is always looking for integration technologies that they could fund and nurture.
An Intel breakup could free these teams from the weight of the corporate bureaucracy. Now is the time for Tech hardware companies to inquire and acquire those intangible assets. PE firms could also extract these groups and reintegrate them into a strategic buyer. Moreover, we are sure some entrepreneurial department heads at Intel would be willing to undertake a management buyout to build on the original vision of these teams.
Intel Corporate could even dictate the terms if it had an outside advisory competent enough to understand all the technology silos and their value-creation opportunities :) But, these types of "micro" carve-outs and aqui-hires are messy. To make it work, the management team has to streamline the process. It's a fire sale. That rarely sounds attractive to a CEO or its board of directors. Usually, these processes bog down into layers of oversight. By the time everyone puts their hands in the cookie jar, there is usually not much value left at the end of the transaction for it to have been worth the effort. Sometimes, these intangible carve-outs morph into IP licensing deals, but that isn't any better. In short, the corporate dysfunction that put Intel in the hole will probably kill off any viable paths for carving out these nuggets without the help of a competent financial acquirer.
National Champion?
The most popular solution among pundits is to wrap an American flag around IFS. That is, have the US bail out or nationalize the company. Proponents justify this remedy because the US needs a domestic alternative to TSMC for national security. With some leap of faith, the arguments for this scenario end with Intel finding a way to turn itself around. Call us crazy, but it doesn't seem plausible that injecting a corporate bureaucracy with a government bureaucracy will solve anything. In 2013, the London School of Business did an excellent review of state capitalism. To their credit, they took a more nuanced view of bolstering national champions and offered several criteria for success. Unfortunately, a US bailout or takeover of Intel would probably fail several conditions on their checklist.
Let's say Intel, under government control, finally does stand up its 18A process in 2025. Then what? How does the company win customers, turn a profit, and invest in the next node (14A/P)?
The advanced foundry business is a winner-take-all game. Fabless IC design companies make significant upfront engineering investments to design chips on a specific foundry process. They will almost always take the safe bet and go with the market leader. Why add another layer of product development risk?
Intel already had trust issues as a foundry because the Products Group was a potential competitor to most foundry target customers. Even if the Products Group is wholly severed, there have been enough missed milestones and rumored yield issues to create uncertainty. And with Mr. Gelsinger's unceremonious exit, the trust gap has widened further.
Market leaders in capex-intensive industries (like TSMC) have an overwhelming competitive advantage. Mr. Gelsinger understood this reality when proposing the 5N4Y strategy. The goal for Intel was always to be the number 2 foundry by 2030. The company expected to pick off some foundry customers from TSMC while relying on the Products Group to drive volumes and stay cost-competitive. However, with the Products Group severed, and the option to walk away after some (assumed) contractual obligation, that safety net will no longer be available. IFS will be in the same competitive position as Rapidus in Japan. And Samsung's foundry service will likely join them very soon.
As we outlined in our post two years ago, the US has two national security needs with respect to semiconductors. First, the US needs to tread water and get access to the existing state-of-the-art fab processes. Second, the US has to stay at the leading edge to sustain a competitive advantage.
At the moment, there is much discussion on the first goal because it is the most urgent. The CHIPS Act was a bet to get advanced wafer fabrication onto US shores. So far, the subsidies to TSMC have worked out. From now on, it would be much easier, and a demonstrably safer bet, to convince TSMC and Taiwan to abandon its "n minus one" strategy and extend their contracts to build the latest node processes on US shores. Some US customers have already committed to absorbing the higher costs of a domestic fab, and the US military doesn't need a large facility to meet its supply chain requirements. With that said, "n minus one" may even be good enough for the foreseeable future given the long design cycles inherent in the defense industry.
Need of A Long-Term Plan
It's the second goal, the long-term sustainability, where the public discourse falls short. State of the art today is trailing edge in two to four years. The US Defense Department must ensure it has access to the latest technologies in perpetuity. Unless the industry's economic structure changes, the US will stay on the hook for continued subsidies and bailouts.
Semiconductor industry subsidies and bailouts are nothing like the temporary automotive or airline bailouts of the past. There is no competitive advantage to regional focus (like what the CHIPS act is trying to accomplish) when the technology changes rapidly. Proponents of bailouts and nationalization pretty much concede that its the cost the US government will have to continue to pay if it wants the strategic advantage leading edge wafer fabrication.
But the US strategic advantage was never its access to the latest generation of advanced wafer fabrication processes. It was having a commercial market that incentivized the development of advanced wafer fabrication from which the military could leverage.
There are only two situations in a commercial setting that would allow a challenger (like Intel) to overtake a market leader (like TSMC) in a high capex business and create a virtuous investment cycle. First, the market leader could trip over itself and leave an opening for second-tier players (think Boeing and the windfall for Airbus). Second, a technology or market disruption allows a challenger to change the game. Ironically, TSMC capitalized on both scenarios to overtake Intel. Intel continuously faltered over the last decade while the fabless/foundry business model rapidly ascended.
In his widely read Stratechery newsletter, Ben Thompson suggested that the US could support IFS by piggybacking onto a new Manhattan project recently proposed by a US government agency. The agency proposed funding the development of general artificial intelligence (AGI) to compete with China. Mr. Thompson is trying to come up with a way to replicate the software moats imparted to hardware in past technology waves. The canonical example is Windows and the x86 processor. Or Apple iOS and the mobile application processor. Or, for a modern example, consider the Nvidia CUDA software stack and its GPU processors.
Mr. Thompson aims to link a government-run IFS organization with this new Manhattan project by incentivizing AGI model builders to integrate with Intel silicon. AI would provide a moat for IFS, which will drive fab volume and allow for further process technology investments.
We are big fans of Mr. Thompson's writing but aren't fans of this idea. By his admission, the idea was still "fuzzy." Our takeaway is that, unlike past hardware/software linkages, the Manhattan AGI project and Intel foundry integration is more like a shotgun wedding. It is artificial and adds no value to the end customers. Worse yet, it's a national tax on innovation, ultimately counterproductive in a geopolitical arms race. Advanced fab manufacturing is not fungible. There is a competitive disadvantage to not using the best and most trustworthy processes. Viewing all manufacturing as fungible is what got the US in this situation in the first place.
Moreover, a government-led AGI Manhattan project isn't needed. AGI is already attracting enough private capital. In fact, animal spirits are running wild. The private sector will do a 1000x better job of innovating. The military would be far better off leveraging those innovations than trying to co-create them.
However, IFS and the US could benefit from an alternative Manhattan project. We outlined the idea in these posts two years ago. The US government could play a critical role in funding R&D to change the underlying economic structure of the IC fab industry so that it can sustain itself with a more dispersed supply base. That would benefit IFS and a lot of other companies in the Tech hardware supply chain.
It begins with lowering capex. The moonshot proposal was to re-engineer advanced wafer fabrication so that the capex scaled down with lower production volumes. Lower throughput tools, nano-imprint lithography, and industry standard tool clusters play an important role. The option to build smaller fabs with advanced fabrication capability and without a scale disadvantage would lower entry barriers, allowing many more companies to enter the market. Instead of waiting and hoping for TSMC to trip up, it changes the game, allowing challengers to disrupt the business.
The benefits would accrue everywhere in the supply chain. Fabless suppliers would have viable second source options in an industry vulnerable to TSMC raising prices and eroding their margins. They may even consider vertical integration and a return to the IDM business model if the investment costs and volumes match their market needs. Fab equipment makers would also welcome having more customers and avoid the bargaining leverage inherent with a dominant customer.
That would be a Manhattan project worth funding.
If you find these posts insightful, subscribe above to receive them by email. If you want to learn more about the consulting practice, or how we can help with your product strategy, contact us at info@rcdadvisors.com.
From RCD Advisors, we wish you a joyous holiday season. If you are in the AI supply chain, we hope models continue to scale and 2025 continues to bring good cheer. If you aren't, we expect consumer spending and automotive to rebound, and 2025 will hopefully be better.
edited on 12/26/2024 for clarity