It won’t be easy, but it could be done. This is How Intel Gets Out of Manufacturing...
On the heels of activist hedge fund Third Point calling on Intel’s board to explore “strategic alternatives,” like spinning off its fabs and/or divesting itself of unsuccessful acquisitions, Intel’s board reacted with a rapid change of CEOs, tapping Pat Gelsinger to succeed Bob Swan. However, in his first address to the financial community, Gelsinger committed Intel to remaining an integrated design manufacturer (IDM) — and, yes, that might require outsourcing some manufacturing to third-party fabs in the near term. While this may not please some investors, it is the correct decision at this time. However, I have a theory on how a spinoff could be successful, theoretically, in the long-term, say five years or more down the road.
The suggestion of spinning off Intel’s manufacturing is nothing new. In fact, I recommended it approximately seven years ago. As the cost of operating a fab, and developing new process technologies and new manufacturing equipment has increased over the past 20 years, most semiconductor companies have outsourced most, if not all, manufacturing, while the rest, including Intel, have adopted a hybrid strategy leveraging some foundry resources when it makes sense.
Many of Intel’s acquisitions over the years were already using foundry resources, particularly Taiwan Semiconductor Manufacturing Co. (TSMC). However, Intel has mostly remained true to its roots and manufactured its x86 and other key products in Intel fabs. AMD had a similar stringent strategy under founder Jerry Sanders, who went so far as to say, “real men have fabs.” But due to financial struggles AMD eventually spun off its fabs (as GlobalFoundries) and now uses multiple foundries to manufacture its products.
Rather than spinning off its fabs, Intel tried to become a foundry itself to other semiconductor companies. That effort, while good in theory, failed due to several factors. First, Intel has its own process, tools, and in some cases even terminology that differs from those of other semiconductor manufacturers. This made it difficult to transition products manufactured on other foundries’ processes to Intel’s unique process, even if Intel offered a better price.
In addition, Intel was reluctant to modify its process to accommodate the requirements of foundry customers. While Intel has been extraordinarily successful with its copy-exact fab model for its own manufacturing purposes, what is good for an x86 processor is not necessarily good for other forms of logic. Even Intel’s mobile group suffered from having a manufacturing group that, at the time, was focused on performance, not low power and performance efficiency — critical factors in a mobile device.
In 2016, Intel even announced a collaboration with instruction set rival Arm to support manufacturing of Arm-based processors in Intel fabs starting at the 10nm node. Unfortunately, Intel has struggled ramping production of its last two process nodes, 14nm and 10nm, and there was little interest in using Intel when the other foundries were already accustomed to manufacturing Arm-based products. Intel did act as a foundry for Altera FPGAs in 14nm and eventually acquired the company.
Note that Intel’s foundry effort began at a time when the company had excess capacity. While Intel is currently ramping new fab capacity, it has been capacity constrained for more than a year in part due to trouble ramping the 10nm process node. Intel has even opened the door for AMD to gain share in certain markets like desktop PCs, while it focused its manufacturing efforts on its highest profit segments, particularly servers.
Trying to spin off its fabs now, when the company can sell everything it can produce and has exciting new products coming to market, would be a disaster.
Even spinning off the fabs during a period of excess capacity would be challenging. GlobalFoundries found this out the hard way. It took more than a decade and several management and strategy changes for GlobalFoundries to find success under its current CEO Tom Caulfield. It is not easy changing an entity’s business model while transitioning from handling one demanding customer to many demanding customers at the same time. Not to mention the added financial burden of investing in new fabs and new process technology without a parent company with good profit margins on its products.
The Plan (in theory)
Now, everything I said until now indicates that it’s a bad time and a huge challenge to spin off Intel’s manufacturing, so I propose going the other way. Rather than spinning off its manufacturing group, Intel should instead acquire GlobalFoundries.
This would give Intel instant access to more fab capacity for some of the product lines that are already outsourced and even some Intel products that can be manufactured on older process nodes. GlobalFoundries has fabs in Germany, Singapore, and the US. And although GlobalFoundries has stepped away from competing with TSMC and Samsung on the latest process node, it is still one of the top three semiconductor foundries in the world and the most geographically dispersed.
For GlobalFoundries’ largest investor, the Abu Dhabi Government through the Advanced Technology Investment Company (ATIC), the company’s progress to success has been slow and any hopes of building semiconductor fabs in Abu Dhabi appear to be gone. Now that GlobalFoundries is profitable, it can no longer be purchased at a bargain price, but I am sure the Abu Dhabi government would be interested in investing in other industries more likely to diversify the region’s economic base.
Through the acquisition, Intel would gain a management team that understands how to be successful as a foundry, and personnel accustomed to working with outside semiconductor customers. So, I would propose that the GlobalFoundries team lead the integration and eventual transition to being a foundry. Intel could continue to invest in more capacity, including at GlobalFoundries’ newest site in Malta, New York, which has the infrastructure and land for additional fab capacity.
Then, when Intel is in a good position competitively and has extra capacity, it could spin off the manufacturing group while maintaining an interest as both an investor and its largest customer. This would free Intel from the financial and capital overhead of being a semiconductor manufacturer while assuring that it has ample fab capacity for at least the foreseeable future. Intel would still have the option of leveraging TSMC as a manufacturing partner as well as providing the company with a dual foundry capacity with what would likely be the two largest foundries in the world.
The one caveat to this is the Intel culture. As my colleague Francis Sideco will quickly point out — the Intel culture is not amiable to outsiders. An example of this was Intel’s acquisition of Infineon’s Wireless Solutions business to help the company be successful in the mobile, a segment Intel struggled to enter. The newcomers were not welcomed and found it difficult to implement the necessary changes to Intel’s product and go to market strategies. This story holds true for many acquired companies and executives hired from outside of Intel.
New CEO Pat Gelsinger is not facing this backlash because most of his career was at Intel and he was a well-liked executive by many at Intel prior to his departure eight years ago. However, he does face the challenge of changing the culture. I would argue, however, that a change in strategy could be accompanied with a change in culture. This would make the proposed scenario possible if chosen and supported by the executive team. But it remains to be seen what changes the new CEO will make.
As I indicated at the beginning, this is just my theory on how a spinoff of Intel’s manufacturing could be successful for both Intel and the new foundry entity. At this point, there seems to be no interest in pursuing a separation of the manufacturing group, and that is the right decision at this time. But the future holds many possibilities.
— Jim McGregor is principal at Tirias Research.