The big picture: Intel has ambitions to build a foundry business by making chips for other companies. This is an important strategic move that the company needs to recoup the massive investment it is currently making in fabs around the world.
Most analysts agree that this proposal is a no-go until Intel can catch up on its manufacturing process. And rightfully so, without it Intel Foundry Services (IFS) would lack serious competitive differentiation (yes, packaging, but that’s not enough). And we’ve also warned that the company is running out of muscle when it comes to customer service after all the fuss about the fabs team running the show.
Guest author Jonathan Goldberg is the founder of D2D Advisory, a cross-functional consulting firm. Jonathan has developed growth strategies and alliances for companies in the mobile, network, gaming and software industries.
But beyond that, there are even more concerns.
When we talk about semi-finished products, we tend to treat the various foundry nodes as fungible, as if a customer could easily trade Samsung for Intel for TSMC. This is not correct. In reality, each fab has a different way of doing manufacturing. Intel’s process is designed to make processors, and these processes are not the same as those needed for other types of chips. For example, mixed-signal chips that must process digital versions of analog signals (like cell phone signals) while performing pure digital logic require a very different set of manufacturing steps—different machines, differently calibrated, in a different order. To put it politely, Intel does not have a good track record in producing its own mixed-signal parts.
And then there are the tools. An important part of the foundry customer experience is the software tools used to manage the communication between the fabulous customer and the foundry. It’s not as easy as sending a few files via email. The major foundries have all invested heavily in integrating the tools they use to manage production with the software tools their customers use.
This is a large part of the competitive influence that EDA vendors such as Cadence and Synopsys have over the industry. To be clear, Intel has its own set of tools to handle the production of its own chips, but these are surprisingly proprietary to Intel. From what we hear, even Intel employees don’t like the experience. Will IFS expect customers to learn these tools? More likely, Intel will have to invest heavily in software to create an entirely new set of tools that customers will be familiar with.
All of this leads to the question: If Intel can get its manufacturing back on track (a big if), which customer will want to switch to Intel? Even if Intel can surpass TSMC’s process by 2025, as they claim (or is it 2026?), it will take fab customers a significant amount of time (measured in years) before they feel comfortable sending real production orders to IFS. .
The industry is full of horror stories about companies stuck because their casting process doesn’t go as planned.
It’s important to remember that every time a company sends a chip to the foundry for production, the company is taking a risk. There’s a reason they call them “Risk Starters.” For established foundries, this risk is measured in terms of yield, the percentage of chips produced on each wafer. But for a new process, let alone a new foundry, there is a real risk that the process won’t work. The industry is full of horror stories about companies stuck because their casting process doesn’t go as planned.
Given all this, it seems highly unlikely that any of the major chip companies would be in a big rush to sign up to IFS. Could Apple risk an iPhone cycle failing because its foundry isn’t ready? Just going from TSMC 5nm to TSMC 3nm is scary enough. If Intel can fix its process (again, a big if), big customers will certainly take a serious look at IFS, but that serious look will require years of meetings and extensive due diligence.
Having said that, there is one group of customers who could be very interested in rolling the dice on IFS – RISC V chip designers. Today, these companies do not have many good options. They can get access to TSMC, but they are usually very small, so they don’t get the “A Team” service or pricing.
RISC V chips also have the advantage of being a fairly new option, so new that no one has much experience fine-tuning casting processes to make them. Remember back in February when Intel announced that they were making major investments in RISC V? At the time, we noted that Intel was not going to upgrade its processors from x86 to RISC V, but they were interested in the wider ecosystem.
Much of the investment they talked about was in creating the tools and processes needed to attract RISC V-fabulous designers. It’s worth pointing out that Intel’s RISC V team appears to be largely located within the IFS organization.
It’s not a bad strategy. Under normal circumstances, we would compliment Intel’s wisdom. RISC V companies could easily be a springboard for IFS, a whole bunch of beautiful captive guinea pig customers to experiment with to hone their customer service skills. The problem is that all of this is still a long way off. We recently pointed out to a friend that we don’t expect to see the IFS truly operational until 2030, and he replied, “Oh, so you’re being optimistic.”
Intel is on the right track, it’s just a long way off.
Editor’s Note: Shortly after this article was written, Intel Foundry Services President Randhir Thakur told EE Times that the US Department of Defense is IFS’s “#1 customer” under the DoD SHIP program. There may be obvious political implications in Intel’s choice, but it still indicates that Intel has yet to execute and prove itself as a foundry that supports customer projects.