Thankfully, I am not in the shoes of Intel’s new CEO, Pat Gelsinger. He obviously relishes it, but very few are quite up to his new challenge.

Gelsinger sees the opportunity in the obstacles that have provoked negative commentaries about Intel, including the brow beatings intended to drive the company out of its internal manufacturing operations.

Intel Foundry Services

A few things stand out from Pat Gelsinger’s rapid response to righting the ship. It boils down to flexibility. He might be pushing Intel to do what it once did while encouraging agility in response to current and future trends. I’m sure we will hear a few accusations that Intel is simply reaching back attempting to grasp at past glories.

This is something much bigger than that.

Gelsinger was clearly saying it is okay to restart what they once were doing, but it was definitely not okay to keep doing what they are doing. The trajectory at the end of 2020 suggested a very unpleasant landing.

Before the foundry business kicks off, Intel will be operating as an integrated device manufacturer or IDM. The move to create the foundry business prompted them to brand their new future as “IDM 2.0.”

The success of IDM 2.0 boils down to flexibility. A useful trait that tends to evaporate from very mature companies. Intel’s newfound openness is built on four key areas: foundry independence, advanced packaging, IP ecosystem, and geographic diversity.

Arguably the main message behind the Intel foundry is not the massive $20 billion commitment of capital. It is actually that Intel Foundry Services (IFS) is set up as a standalone company. If it were not, the question of priorities given to internal versus external designs were likely to cause hesitation amongst potential clients. This was at least part of the failure of earlier dalliances with the foundry business.

Indications from the CEO are very strong that IFS operate very much as a pure play foundry.

Packaging

When Gelsinger says “system-on-package,” it isn’t a subtlety or a slip. Intel is positioning to provide full scale system-level integration. It is not the old system-in-package.

System-in-package concepts (source: Amkor)

Adding ever more confusion to the jargon landscape, system-in-package was a bit of the terminology getting ahead of the technology. The original “systems” were typically just two chip types in the same package.

The two descriptions of components bundled into a single package could not sound more alike while attempting to describe such distinct concepts. The assembly houses like ASE and Amkor are still using in package, but I would watch for Intel to continue to push the system-on-package  term to describe a new generation of integration.

System-on-package is the next phase of miniaturization for electronics. The motherboards of bygone days will be implemented into one packaged unit. A viable chiplet ecosystem will enable this change. Intel is pushing that too. It isn’t just dabbling. Intel has physical integration on the packaging side with Foveros and EMIB. They are also actively developing interface standards for chiplets, an often discussed sticking point for chiplets to get off the ground.

System-in-package concepts (source: Amkor)

It is not part of the technology strategy, but a significant aspect was Gelsinger’s promise that Intel would accept wafers from anywhere for package integration. A closed shop mentality for wafer foundries with advanced packaging in-house would slow innovation. A perceived unwillingness to allow a full mix and match manufacturing ecosystem has created some hesitancy for design companies to get on board with a transition to dissagregating system-on-chip designs into chiplets.

Like every other human endeavor (and the creatures themselves), a company matures and may grow out of its early aggressive, hungry phase. Conservatism takes over. Arrogance sets in, and the approach to the market and the technology itself all face the risk of inflexibility to the point of ossification.

The old Intel might have just felt it was a little too distinguished to stoop to pick up that dirty old dollar bill on the sidewalk. The present and future Intel now would have to veer out into the busy street to avoid the mountain of cash sitting there. Gelsinger indicated that the foundry business could top $100 billion by 2025.

System-in-package concepts (source: Amkor)

Of course, this was the polite way for Gelsinger to point out the opportunity rather than Intel’s earlier failures to act. That inaction led to a series of events from financial analysts and investors calling for an end to internal manufacturing to a new CEO to finally the announcement of IDM 2.0 — all since the final days of 2020.

IP

A key component for a viable foundry service is the availability of design IP for its clients. In Gelsinger’s web presentation, the new Intel flexibility was on full display. Intel promises “a world-class IP portfolio that customers can choose from including x86 cores, graphics, media, display, AI, interconnect, fabric and other critical foundational IP, along with ARM and RISC-V ecosystem IPs.”

IFS will accept a range of third party IP as well as offering Intel’s own including x86 cores. Intel sees this expanded IP opportunity as a differentiator over the other big foundries.

Intel will complete its foundry offerings with design services.

Location, location

Intel believes it can set itself apart with another key feature: supply chain security.

Calculated by revenues, indicated that TSMC manufactured over half the world’s chip content last year. Rolling up UMC and some smaller foundries, the total for Taiwan was 63%.

Samsung’s foundry share was 17%. Operating as an IDM and as foundry, Samsung is a model that most closely resembles Intel’s new strategy and surely a very well-studied business case among Intel management.

Right now, Intel operates semiconductor fabs in the US and Ireland. The two new foundry facilities announced so far will be built in Arizona, so most of the capacity will be in the US.

Although TSMC will also build a new fab in Arizona, and Samsung is operating (and planning to expand) in Texas, Intel would still be the biggest player outside of Asia for customers looking to keep their supply chains geographically diversified.

Intel Worldwide Manufacturing Sites (source: Intel)

But the US isn’t the only jurisdiction that has seen a once proud IC manufacturing capability all but disappear (at least for the most advanced silicon processes). Europe is in the same boat. The EU will surely try to regain some of that former glory because if not now, they will have to be content with never.

And Intel may well be set to satisfy that hunger. As Gelsinger stated, “I expect that we will be ready to announce our next phase of expansions in the U.S., Europe and other global locations within the year.”

Time will tell what that looks like, although current pace of change under Gelsinger suggests it might not take that long. As pointed out by , Intel is likely shopping around hoping to get a few jurisdictions bidding against each other before they commit to locations for any additional manufacturing sites.

A new approach is usually refreshing. The business agility invoked by Intel’s new CEO is new and based on sound reasoning. That combination is uncommon. It might also be a winner.

The post appeared first on .