Intel’s Strategic Quandary

The hits keep coming to Intel. Yesterday, Intel posted earnings reflective of a decent quarter revenue-wise but it was the forward-looking commentary that is much more concerning.

The first takeaway that is relevant to watch for regarding many companies, not just Intel, was management commentary that data center momentum was slowing down. If you look at Microsft’s earnings, you see a similar deceleration related to Azure. Nvidia also reported some datacenter decline in areas as well in their May earnings call. That’s enough data points to believe data center is slowing to a degree, and we will see what AMD says next week as well as Amazon’s AWS commentary next week as well.

But that is not the market dynamic that has Intel’s stock off 16% as of this morning. It is the news that their 7nm process and subsequent chips on 7nm are delayed six months. Given the challenges Intel faced with moving from 14nm to 10nm, this would stack back to back process generation delays, which only further puts Intel behind their competition and will likely lead to continued competitive challenges and market share loss to competitors, specifically AMD.

I will be the first one to defend a position that says process technology leadership is not everything. However, being competitive from a product standpoint does require not falling dramatically behind from a process technology standpoint. It was AMD that demonstrated this as they first sold off their own internal manufacturing when it became clear they could not keep up with Intel, and then similarly moved away from Global Foundries (which was the AMD fab assets spun-off) and joined at the hip with TSMC which has now paid off in dividends strategically and financially. Falling too far behind in process technology is the absolute worst-case scenario for Intel, no matter how you spin the narrative.

To help you understand the timeline, Intel is now saying late 2022 or early 2023 for their 7nm products. Intel CEO Bob Swan indicated they figured out the root issue and believe the 6-month delay is all that should be expected. But let’s look at that same timeline and see what TSMC is saying. TSMC’s management talked about 4nm in H122 and 3nm in H222. The takeaway from that is Intel will be rolling out their 7nm products at the same time TSMC is rolling out 3nm products. In a best-case scenario for Intel, they will be 1.5-2 generations behind customers using TSMC process technology. With AMD being chief among their competitors in CPU/GPU, it is hard to not firmly believe AMD has a great deal of potential growth ahead to take share from Intel in both datacenter and PC clients.

The data center is a critical battleground because of how lucrative it is, even though the volumes are low compared to PC clients. The data center is the biggest chunk of Intel’s revenue, so this is of higher strategic importance. One thing Intel management has now started saying, that many views as a positive are their willingness to look at outside fabs for assistance in manufacturing. Intel stated they would use an outside foundry, most likely TSMC, for their upcoming data center GP-GPU Ponte Vecchio to keep it on track. I believe this upcoming GPU from Intel is of utmost strategic importance to their product portfolio, and that is telling that it will be built on someone else process technology. To be entirely honest with you, there is a part of me that simply can’t believe it has come to this for Intel.

It’s also worth mentioning a few things about PC client hardware. While less than 30% of Intel’s revenue, client is where a lot of the glory is in semiconductors because it is the tangible part of the business most humans get to see and touch. Remaining relevant in client PC hardware is extremely important. With Intel’s delays now in 7nm, this makes Apple’s move to their own silicon that much more interesting and potentially strategically detrimental to Intel.

Intel has historically been the company to bring leading-edge process technology products to the PC client space. The leading-edge process means better performance per watt, which is extremely relevant in bringing better performance with better battery and power management to notebooks. So you can argue in PC notebooks, where performance per watt is extremely relevant, then competing with leading-edge process tech is incredibly important.

AMD is bringing their 7nm Ryzen products to notebooks potentially by the end of the year but early 2021 for sure. But the interesting story here is Apple. I am not sure what process generation from TSMC Apple’s upcoming silicon for Macs will use, but even if it is on TSMC 7nm, Apple will still be out the gate first with leading-edge process technology. It is also a safe bet Apple will be first with 5nm, and beyond due to TSMC now leading the charge in process technology.

All of that to layout the point that Apple may very well put much more competitive pressure on the PC OEMs than I first thought last month. If Apple’s Macs, running Apple Silicon, take the clear lead in benchmarks, and a big reason for that being not just their design. Still, because they are running better process technology, then their devices will stand head and shoulders over competing PC OEM hardware. If Intel has no solution even close, then PC OEMs will increasingly turn to AMD, who should “technically” remain more competitive with Apple. To emphasize my point here, to better compete with Apple, OEMs will need to more deeply embrace AMD if this scenario plays out.

So now, the biggest question for me is what does Intel does with their fabs. Being vertical, designing, and manufacturing their chips was once an incredible strategic advantage for Intel. Yet, it is now looking like one of their biggest strategic disadvantages. The strategic quandary they are in is they can’t move their CPU manufacturing anywhere else because TSMC doesn’t have the capacity. Perhaps Samsung could, but even that is somewhat questionable.

Intel can outsource more of the promising companion parts like GPU, AI accelerators, etc.. Still, they have to deal with this issue with their CPUs being tied to their process, which is consistently causing problems. In the back of my mind, I always had a worst-case scenario for Intel, where they had to pivot with regard to their fabs. Either sell them off or have someone like Samsung or TSMC collaborate on designs made in Intel fabs. We are sadly at the point where to meaningfully move forward, structural change must occur for Intel.