Late last week, Intel announced that they were delaying their next generation of chips. This is big news and the stock took a tumble after the announcement. This has big implications for the whole industry, and arguably the world.
The issue is that they announced a delay of their 7nm chips by six months. They were already behind schedule with 7nm, by contrast TSMC has been shipping others’ 7nm products for over six months. The fact that this comes on top of Intel hitting 10nm roughly two years behind schedule speaks to two big challenges for the company.
First, this has all sorts of impact on their financial results. Not least, they do not have sufficient capacity for their trailing edge, 14nm process, which has crimped revenue from their inability to meet demand. This hurts the whole PC ecosystem, as large companies delay their purchases of hardware which ripples through the whole supply chain.
This also opens the door for Intel’s competitors. On the PC side, this makes Apple’s move to their own silicon even more important. Apple likely knew of Intel’s delay a few months ago, and likely catalyzed them to move even faster with their transition. More importantly, it creates a big problem for Intel in the data center, which is their largest profit driver. Server customers want the best performance. Cloud scale providers – who consume something like 70% of Intel’s data center products – make purchases based on performance per dollar per watt. Any competing product just got a big leg up in that math. All those competitors are using 7nm now, and basically have a year to further erode Intel’s once-dominant share. There are other factors that will keep those customers using Intel, but the math plays a big role in determining the purchase decisions. For companies like AMD and privately-held Ampere, this is very good news. It also gives a big boost to anyone else who still has an ARM CPU in the works, companies like Huawei and Amazon.
Beyond those immediate financial problems this news poses what we see as Intel’s identity crisis. For almost thirty years, the company has prided itself on producing the world’s most advanced semiconductor manufacturing technology. They ceded that with the 10nm miss two years ago, but they could have convinced themselves that was a one-time fluke. Fool yourself once…. The rest of the world has moved to a fabless model, but for Intel that vertically integrated model goes to the core of their identity. Now it is not clear what they are.
After the announcement, the press focused on Intel’s admission that they would shift some GPU production to TSMC. This is a bit misleading as Intel has been producing chips at TSMC for close to a decade. They acquired Infineon’s mobile business Altera for FPGAs and have been running those as fabless through TSMC since acquisition. In hindsight, the fact that they could never shift those to internal production should have been viewed a major Defcon-1 Red Flag.
And this leads into the geopolitical realm. The whole industry has recently woken up to the strategic value of Taiwan’s TSMC. In our earlier writing on the geopolitics of chips we have been careful to call out the fact the US does (did?) have advanced semiconductor manufacturing capabilities. With Intel lagging, TSMC’s unique position becomes all the more apparent.
Intel still has a lot working in its favor. They still make a lot of money, $10 billion in the latest quarter. They have a tremendous reserve of talent, especially if recent executive changes means they can rehire employees who have exited over the past few years. They will get to 7nm eventually, and still have a credible path to 5nm and whatever lays beyond that.
That being said, Intel needs to think long and hard about its future. Do they want to be a 3rd party foundry and produce chips for others? They could still do that, but it would require tens of billions of dollars of investment, starting right now. Stay the course and milk the markets they still dominate? That sounds tolerable, but it is effectively the strategy they have employed for the past ten years, thus not sustainable indefinitely. Probably the best thing they could do would be to hunker down and invest heavily in their existing team, committing to never missing a deadline again. This would lkely require big organizational change and a major investment. Long ago, we semi-joked that Intel should go private and make the required changes away from the spotlight of the public market. That particular financial structure remains far-fetched, but the depth of change required still pertains. Intel needs to fix itself, and no amount of government subsidy will alter that.
TSMC’s unique position becomes all the more apparent. Unique, indeed, since TSMC is located in China–according to the Taiwanese and the PRC constitutions–and there is nothing the US can do to prevent the PRC from making that de jure reality a de facto actuality.
I think there is a lot the US can do. Whether we are capable of building alliances and common spheres of interest, diplomacy, and applying various forms of pressure and persuasion. This is complex, subtle stuff, and the PRC seems to think the US is no longer capable of doing this. They may be right, but maybe not.
Pingback: Intel Selects Option C – All of the Above | Digits to Dollars·