First, let's take a quick look back at Intel's current state of affairs. First, despite its bold promises to regain technological leadership, it has fallen far behind TSMC (Taiwan Semiconductor Manufacturing Company) in chip manufacturing and, in fact, is increasingly dependent on TSMC to manufacture its latest and future CPUs such as Meteor Lake, Lunar Lake, and Arrow Lake. In fact, it is increasingly dependent on TSMC to manufacture its latest and future CPUs such as Meteor Lake, Lunar Lake, and Arrow Lake.

On the other hand, it has lost market share to AMD in server CPUs, and Qualcomm's Snapdragon X Arm-based chips are a real threat in Intel's largest consumer market, notebooks.

More recently, Intel announced very poor financial results and decided that it needed to lay off another 15,000 employees after already cutting 5% of its workforce last year.

This is only a stark outline. There are several other lesser-known horror stories, including the curtailment of plans to build factories in Europe and the fact that things have gotten so bad that Intel is canceling dividend payments to shareholders. But the obvious question is. How did Intel get it so wrong?

Perhaps one could say that the corruption began when Paul Otellini replaced Craig Barrett as CEO in 2005. Barrett had a PhD in materials science from Stanford University. Otellini was an economics graduate with an MBA. Simply put, Intel's top management has replaced engineering with sales.

Certainly, Intel's next CEO, Brian Krzanich, came from the engineering side, but let's not let such details get in the way of an easy narrative. The story is that Intel became lazy and cynical under sales and marketing-driven leadership, tilted to a near-monopoly position in the x86 market, and it all went downhill from there.

More specifically, the most obvious sign of Intel's decline was its troubles with the 10nm node. Originally scheduled to go into production in 2016, Intel first admitted to delaying the node in mid-2015.

Strictly speaking, Intel released a small number of 10nm dual-core Cannonlake CPUs in 2018. However, 10nm did not become fully operational until the Ice Lake mobile CPUs released in late 2019. Presumably, Ice Lake laptops would not become available in large quantities until 2020.

In the cutting-edge chip industry, this is a catastrophic delay. In fact, if you consider the Intel chips currently available for purchase in PCs, the situation is hardly any better. These broken 13th and 14th generation desktop CPUs use the Intel 7 node, a rebranded and revised version of the 10nm technology that was supposed to be released nearly a decade ago.

The Meteor Lake laptop CPUs get Intel 4 silicon, which is a chiplet design with only the CPU tiles made of Intel silicon. All other functional chiplets, which make up the bulk of the entire processor, are made by TSMC.

Oh, and Intel admits that it is struggling to increase production of even just the small CPU tiles made with Intel 4. Sales of Meteor Lake laptops have been limited by the amount of processors produced, and the latest rumors indicate that this problem is still ongoing.

It is said that the problem is still ongoing.

Whatever the case, it is certain that both active tiles of Intel's next generation mobile CPU, Lunar Lake, will be produced by TSMC, not Intel. Only the passive base tiles, which are essentially good-looking PCBs with wires, use Intel silicon.

It is not entirely clear how Intel's next generation Arrow Lake desktop chips will be manufactured. However, it is certain to be a chiplet design, and rumors suggest that most models will use TSMC silicon for all active tiles, and perhaps a few low-end SKUs will use Intel silicon for a single active tile.

In other words, Intel is continuing to increase the percentage of TSMC in its chips, including those not yet released. It will be many years before Intel will be able to reverse its ad hoc strategy of farming out production on the grounds that it does not have enough of its own production nodes.

It's worth thinking for a moment: now in 2024, Intel is increasing the percentage of production at TSMC for future processors. [The roadmap produced by Intel Foundry Services (IFS) is further evidence that Intel is far from truly turning the corner. This is the business unit that Intel has separated from its fabs (chip factories).

Intel wants to move from using its own fabs to produce its own chips to also becoming a major contract manufacturer of chips for its customers. In its new role as a customer fab, it hopes to not only make better chips in-house, but also compete directly with TSMC.

So there's the problem. According to IFS' own figures, which seem more optimistic than pessimistic, the majority of Intel's production capacity will not use nodes equal to or larger than Intel 4 until 2027. even in 2026, about 65% of Intel's capacity will consist of 10nm-class nodes, now called Intel 7 or earlier nodes, currently called Intel 7.

Furthermore, if there are still rumors of yield issues with Intel 4, what are the odds that the upcoming 20A and 18A nodes to be released will be zinged from day one?

Indeed, given the debacle of the 13th and 14th generation Raptor Lake desktop CPUs, rumors of ongoing problems with Meteor Lake supply, and Intel's latest financial situation and belt tightening, it is hard to be sure things are not continuing to get worse.

This is especially discouraging given that Intel's newest CEO, Pat Gerziginer, will take over in 2021. Gerziginer is supposed to be one of the good guys, an Intel engineer who was forced out of the company during its marketing-driven days and then came back to help turn it around.

Finally, the engineer at the helm was back. But as the years go by, Gelsinger's appointment is becoming less and less a case study in how leadership can save a stalling giant.

It is clear that Intel is not done yet. It remains a giant with vast resources, and in two years' time there will be great CPUs on the market, plus a third generation of Arc graphics cards that may well take the fight to Nvidia and AMD.

Regarding the latter, it would be a shame to see Intel Arc graphics fall victim to Intel's need to cut costs; the PC gaming graphics market desperately needs more competition, and a healthy Intel would be good for us all.

The same can be said for CPUs, but perhaps to a lesser degree. In the long run, it is hard to imagine that PCs will not eventually assimilate to the ARM architecture. Qualcomm's Snapdragon X chip appears to be a serious first step in that direction. And if Arm were to take control of the PC, all kinds of competition could ensue. Not only Qualcomm, but Mediatek, Nvidia, and even AMD and Intel may start producing Arm-based CPUs for PCs.

This is somewhat speculative, but what is clear here and now is that Intel needs to act quickly. The first thing Intel needs to do is to better clarify the status of the Raptor Lake processors. Indeed, they have now announced a two-year warranty extension, but there is still a lack of clarity on this matter.

After that, the fab will need to be sorted out. But it appears to be a far, far taller order.