Pat Gelsinger took over as CEO of Intel a little more than 28 months ago. In that time, the company hunkered down to solve the existential crisis it faced centered on its manufacturing process. Over two years of fighting for its life, the company now seems to have turned the corner, they are going to make it. But like the hero in a zombie movie they are awakening from a coma to find the world radically altered, and suddenly that fight for its life looks like a smaller problem given the direction everything else has gone..

On their last earnings call, Intel confirmed that they have begun sampling their first EUV products including the 20A process which is ‘on par’ with TSMC’s 3nm process. It now looks increasingly likely that they have finally fixed their manufacturing process and significantly narrowed the gap with TSMC. If they had not been able to achieve that the company would be in a very bad place, likely forced to split in two and take itself apart in many other ways. But now that it looks like they can manufacture competitively they have a new lease on life.

Of course, that just means they have moved up the next level of challenges. Now they will have to produce competitive chips, with all the features that customers want. And while they were engaged in that internal struggle, the competitive landscape shifted dramatically. AMD has gained solid market share in the critical data center market. This process began long before Gelsinger took over, but he now has responsibility for stemming the tide. Most people assume that the company can reassert its privileges and return to mastery of the data center. Intel has a 40 year history of making the best CPUs in the world, surely with their manufacturing fixed they will go back to making the best CPUs again. At best, we think the jury is still out and it is not clear how competitive their next generations of CPUs will be. And we are more optimistic than many we speak with who think Intel’s design teams had been coasting for a long time, relying on the manufacturing side too heavily. Last month, Intel announced that they would change the accounting relationship between its fabs and the design side of the house. That laid bare the many ways in which Intel’s sales organization used the fabs as a crutch. It is reasonable to wonder how much the engineering side did the same.

More critically, the pattern of data center purchases has altered dramatically. When Intel entered its downturn a data center was a giant building filled with CPUs and all the equipment necessary for operating them. Now data centers have become fully heterogenous filled GPUs, FPGAs and AI accelerators. There are still a lot of CPUs required but they share their domain with many others. The big hyperscale customers have become ever more demanding and diversified in their purchases. Even if Intel’s CPUs turn out to be highly competitive (a big if), the rest of the portfolio is mixed, at best. They have Altera FPGA’s which should help, but that business has been languishing for years under Intel, so it will require its own retooling. Intel now has GPUs, but these are in early generations and appear to be far away from challenging Nvidia in any meaningful way. Intel has AI products, its Gaudi line acquired via Habana, and maybe these are appealing to customers, but again the jury is out. The last big question mark is how Intel will approach the custom ASIC business, supporting the hyperscalers’ internal designs. This is an important business, but Broadcom, Marvell and AMD are way ahead of Intel in building out this capability. All in all, Intel has a lot of holes in its line up.

And Intel’s problems are not contained to the data center. The virus has spread to the PC market as well. Arm-based laptop CPUs from Qualcomm are rounding error in the market, but the advent of AI offers them a once-in-a-lifetime opportunity to shift that market. More critically, Apple has the most competitive laptop CPU on the market, and is sucking up that segments profits as a result.

In fairness, Intel is full of very capable people and they are not asleep anymore. It’s just that from the outside the list of challenges looks formidable. And Intel does not get any breaks, the competition is running fast and clawing at the door.