Intel, a Love Story

I have a huge amount of admiration for Intel, outweighed only by my disappointment in their execution. I had high hopes when they brought back Pat Gelsinger to lead the company in 2021. But his unceremonious exit last month signalled the end for this generation of Intel’s evolution. This isn’t the first time the company has faced down an existential crisis though, and I’m betting against the odds for them to recover.

Generations 1 & 2

Intel was founded as a memory company, and they were very successful early on with the best selling DRAM chips on the market. However, the technology was quickly commoditized by Japanese competitors who were able to better optimize manufacturing efficiency. So Intel was forced to pivot, and moved into CPU design and manufacturing.

This move had some strategic advantages, namely that they required a lot more R&D and that instruction sets could be made proprietary and force vendor lock-in. CPUs were thus less likely to become commoditized, and with the emergence of the personal computer there was an opportunity to establish hard requirements around ISA compatibility which they could control.

The pivot from memory to CPUs was wildly successful. The commoditization crisis was averted, and they went on to form a duopoly with Microsoft that would make them the dominant semiconductor designer and manufacturer for several decades. However, Andy Grove who was CEO at the time famously said “only the paranoid survive,” and wouldn’t let their market dominance lull them into complacency. Intel created a culture of strong investment into R&D that would keep them ahead of everyone else.

Generation 3

There was no more room for growth in the CPU space. Intel already dominated the entire market for desktops, laptops, workstations, and servers. So, to grow the company they went wide and pursued other semiconductor product lines such as: networking, motherboard chipsets, storage, graphics, IPUs, FPGAs, ASICs, and some really out there stuff like silicon photonics and neuromorphic processors. They went much wider on R&D than any other semi designer, and regardless of the outcome it’s what I admire most about the company.

The problem is it didn’t work. Aside from a few point solutions they weren’t nearly as dominant in any of these spaces as they were with CPUs. This was compounded by their failure to adopt EUV lithography in their fabs, instead hoping (and failing, badly) to get a few more generations out of DUV.

This is where we are today. Their product lines are bloated and underperforming, and their fabs are badly lagging competitors. This generation is over, and the sooner this is acknowleged the better.

Generation Next

So how can Intel recover from all of this? Software. That may seem like a strange take for a hardware company, but they are uniquely positioned to become a vertically integrated hyperscaler. Vertical integration is difficult to achieve and easy to lose, there are calls today for Intel to divest its fab business (as AMD did with GlobalFoundries) but it’s essentially impossible to rebuild that manufacturing expertise once it’s lost.

Intel has the existing product lines necessary to create a competitive hyperscale cloud provider, and their advantage in silicon design would allow them to create application accelerators (as they did with Sapphire Rapids) specific to various managed services. Furthermore, OCP and Oxide are examples of the commercialization of hyperscale technologies. So while the barriers to entry may be substantial, they can reasonably be overcome.

Intel’s semiconductor supply chain and domain expertise would put them at a unique advantage against existing incumbents. Cloud providers are continuing to eat the world, and hardware acceleration of all kinds is quickly becoming a key differentiator.

Bet on Success

Being an Intel shareholder has not been a pleasant journey, but I still think they have a huge amount of talent and potential. The market thinks that I’m wrong, which I probably am. But it would be a hell of a thing to watch Intel turn things around and reinvent itself one more time.