While the large number, such as 90 percent of market share, looks suitable to investors, not all markets are worth perusing.
Bob Swan, Intel’s CEO, who used to be a Chief financial officer, is getting Intel back on the track. It is hard not to mention that we were right that Bob will be the new CEO, months before everyone else. What is crucial in the semiconductor industry is making money out of the silicon. In other words, it is financially much better to sell a single Core i5 than dozen if not hundreds of low-cost IoT chips, as the margins are better. Since Intel has FABs, getting them on track is one of the critical targets.
Away from entry-level
Intel has traditionally left the bottom-feeding segment in PC and notebooks to competitors like AMD and many others. While the company still covers this market well with Celeron, Pentium, and what use to be called Atoms, the company is not willing to sacrifice margins. Intel like every other company needs to focus on what they are good at and spend the money, IP and engineering on that goals. Instead of making the mediocre 5G modem, they can make a great AI accelerator, best IPC next generation CPU core or the best data center platform. That is what the company is good, not in making a few cent chips for the simplest IoT device. The company will leave this to others.
Market share ≠ revenue
I remember a conversation in early 2017 with a high-profile executive at Nvidia. Back then, I told him that AMD is building the channel with underpriced inventories in both CPU and GPU business. The key objective was to win the market share. Investors react well to Lisa’s announcement that the market share grew. That was the principal target, and the goal was accomplished.
RTG did this without product
Raja Koduri formed RTG (Radeon Technology Group) in late 2015 and managed to grow the market share without any new product at the time. It was all built in the promise that AMD is getting committed to the graphics in a big way, and the revenue and products did follow. They were not wrong with the commitment, and they did cause some severe fuss about the graphics and did come close to Nvidia in some market segments. Betting on HBM2 was a good and sturdy strategy but one that simply did not pay off. Nvidia is very vigilant, especially when they are winning.
AMD EPYC strategy
Lisa Su and her team at AMD used the same tactics with Zen and server EPYC.
AMD’s game is market share first, get the designs wins, and then try to inverse the margins and make some decent money. This strategy worked quite well for the company over the past few decades. The tactics work as AMD went from $1.67 in July 2015 to around $44 at press time.
Nvidia was always vigilant about making good margins. Nvidia got rid of mobile phone business as companies like MediaTek swept in the market and destroyed the margins. Nvidia, who likes to operate in very healthy margins, decided to redraw from this market and focus on Automotive and AI, both bets played well for the company.
Intel double CFO strategy
Now back to Intel after a short history lesson. Bob Swan was a CFO, and he hired ex Qualcomm CFO George Davis to run the finances as of April 2019. It is essential to understand that these guys are calling the shots and want to come back to a healthy financial environment. Intel is still making good margins, but it has to come back on track with manufacturing and product execution.
One of the most recent business decisions/cuts was the modem, a business unit that sold to Apple. The tough decision is clearly a business sense driven decision. Business-wise it is not healthy to make a modem for a single customer, even if that customer is Apple. Just remember what happened to Immagitionation technology, the sole provider of GPU IP to Apple as Apple replaced them with its own GPU.
Under the leadership of Aisha Evans, Intel has spent billions of dollars developing modem and hoping to fight Qualcomm in 5G and realized that this is the fight it cannot win. The CFO decided to scrap the project and refocus in 10nm and, more importantly, 7nm roadmap, products and manufacturing. Intel has to get to Exascale with its CPU, GPU, Memory, FPGA and AI accelerators as quick as possible. That is the target.
5nm, 7nm Intel
The 7nm that is the equivalent to TSMC 5nm is expected in Q4 2021, almost a good year after Apple launches if first 5nm iPhone SoC solutions in September 2020. It is one thing to make a tiny mobile SoC and another to launch a full-blown GPU in 5nm. TSCMs 5nm is Intel’s 7nm; it is all about the marketing, isn’t it, and 5 nm does sound much better, doesn’t it?
Intel already reported that the GPU is the first 7nm product as the new process brings a promise of both surface gain as well as lower power per transistor. It is all about the compute, machine learning, AI part of the GPU, not betting Nvidia in gaming. Going after AMD and Nvidia on the desktop is undoubtedly one of the goals, but not the top target.
Intel is committed to challenging Nvidia and AMD in the desktop and mobile market, but more importantly, Intel needs a GPU for its data center business. The goal is to have Exascale performance in 2021 with two Xeon scalable Sapphire Rapids and Six Pointe Vecchio GPU in a single Aurora node. That is what 7nm for Intel is all about.
Execution and leadership
Execution and manufacturing hurdles are improving, but companies like AMD are catching up and putting a lot of pressure on both the desktop and server/data center market. Intel won’t give this without a fight, while the company still dominates the notebook market.
Intel still has the IPC leadership even with 14nm or it is losing very tight to AMD on desktop. Still, the market wants 10nm in both server and desktop as soon as possible, but for now, 10nm was the right call for notebooks, and it will expand to other parts of the markets in 2021.
Intel’s two CFO strategy and the recent acquisition of Habana AI accelerator is another good shoot. They got Nervana two years ago, but obviously, Habana has something that Nervana and Intel don’t.
Mobileye from where I am standing today was an excellent acquisition, but it didn’t seem reasonable back than 2017, as 15.3 billion sounded like a lot of money for the „camera company.” It is easy to see today that Mobileye is way more than the camera company and has a solid ADAS and self-driving roadmap.
Intel’s old guard including Bryant GB in charge of client products, Sandra is chasing Modems, Navin in charge for data center and many others who are convinced that things will get brighter under Bob’s new leadership. Having some world-class talents like Jim Keller, Murthy and Raja indicate that the company’s future is in good hands.
With the help of the women and man at Intel, Bob and George Davis will make the acquisitions, hire the right talents, fix what is wrong with the Fabs, maybe cut some loose ends in the process, but will get the big compute giant on track.