Last week we published a piece arguing the not-so-outrageous case for Intel to get acquired by private equity as a forcing function for the drastic cultural change it needs to save itself. Judging by the number of people who read that piece with “Intel.com” in their addresses, we suspect that we struck a nerve. In that piece we noted that Intel is now heavily reliant on its data center products. This week Intel hosted a press event and released a slew of press releases about its roadmap. Their goal was to try to assuage the Street’s fears about the slowdown in its manufacturing prowess. Judging from its share price and a slew of downgrades, they did not accomplish that goal.
The Street went into the event already skeptical, and so it is not surprising that a much bigger topic of conversation was a report from a third party industry analyst group. That firm, Semiaccurate, painted a very bleak picture about Intel’s roadmap. Some will argue that Semiaccurate has a bit of a history of being Intel bears skeptics, and that we should discount their comments. However, a quick read of their piece shows that they are mostly just stating facts, with a few insights and opinions. Say what you will about Semiaccurate, but they have technical chops beyond most in the industry.
The gist of their article holds that Intel’s roadmap is much worse than the company has been admitting publicly. With essentially no one left in competitive landscape except for AMD, the post makes clear that AMD, using third party foundries, is moving way ahead of Intel in this crucial data center space. Read their article if you want the full technical details, but the high-level summary is that AMD’s products are now on par with Intel for performance, and their roadmap actually propels them ahead of Intel parts later this year. Intel looks set to remain behind on performance until 2022, which is a very long time in semiconductors. And that is just comparing raw performance. If you look at Total Cost of Ownership (TCO), AMD is already ahead because their products are much cheaper and arguably consume less power.
If true, this is bad news for Intel along several axis. First, there is the very real problem that they are going to lose share in the data center for the next few years. Now that market is growing very strongly as the world builds more cloud data centers, which should buffer the impact for Intel somewhat. However, it does seem likely that Intel will have to cut prices, hurting margins. Recall from our last post that Intel needs to keep its fabs (manufacturing lines) full or risk a much more serious decline in margins. The secular decline in PC demand has put serious strain on this model, and losing margin on data center products has the potential to be a bigger problem.
Another concern is that Intel is dependent on its CPUs to drag other products into sales cycles. So long as customers need Intel CPUs they are willing to buy Intel FPGAs, memory, machine learning and networking products. However, customers are already reluctant to take any more Intel product. If they stop buying CPUs, they are likely going to buy less of all the other parts as well.
That being said, there are still many ways that Intel can hold the line until they get their manufacturing mojo back. Anyone who has ever competed with Intel is likely familiar with the Intel playbook. First comes the FUD marketing campaign, which clearly kicked off with this week’s event. Next, the Intel sales force will go into full channel control mode. Intel has many levers and points of access at its customers. For instance, memory is possibly even more important to data center operators than CPUs, and Intel has a big business there. Intel is also a crucial supplier to the ODMs and OEMs who assemble and integrate servers. The Taiwan electronics complex is still heavily reliant on Intel and will be reluctant to antagonize the company. At least for a little while.
Intel can also gain some comfort from the fact that they really only have one competitor in the data center, AMD. No one else is running a scale server CPU business. Cavium has a solid ARM-based product, but they have just been acquired by Marvell, and integrations always cause uncertainty. There are also companies in China producing processors, but that is a whole other topic. In Intel’s defense, we know that many customers are a little bit wary of AMD. This company has its own historical baggage to overcome in the minds of customers. As one architect told us, “AMD can never seem to keep a roadmap together for more than three years.” In AMD’s defense, the company is run very differently now than in people’s memories. As they say, past performance is no guarantee of future mis-performance.
Intel is still a massive company and will continue to throw off enormous profits for the foreseeable future. Nonetheless, we think many people at Intel have to be very concerned about the company’s future.
DISCLOSURE NOTE: We own a small position (200 shares) of AMD.
Pingback: And then there were four…or three… or maybe it’s two | DIGITS to DOLLARS·
Pingback: Did Qualcomm Just Buy a Server CPU Company? – No | DIGITS to DOLLARS·