I am working in software engineering but not as familiar with the hardware side of vendor lock in. There have been some extreme cases of cloud providers offering huge discounts to gain market share. Around 2017 Microsoft was offering any startup up to 1M in cloud Azure credits, or basically a startup could run on Azure for free until you spent 1M. This created a tough decision for a lot of startups who would much prefer to be on AWS, but were tempted by Microsoft’s offer. There was also the case during covid where Oracle offered two years free to Zoom for using Oracle Cloud, because they saw the value of having Zoom locked in.
A lot of the companies looking to acquire AI servers right now are planning a multi-year build out. They are not just buying a round of server racks and then taking a break on purchases. The CapEx is planned out for many years. The companies cannot afford to skip a generation of hardware because of how big the speeds up are. For example, Blackwell is supposed to be 30x faster and 25x more energy efficient than Hopper. I’m guessing there is a certain inertia towards working with a vendor repeatedly which is what Supermicro means by marketshare.
There are some cases like Tesla buying 50/50 on Supermicro and Dell servers. What I am not sure about this situation is whether Supermicro couldn’t meet the demand Tesla had and the overflow orders went to Dell (or visa versa). Or it could be that Tesla is hedging their bets and seeing which vendor provides a better product and service. It will be telling if Telsa does another round of server purchases and who they decide to go with.
Another factor may be Supermicro’s service contracts and I’d be interested to find out what the margins are like on this part of the business and what percentage of their revenue comes from service. Here’s a job posting they have in for a service technician. The job description is,
Supermicro is seeking a Sr. Service Technician (ST) that will be responsible for field work in local area. The ST will work exclusively at customer site to resolve and repair server systems. The ST are a critical part of post-sales support and need to be able to demonstrate exceptional technical aptitude in working with server products in addition to excellent interpersonal skills.
One of the details of job is to “provide on-premise field repair at customer facility”. My understanding is this type of job role goes to where the customer’s servers are. I’d also be interested to find out what percentage of customer’s opt for getting their servers service done by Supermicro. It is my understanding the customer can self service their own servers, or pay Supermicro for the maintenance service.
There’s 401 listed open engineering roles on the Supermicro careers page. For comparison there’s 79 job openings in sales, and 82 job openings in product managers, 12 in finance, and 41 in product and warehousing. It just shows how much engineering work goes into making the product. The CEO Charles Liang said that it took thousands of engineers working “around the clock” to build their liquid cooling solution over a two year time frame. I see Supermicro as an innovator on their technology, and I don’t believe this is a commodity business as some of the bear cases against this company assume.