# data center revenue — X 热门讨论 (2026-09-27 20:12 UTC)
## @lugaricano (Luis Garicano 🇪🇺🇺🇦) · 09-27 18:56 · ♥41 ↻11 💬1 "in order for all data center owners to break even on their AI investment... their annual revenue would amount to roughly 9.2% of projected 2032 GDP" I wanted to make time to write about @SVNieuwerburgh 's excellent paper in our Brookings panel but @HannoLustig beat me to it. Do read the full post (or Stijn's paper!). https://t.co/xapKMPbQoA https://x.com/lugaricano/status/2104284046012350699
## @AuditTheHerd (Audit The Herd) · 09-27 13:27 · ♥36 ↻1 💬3 $SITM reminds me a lot of $AMD 10 years ago. I had around 60% of my portfolio in at $15-20 and sold way too early. I am not making the same mistake now.
If you have spent any time trying to understand what actually makes modern electronics work esp with the ai build out, you eventually land on timing. Not something you would even consider tho right? Every system needs it, but no one talks about it, which is exactly why SiTime has stayed under the radar.
Let’s think about what SiTime replaced. For decades timing came from quartz (yes the actual rock). Quartz is analog, fragile, it drifts when it gets hot, it struggles with vibration and it does not scale well. SiTime built timing in silicon using MEMS. It is programmable, so a customer can set the exact frequency and stability in software, it is resilient in harsh environments and it is manufactured like a semiconductor. It sounds like a small difference but it’s actually not. With MEMS, engineers are able to develop an actual system.
I hate the term MOAT, so I will use the term value proposition. First there is the intellectual property. More than a hundred patents covering how you design, package and compensate a MEMS resonator, and that took over ten years to build. Second there is the library. SiTime now has thousands of programmable parts that drop in where quartz used to be, so once you are in the design system you become the default. Third there is the pain of switching. If you are building an AI server, a networking switch, an aerospace box or an automotive compute platform, you do not want to requalify timing. If timing fails the whole system fails, so you stay with what works. That makes wins very sticky. And fourth there is how it is built, through partners like TSMC and Bosch, which gives it semiconductor economics rather than crystal economics.
The TAM story started a bit small but it is growing exponentially every day. It is in almost every sector, space, data centers, robotics, autonomous cars, ect. Quartz was fine when electronics lived in benign environments and data rates were modest. The next wave is the opposite. AI clusters need to stay synchronized across thousands of nodes. Data center and optical gear need lower “jitter” at faster speeds. Cars that drive themselves and aerospace systems need timing that survives heat and shock where quartz just quits. Edge devices need tiny, low power timing. Every one of those trends means more SiTime content per box and more boxes overall. That is why the market for precision timing keeps growing at a high single digit to low double digit pace and MEMS is still early in terms of penetration.
So why do I think it is still cheap to own on a forward basis? It is invisible, it is small cap and it had a rough couple of years. Because the product is deep in the supply chain it never makes it into the popular AI narrative, even though AI cannot run without it. Coverage is light and most screens still bucket it as a boring component. Then you had the inventory correction that hit all small cap semis. If you look at it on a TTM basis it screens poorly. Profitability looks compressed and price to sales looks high because you are looking at the wrong piece of history.
Flip it to forward estimates and it looks different. Revenue has turned, mix is moving to higher value precision parts and gross margins are back in that high fifties to low sixties range that you expect from an asset light design driven business. At around 40 times forward earnings estimates you are not paying a crazy price. You are paying a decent price for a sole source supplier that owns a foundational layer with a real moat in a TAM that is getting bigger every year. That disconnect between how it looks on trailing numbers and how it looks forward is why it feels like a hidden gem right now.
If interested my write up is in the post below https://x.com/AuditTheHerd/status/2104201194805596331