Audit The Herd
CPA, Corporate Finance Professional
SiTime's quartz-to-MEMS shift extends beyond AI servers, supporting growth if AI training spend slows. Q2 CED rose 181% to $101M, AIED (auto, industrial, aerospace and defense) 51% to $25M, and MIC (mobile/consumer, mostly Apple) 85% to $31M; CED is still largest and fastest. About a quarter of acquired Renesas timing revenue is industrial and auto. Design-in timing content tends to stick even if server builds pause. Space and robots have separate potential demand. x.com x.com x.comÂ
SiTime's MEMS can win high-performance timing sockets from quartz as management's precision-timing market doubles to $22-25B by 2030. At 25% share, $5.9B revenue implies about $37B in value, roughly double the stock; even 25% feels low. The base expectation is between 35% and 50% share, versus roughly flat at 15%. The Renesas clocking deal broadens the path, though execution remains a risk. SiTime is the best company in the market at current prices. x.comÂ
SiTime's timing content multiplies as AI data-center bandwidth scales, driving continued hypergrowth. Moving from 800G to 1.6T optical modules splits links into more, faster lanes with tighter jitter budgets and extra clock domains, so each module needs more and better timing parts, lifting volumes and prices. High-radix switches multiply timing content the same way. Its discrete parts sit apart from GPU and switch ASICs, so demand rises in several places in NVIDIA-based racks at once. x.comÂ
SiTime is an under-the-radar AI infrastructure winner posting triple-digit growth while likely trading below 40x NTM EPS. Its silicon timing chips replace quartz across data centers, robots, drones, defense and robotaxis. Q3 guidance implies ~250% YoY revenue and ~315% EPS growth; the Renesas deal adds ~$85M, leaving ~30% organic QoQ growth. It sells more into each rack, expects 1.6T optical revenue to double in 2027 with higher share and ASPs, and customers order 12-18 months out. x.com x.com … Show more
SiTime looked set to disappoint on guidance after likely losing a hyperscaler, probably AWS, while priced for perfection near its $564 high, with $400 possible. The market assumed the Renesas jitter-cleaner assets would simply bolt on, but a 3-6 month integration during the 1.6T cycle risks losing orders to NDK quartz; long term it should fix the issue. On the call, sequential growth was down, but SiTime appears to have replaced Amazon with Meta and Nvidia, and Renesas revenue eases debt fears. … Show more
Slowing AI capital spending keeps the semiconductor growth scare in play. Meta's cash flow is negative and Lam Research's capex is slowing, further evidence as the sector faces a growth-to-value rotation and a Fed pause next week. x.comÂ
SiTime's opportunity is being missed because AI inference needs far more timing content than training, so even a pause in training doesn't really matter. Per its Q1 commentary, inference infrastructure on newer XPUs needs two to four times more timing content per system; in Q2 it said synchronization adds several hundred dollars of content per rack across accelerators, switches and other subsystems. Demand is also spreading to enterprise, captive, neocloud and sovereign data centers. x.comÂ
TSMC's capex plans suggest the AI chip trade is not over for semiconductors, Nvidia, AMD and Micron. CFO Wendell Huang cites robust multi-year demand, including agentic AI, behind $60-64 billion of 2026 capex; TSMC expects 2026-28 spending to be even more significantly higher than in the prior three years. Hyperscaler earnings should reinforce the higher AI capex trajectory. CoreWeave-driven fears feel overdone, though the neocloud model does not appeal long term. x.com x.comÂ
Robotics component suppliers get paid on every robot no matter which maker wins, as capital floods into the sector. Robotics VC hit a record of over $16 billion in a quarter, and Goldman sees the humanoid market growing from under $2 billion to $38 billion by 2035, with the inflection around 2027-28. Nvidia supplies the compute and training platforms, Ambarella edge AI chips, Ouster lidar, Vishay Precision force sensors, and memory makers the memory every robot needs. x.com x.comÂ
Aehr Test Systems' AI wafer-level burn-in business is scaling, with a $22M follow-on order from its lead AI processor customer to expand capacity for high-volume production. That customer is already planning more capacity, and Aehr is expanding talks with other hyperscalers. The stock reached fresh all-time highs, and its recent strength looks like institutional accumulation. x.com x.com x.comÂ
Semiconductor hardware faces pressure as AI shifts from racing on capex to competing on capital efficiency. Meta selling surplus compute amid slower-than-expected agent progress raises doubts about returns on AI spending, while Anthropic is reportedly exploring custom chips with Samsung. Nvidia moving to revenue sharing and Meta renting out capacity squeeze resellers like CoreWeave. Makers with capacity but no pricing power will be hit first if demand growth slows. AI demand has not collapsed. … Show more
Semiconductors will struggle to rise, with any rally sold, because Meta's move to rent out compute, exiting the arms race like xAI, has the market pricing a broad capex slowdown. Meta's spending is top tier, so the selloff prices in who is next; Microsoft may also pull back and drag semis down. Even in a benign reading, Meta is responding to market aversion to heavy capex, a sentiment top signal. If OpenAI and Anthropic ARR growth slows in a price war, QQQ could fall back to the 600s. x.com … Show more
