# chip earnings — X 热门讨论 (2026-10-11 05:16 UTC)

## @HunterAllen4 (THE GAP FATHER) · 10-10 13:25 · ♥31 ↻4 💬15 $EME

$1,000 is coming. 👀🔥

EMCOR is one of the most important infrastructure plays hiding in plain sight. AI data centers don’t build themselves. They require massive electrical systems, power distribution, cooling, mechanical infrastructure, and specialized contractors capable of bringing these facilities online.

$EME sits directly in that spending pipeline, with exposure to data centers, semiconductor facilities, healthcare, and other mission-critical infrastructure.

The growth numbers are getting serious. Q2 2026 revenue hit a RECORD $5.15B, up 19.8% year over year, while diluted EPS surged 34.8% to $9.06.

This isn’t just a company benefiting from a hot narrative revenue is growing, earnings are expanding, and demand is translating into real financial results.

Even more interesting? Remaining performance obligations climbed to a record $17.14B, up 43.9% year over year, with approximately 95% of that growth coming organically.

That’s a substantial pipeline of contracted work that gives investors a clearer picture of future demand, although backlog still has to convert into revenue and profits.

The AI infrastructure tailwind is showing up in the business itself. Electrical construction revenue grew 24%, while data-center revenue within that market surged 45%.

As hyperscalers continue investing in computing capacity, the electrical and mechanical systems required to power these facilities become increasingly critical. EME is positioned to benefit from that buildout without having to bet on which AI chip ultimately wins.

Management also raised its full-year 2026 guidance to $20B–$20.5B in revenue and $32–$33.25 in diluted EPS. That’s the kind of combination I want to see: accelerating demand, a growing backlog, stronger earnings, and higher expectations from management. The big question is how much of that growth is already reflected in the stock’s valuation.

$1,000 is the target I’m watching. Nothing moves in a straight line, and even great businesses can experience brutal pullbacks when expectations get ahead of fundamentals.

But EME has the earnings power, infrastructure exposure, and backlog growth to keep this name firmly on my radar.

The AI buildout is bigger than just chips and the companies building the physical infrastructure could be some of the biggest beneficiaries. 👀💪 https://x.com/HunterAllen4/status/2108911678037393563

## @SemiconductorsX (Semiconductor Insider) · 10-11 02:55 · ♥34 ↻2 💬4 Robot DRAM demand is projected to rise more than 20 times by 2030. The per-unit content only doubles. The rest is the number of robots.

As per Counterpoint: total humanoid DRAM demand goes from about 8,300 terabits this year to about 172,000 terabits in 2030. Average DRAM per robot rises from 19GB to 39GB.

A humanoid reads cameras and sensors in real time and decides the next move. As the AI models get heavier, the memory needed to process that data rises with them. Battery power matters, so the industry expects the pull to be low-power DRAM and high-capacity NAND, not a server RDIMM.

Micron CEO Sanjay Mehrotra said on the September 30 earnings call that a Level 4 or higher autonomous car typically carries more than 200GB of memory and several terabytes of storage, and that a humanoid will need a similar load. Counterpoint also sees compute-chip demand rising with the models, including NVIDIA GPUs.

HBM is the data-center bid. This is a new end market for the same three suppliers, on a different product.

172,000 terabits in 2030 is a large multiple off a small base. It is not yet a second HBM.

If one robot takes 39GB and the total is 20 times today’s bits, how many humanoids does the model need to ship before it moves the DRAM price? $DRAM https://x.com/SemiconductorsX/status/2109115711625064818