# data center revenue — X 热门讨论 (2026-09-30 18:04 UTC)
## @CKCapitalxx (CK Capital) · 09-30 16:30 · ♥72 ↻10 💬21 $MOD is about to become a different company on Thursday, and I think it's one of the most interesting setups in data center cooling right now.
Modine is spinning off its auto and industrial cooling business into Gentherm. That's $1.1B of revenue, shrinking 3%, at a 13% EBITDA margin. Shareholders get about 0.4 Gentherm shares per Modine share plus $1.90 in cash. Modine gets $210M to pay down debt.
What's left is two segments:
Data Centers: $349M last quarter, up 90%. Guided to 60 to 80% growth this year. Margins are 15% right now because they can't get fans and coils fast enough, guided back to 19 to 20% next quarter and above 20% in the second half.
Commercial HVAC: $262M, up 22%, guided to 18 to 20% margins by year end.
That's the whole company after Thursday. No auto exposure, net debt around $200M, and two segments growing 90% and 22% instead of being averaged down by one shrinking 3%.
Hunterbrook reported a $4B+ Google chiller contract running 2027 to 2029 and a large Amazon air handler pipeline from an internal database. A new 3MW chiller starts production in October with first delivery in January. The constraint is parts and engineering hours, same as $AAON, same as $VRT.
CEO Neil Brinker took this from $10 to $180 by cutting what didn't work. Thursday he cuts the last piece. Q2 report in late October is the first look at the clean company. Data center sales near $450M and margins near 20% and this rerates. https://x.com/CKCapitalxx/status/2105334408634937629
## @Forbes (Forbes) · 09-30 14:31 · ♥34 ↻9 💬7 Bob Clark has been building data centers since the 1980s and completed 16 of them last year. He has another 77 in the works across 13 states. Data center construction is expected to account for 75% of Clayco’s nearly $12 billion in 2026 revenue, helping boost revenue by 43% over 2025.
“I’ve been in the data center business for most of my career, but never would’ve imagined what’s happening today,” he said. Read more: https://t.co/Cb90magyht 📸: Aaron Kotowski for Forbes https://x.com/Forbes/status/2105304583161757841
## @HunterAllen4 (THE GAP FATHER) · 09-30 12:38 · ♥32 ↻1 💬15 $HPE
EARLY BONER CANDLE. 🍆🔥
I told you all ATHs were coming.
HPE just landed its first $1.2B Vultr order for AMD Helios AI racks, combining AMD Instinct MI455X GPUs + EPYC Venice CPUs with AMD networking/ROCm and HPE’s rack-scale systems, Juniper networking, cooling and integration.
$AMD supplies the compute → HPE builds the AI infrastructure → Juniper provides the networking → Vultr deploys it.
But the networking story may be even bigger. At today’s Networking Investor Day, HPE raised its FY27 Networking outlook to high-teens to low-20s growth, expects share gains across Data Center Networking, Routing and Campus & Branch, and sees Routing growing at a low-to-high 20s CAGR through FY29.
Add the multi-year Wi-Fi 7 refresh cycle, deeper security integration and $800M in annual Juniper synergies by FY28, up from $600M.
The underlying numbers are already strong: Q3 revenue hit a record $12.2B, up 34% YoY, networking revenue jumped 75%, data-center networking surged 112%, and AI systems backlog sits at $6.8B.
HPE’s current FY27 revenue guide is 14–17%.
FY27 growth reaching 25–27% and potentially approaching 30% if Helios and the broader AI infrastructure ramp accelerate. That’s my thesis, not HPE guidance.
I see $HPE becoming a much broader AI infrastructure platform spanning compute, networking, cooling, security and full rack-scale deployment.
Early candle. Let it cook.🔥 https://x.com/HunterAllen4/status/2105276094630826452
## @Dagnum_PI (Dagnum²) · 09-30 16:57 · ♥34 ↻6 💬2 $1.35 billion was stolen from DeFi in the last year, according to DefiLlama
That number is a big part of why Constellation closed PacaSwap and went all in on data integrity and AI
AI lets anyone point a swarm of agents at open code and find a bug in hours
PacaSwap was exploited twice in a few weeks
A small team can't win that race with other people's money on the line
The node cut follows the same logic
~400 to 500 nodes across mainnet and two test networks were supporting a handful of active metagraphs....all paid from emission
Trimming mainnet to 30 validators and ending test network rewards is what lets inflation fall from about 6% toward a target of 2%
Running a node used to take 250,000 DAG, a server, a spot on the allowlist and months on a waitlist
Ao only a few hundred people ever had one
The inference network opens rewards to everyone else
If you have a Mac/PC with 32 GB of memory or a GPU with 16 to 24 GB of VRAM that machine can serve AI models to Gate customers overnight while you sleep and you earn $DAG for the work
There's no allowlist and no cap on who can join
If you'd rather not run hardware delegated staking stays
You can still delegate DAG to the validators and share in the rewards
For holders asking why DAG still matters
Every request a Gate customer sends to the inference network is paid out to operators in DAG
That gives DAG a job tied to real revenue while inflation heads down
I came to Constellation in 2021 because someone told me that if I liked Palantir then I should look at a network built to verify data at the source
This is that network with DAG at the center
Meanwhile the data integrity work keeps moving
Last week Lithuania's Ministry of Economy and Innovation named Constellation
"a developer of secure data transmission and data integrity solutions,"
Yes Constellation's work with Panasonic is still in progress
$DAG $AIAI > 引用 @Conste11ation: Constellation is going all in on AI.
Ben, Alex, and João discuss the changes, why it had to happen, and what it means for the future of the network.
Full details in the blog: https://t.co/cY5AVKsayx https://t.co/wxWMFbgKwq https://x.com/Dagnum_PI/status/2105341215424319603