# data center revenue — X 热门讨论 (2026-09-26 16:50 UTC)
## @Funmentalist (The Fundamentalist) · 09-26 14:18 · ♥60 ↻3 💬7 Nebius Thesis (Updated - September 2026) https://x.com/Funmentalist/status/2103851611403465206
## @rohanpaul_ai (Rohan Paul) · 09-26 06:41 · ♥31 ↻5 💬9 How much revenue each unit of deployed compute would need to generate for 182.7 GW of additional U.S. data-center capacity between 2025 and 2032
This Columbia Business School report "Financing the AI Buildout" ans that.
182.7 GW is its central buildout scenario which is the U.S. data-center capacity between 2025 and 2032.
Using the GB300 NVL72 configuration, that capacity works out to roughly 1.07 million racks / 77 million GPUs, which lets the author express the required $3.725 trillion of mature annual revenue as about $5.5 per installed GPU-hour at full utilization.
This required unit revenue sits within the $6–$10+ per GPU-hour for current high-end NVIDIA capacity, so the central scenario does not require some radically higher compute price to make the arithmetic work. https://x.com/rohanpaul_ai/status/2103736651663208491
## @HunterAllen4 (THE GAP FATHER) · 09-26 02:41 · ♥31 ↻1 💬7 $KEYS
Massive bull flag cannon? Yes.
Testing is becoming the next bottleneck, and KEYS is sitting directly in the middle of it.
Its niche is much broader: high-speed electrical and optical characterization, network test, protocol emulation, RF/mmWave, semiconductor validation and software.
As AI infrastructure moves from 800G → 1.6T → 3.2T, silicon photonics scales, and co-packaged optics becomes more important, engineers need to validate increasingly insane amounts of bandwidth, signal integrity and system performance before these technologies can scale.
That is where KEYS gets interesting. The company is involved across the workflow pre-silicon validation, component characterization, rack and cluster emulation, high-speed Ethernet testing, optical validation and eventually manufacturing.
More complexity means more things that can fail, which means more testing. You don’t just build a 1.6T or 3.2T system and send it into a hyperscale data center. You have to prove the damn thing works.
And we’re already seeing the demand. Q3 FY26 revenue was $1.85B, up 36%, while orders hit $2.09B, up 56%. CSG revenue jumped 43%, commercial communications crossed $1B for the quarter and wireline orders more than doubled.
Management has said demand is exceeding available supply, with roughly $100M of revenue sitting beyond the company’s normal order-to-revenue window. That’s an unusual setup: the bottleneck is KEYS’ ability to fulfill demand, not a lack of demand.
The financial profile matters too. Gross margin was roughly 69% and operating margin 33.2% in Q3, while Q4 guidance calls for $1.93–1.95B of revenue and $3.34–$3.40 of non-GAAP EPS.
If KEYS can continue capturing the shift from pure R&D testing into higher-volume manufacturing workflows while maintaining those margins, the opportunity gets even more interesting.
RSI around 66, MACD turning higher and the longer-term moving averages still stacked underneath price. ~$367–375 is the area I’m watching for the next breakout.
If it doesn’t go, $340–350 is the nearer technical zone, while that massive $240–260 gap remains much further below.
AI infrastructure needs more than chips, power and data centers. It needs a way to validate all of that increasingly complex hardware.
That is the $KEYS niche.
And I think testing could be one of the next major bottlenecks the market starts pricing in. https://x.com/HunterAllen4/status/2103676195078369479
## @ValueInvestShow (Value Investor) · 09-25 18:19 · ♥31 ↻1 💬4 $BABA is one of my worst stock picks ever, not because it's down significantly but rather because since I invested 4.5 years ago, the stock has done nothing
However here's 3 reasons why this could change: 1) Jack Ma bought over HK$600M of shares right after the HK$80B placement. Joe Tsai added ~HK$81M and CEO Eddie Wu picked up 350k shares. Insiders don't pile into a raise they think destroys value. 2) The spend is going somewhere specific: a 20GW data center network by 2032, plus new cloud regions in Turkey, Finland and the Netherlands in the next 12 months. Citi thinks that network can support $160B+ in external cloud revenue. Goldman also expressed optimism. 3) The lawsuit headlines this week are law firms chasing an Oct 5 deadline over June news (the DoD military-company listing and the distillation allegations). That was priced in months ago.
The risk almost nobody's tweeting is free cash flow: negative RMB 44.7B, with capex up 75% YoY. That's the number I'm watching next quarter. Hopefully this stock can turn around.. https://x.com/ValueInvestShow/status/2103549847215169570