# AI capex — X 热门讨论 (2026-10-08 20:04 UTC)
## @stevenfiorillo (Steven Fiorillo) · 10-08 19:17 · ♥122 ↻5 💬19 $ZETA just dropped the biggest set of announcements in company history at Zeta Live 26 which included AthenaOS, Athena MCP, A new Athena Inference Model and A specialized chip.
@amitisinvesting and I sat down with @dsteinberg10000 this morning for an episode of @basispointpod. The part investors should not miss is the chip.
A third party is building it, paying for it and manufacturing it. $ZETA is NOT spending a dime of CapEx to produce it.
That is capital light by design. It does not hit margins. It does not hit free cash flow. It does not touch the buyback. Margins and Profitability stays intact.
As an investor I love that. All the upside. None of the balance sheet drag as it deepens the flywheel into the $NVDA Nemotron models Athena runs on.
Here is the stack that landed on stage.
AthenaOS is the enterprise intelligence operating system. It knows, Decides, Acts, Learns.
You start with the outcome you want. Athena brings the intelligence, apps, and actions to get there.
Your data stays yours.
Athena MCP puts that intelligence inside tools teams already use. Day one integrations with ChatGPT, Claude, Palantir, and Gemini.
The Athena Inference Model is being built with @FireworksAI_HQ on @NVIDIA Nemotron. Workloads that run on Nemotron keep more enterprise inference inside Nvidia’s ecosystem.
Foundry handles the data and operations layer with $PLTR. Proprietary customer data sits at the core.
This also ties together the Senso and Digital Audience deals. Three major AI moves in about a week. Experience layer. Application layer. Orchestration layer. AI infrastructure. Context and data layer.
That is the AthenaOS stack in plain English.
@dsteinberg10000 is crushing it. As an investor I want the intelligence layer and the capital light chip. Not a CapEx story dressed up as AI. $ZETA $NVDA $PLTR https://x.com/stevenfiorillo/status/2108275667209015684
## @Banana3Stocks (Banana3) · 10-08 19:02 · ♥121 ↻4 💬16 $SPY $QQQ $MRVL $NVDA #OpenAI #Anthropic
Today’s selloff is the market taking one messy revenue headline and treating it like the whole AI trade just cracked. It didn’t. The indexes heard “$20 billion light” and sprinted for the exits like someone yelled fire in a data center 🔥 🕋
OpenAI being light by about $20 billion on annualized revenue is a definition problem, not a demand problem!!!‼️
The $70 billion number that was floating around was investors trying to line them up with Anthropic, and those two don’t even count the same way 🤦♂️
Most don’t know this unless you know AWS like the back of your hand but Anthropic includes cloud partner sales, OpenAI doesn’t 🧐
What they actually told investors is closer to $50 billion annualized, with their book still ripping! That’s about 77% total run rate growth in the quarter and enterprise run rate up about 107%. Consumer revenue in Q3 alone was already bigger than all of 2025. That’s not a company falling apart. That’s Wall Street doing math on a napkin and then acting shocked when the napkin was wrong and then dumb ass algos running wild🧮 🤦♂️🤦♂️
As I have said, this is not a one size fits all read on the buildout!!!
$MRVL just had its investor day and basically told the Street it was sitting in the cheap seats ⚾️. They took FY28 revenue to $20 billion, up from the $18 billion they had out there in August, and above the roughly $18.2 billion most analysts were sitting on. Then they dropped a FY31 framework of $70 to $90 billion in revenue and $30 plus in EPS 🤯 . Custom revenue up 200% or more into FY29 🤯. Data center alone approaching about $18 billion in FY28 🤯
Matt Murphy (MRVL CEO) went on and called the guide “pragmatic, judged and achievable.” Morgan Stanley said the targets were “at least 20% higher than we had forecasted” and that management “guided up next year by about 10%.” Needham took the target to $400 from $300👀. Jefferies went to $450 from $325👀. That’s the inside plumbing of this trade talking, do you hear it??? 👂 That means Interconnect, and optics are going gangbusters 💪
Meanwhile… the indexes are panic selling because one lab’s spreadsheet had a footnote. Classic BS 🤣
We’re still early in the capex boom. And that spend is already starting to show a return because demand for compute is through the roof!!!
These guys aren’t building empty buildings and hoping a customer wanders in🤦♂️. Capacity is the constraint. Watts, racks, optics, custom silicon. Customers are literally lined up with backlogs growing and growing 🌳 🌳
Enterprise adoption is early innings ⚾️ ⚾️. Most businesses are still poking at pilots like it’s a group project they can turn in late, they’re just figuring it out. The ones that don’t actually adopt this get left behind on cost and speed 🏃 🏃
Consumer agents? We’re not even in the first inning ⚾️!!!
That stuff has barely been out for a few months. Judging a multi year infrastructure cycle off one private lab’s run rate definition is how you miss the move and then tweet about it later 📈
Not financial advice!
Much love! 💜🍌🍌🍌 https://x.com/Banana3Stocks/status/2108271717374443631
## @trueinvestings (True) · 10-08 19:06 · ♥54 ↻4 💬3 Portfolio took a pretty massive hit off that OpenAI news of unrealized revenue miss of $20B. Every levered AI name took a hit and sold off massively.
If we look at the overall trend OpenAI numbers are still good we were at $25B in February and $50B now. Thats a 2x increase in growth so far. It targeted overall 29B for all of 2026.
Missed against a leaked unaudited number. Whether gross or net or just leak ahead of raise, encourages the demand for AI is still there.
The infrastructure of these names don’t change at all. Memory, networking, compute sell into the total AI capex. OpenAI run rate shouldn’t change anything in my opinion.
Something that would make me think we are in near the top would be a change in hyper scales cutting capex guides.
I have been accumulating some dips on names I have and new names I’ve added. https://x.com/trueinvestings/status/2108272889984082178
## @evergreencap3 (Evergreen Capital) · 10-08 17:34 · ♥30 ↻2 💬4 The market is overestimating how much AI capex depends on OpenAI and Anthropic’s revenues. Capex is a function of the supply and demand for tokens. And while I remain bullish on the labs, that imbalance will intensify regardless of the dynamics between these two companies. https://x.com/evergreencap3/status/2108249667834179803