Amazing quarter by $SOFI! Even executing in a tough macro environment they are growing in excess of 40% YoY and we will have $HOOD earnings today with arond ~15% YoY growth and similar revenue but almost 5 times the market cap of $SOFI. Something has to give!

$ZETA Big earnings coming up. Cant wait! 🚀🚀

$NBIS Got some shares after a long time Hopefully closer to the bottom

$IREN Anthropic releases S1 today at 9 AM EST. Wonder if we see any name of $IREN in there. This is what is known so far: - Broadcom: $160+ billion to lease custom Google TPU chips through a direct partnership. - Google Alphabet: $111 billion in legacy infrastructure commitments. - Amazon Web Services (AWS): $110 billion in cloud computing commitments. - Lambda Labs & Nscale: $86 billion combined to rent Nvidia chips through 2033–2034, which includes a specific $45 billion deal with UK-based neocloud Nscale. - SpaceX / Colossus 1: A contract for the entire Colossus 1 cluster, securing access to over 220,000 Nvidia GPUs. - Microsoft: $31 billion in cloud infrastructure obligations.

You'd think with all the $SPCX xAI compute power currently draining the electric grid Grok would at least recognize its own reflection. Maybe Elon should focus less on tweeting doom and gloom predictions about the collapse of AI neoclouds and actually focus on training his own model. Bullish $NBIS $IREN $CRWV grok.com/share/c2hhcmQtMw_90…

Doesn't apply to superior Neocloud*cough Hyperscalers like $NBIS @SpaceX is already using @tavilyai. We should see a deeper partnership soon

$NBIS About time

$IREN CEO says the prices aren't going down but he didn't confirm if they're going up. One of the reasons where $IREN might be compromising in return for moderate pricing is how much prepayments can they get to fund the build out. More the better That's their main focus it seems as it should be

S&P is still close to ATH's what happens to financials and data center stocks if the S&P corrects itself by 5-10% from here ❓

Following the playbook of $ZETA the ultimate differentiator for $NBIS will be its ability to demonstrate a superior Return on Compute Spend compared to its industry peers.  Currently, $ZETA maintains its market premium by consistently reporting exceptional economic efficiency, boasting a Return on Marketing Spend above 600% and a Return on Technology Spend exceeding 295%. As the AI infrastructure market matures and raw compute supply inevitably catches up with demand, hardware providers can no longer rely solely on raw processing power or AI workload capacity to justify their valuations.  If $NBIS can successfully differentiate itself by proving exactly how much tangible value and financial return its customers achieve per dollar of compute spend ( through platforms like Token factory, recent acqusitions that bolster the performance ), it will provide a powerful catalyst for long term market leadership imo

It is inevitable that every major global enterprise will eventually become a $NBIS customer. By 2030, if a company is not utilizing NBIS for its AI workloads, it likely will not be able to generate competitive returns.

$NBIS This boring routine will help us get $200B financing to build out 1GW of compute per year 🥱

$SPCX WE ARE SECURING DEALS WELL ABOVE THE $30M-$50M/MW RANGE, ENTIRELY DRIVEN BY DEMAND This is crazy!!! Remind me again why the data center stocks dipped so much? $NBIS $IREN

$SPCX Elon: "I would be very surprised if we are not able to bring up to 13GW of compute by end of 2027" Ruthless execution. Separates them from average Neoclouds Only $NBIS next in line when it comes to execution

$CRWV Nothing is helping Debtweave today

It is honestly comical how international diplomacy works these days. We have the US acting as a mediator between Russia and Ukraine, while simultaneously, Russia is attempting diplomatic mediation between the US and Iran. The funny part is that both of these countries are responsible for conflicts they never should have started in the first place. Global politics has turned into a total clown show run by a select group of government officials with way too much power.

This was sick af This is literally all of us tech nerds in the audeience when $NBIS finally hits 2000

$NBIS Expect to hear things which will re rate the data center trade again

Anyone else tracking the 2027 compute math after the CoreWeave bond drama this week? - $CRWV locked down a monstrous $100B+ backlog, but they did it early. Their older contracts yield around ~$10M/MW. Meanwhile, hardware costs have spiked Blackwell racks and Vera Rubin chips mean total build costs are hitting $35M to $47M per MW because memory got so expensive. Needing 4 years to earn back a build at $10M/MW while unsecured debt yields 16% is why the market is sweating (even if demand is fine). - $NBIS learned the lesson. They have a ~$40B backlog and Meta's massive deal coming online, but management is intentionally holding back capacity instead of selling out 2027. They're charging double per MW to offset inflation. - $IREN sold out 2026, but their 2027 is mostly a blank canvas. Managemengt did they they are in late stage discussion for some 2027 capacity and hopefully we hear more about them on the next ER It boils down to a a tricky trade off: Lock in early for cheap debt and get stuck with cheap pricing, or wait for high spot prices and get squeezed by modern credit markets. A month ago I thought waiting was a no brainer. Now not so sure.

I have no doubt we’re going to see a huge shift in adoption to open weight models. My base case: open models go from ~18% of all tokens today to ~44% by 2030. Total tokens go from ~120Q a year to ~1,400Q. Most of that growth is agents, not people chatting. Agents already use more tokens than humans on OpenRouter, and each agent request burns ~15x more. When a workflow fires off thousands of calls, you don’t send the routine steps to a $30/M token model. You send them to an open model at a dollar or less. Just this week $NBIS put three new open weight models live on Token Factory on day one: $Meta’s Muse Spark Open, $NVDA's Nemotron 3.5 Super VL and Reflection’s Beam. Beam is $1 per million output tokens, Nemotron $0.90. That matters because a lot of enterprises weren’t comfortable running DeepSeek or Qwen. Now they don’t have to. Even at 44% of tokens, open models are probably under 10% of spend. Closed labs keep the money, open models take the volume. Here’s why that matters for neoclouds. Only OpenAI or Anthropic can serve their own models, so a neocloud’s job there is renting them GPUs. Open models anyone can serve. That’s where $NBIS gets to sell the whole stack, not just the hardware. Does that mean better margins? Maybe. Token Factory workloads tripled last quarter, and they’ve bought three inference optimization companies (Eigen AI, Clarifai, Inferize) but open inference is a commodity. Everyone serves the same weights, so margin comes from running GPUs more efficiently, not charging more. Upside goes to whoever’s best at that.