Alpha～

While the market is still debating who'll become the next AI winner, NVIDIA CEO Jensen Huang has already given one of his own answers. He's called $Nebius(NBIS.US) and $Coreweave(CRWV.US) "World-Class AI Clouds." That's not just a passing compliment — it sits on top of a series of large-scale partnerships and capital commitments already underway. Start with $Nebius(NBIS.US). One of the most closely watched developments around Nebius right now is its $27 billion AI cloud partnership with Meta. At the same time, the company is rolling out a 5GW-scale collaboration with NVIDIA — and NVIDIA itself has invested $2 billion into Nebius. For the AI infrastructure space, customers, compute, and capital are often the three most critical variables, and Nebius has secured backing on all three at once. Now $Coreweave(CRWV.US). CoreWeave has become one of the most talked-about names in AI cloud over the past few years. The company struck a multi-billion-dollar cloud services partnership with Anthropic, and signed a $21 billion cloud services agreement with Meta. Much like with Nebius, NVIDIA has likewise invested $2 billion into CoreWeave. Looking at the partner lists — Meta, Anthropic, NVIDIA — each is one of the most important players in today's AI value chain. That also means the links between AI model companies, AI platform companies, and AI infrastructure providers are getting tighter and tighter. Worth noting: over the past few years, the market focused more on the model layer and the application layer. But as training scale keeps expanding and inference demand grows rapidly, the importance of compute, data centers, power supply, and cloud infrastructure keeps rising. And what NBIS and CRWV have in common is that they're both deeply involved in this round of AI infrastructure expansion. When NVIDIA pours billions of dollars of capital into both companies and keeps pushing the partnerships forward, the market naturally starts to re-assess where these AI cloud platforms sit in the value chain ahead. AI competition may not just be a contest between models. What more and more capital is starting to watch is the infrastructure network running behind those models. And in that space, NBIS and CRWV have already become names too big to ignore.

1/Jensen confirmed it. Vera CPUs are now at Anthropic, OpenAI, SpaceX, and Oracle. 🤖This is Nvidia's second data center CPU. First was Grace (2021). Vera is purpose-built for one thing: agentic AI.2/Grace = general HPC. Vera = orchestration layer for AI agents.When Claude or GPT-4o runs a multi-step autonomous task, the GPU handles inference. The CPU handles the decision logic, tool calls, memory retrieval, loop control. That's Vera's job. 💻3/Why those four customers first?Anthropic + OpenAI: production agentic systems already running at scale.SpaceX: Starlink's autonomous ground network needs exactly this kind of CPU.Oracle: cloud infrastructure partner, early deployment makes sense.4/Jensen also said China will "gradually open." No timeline.But Vera going to the top four AI labs while China sits out the H200 trade tells you where Nvidia's near-term growth is coming from. It's not China. It's every other AI lab on the planet. 🇺🇸Watch Nvidia's Q3 earnings. Vera attach rate in data center revenue is the new number to track.

Intel Jumped 4.43% on Its 18A Reveal. Is the Comeback Real, or Just a Good Day?

NVIDIA up 6% and ARM up 16% in the same session. That 10-point gap between the two numbers is telling you something. ARM licenses its CPU architecture to every company building ARM-based chips: Apple, Qualcomm, NVIDIA. Every RTX Spark laptop that ships this holiday season, every Vera Rubin data centre rack that goes live, every new ARM-based device from any manufacturer anywhere — ARM gets a royalty. NVIDIA using ARM for RTX Spark is a direct ARM revenue event. 📈Here is the RTX Spark picture: 20-core ARM CPU, 6,144 CUDA cores at the level of a desktop RTX 5070, Blackwell GPU architecture, 3nm process, 45–80W power envelope. Dell, Lenovo, Asus, MSI, and Microsoft Surface are all launching devices before the 2026 holiday season. This is NVIDIA's first laptop chip. The PC market was the one segment they did not own. Now they are attacking it directly, and Qualcomm and Intel are both in the crossfire at the same time.The market consensus going into Computex was that NVIDIA was a data centre company. The consensus coming out is that NVIDIA is an AI compute company across every form factor: data centre, workstation, laptop, and eventually edge. That TAM re-rating is exactly what the 6% move reflects. ARM's 16% reflects the same insight but one layer up the stack.

NVIDIA × Microsoft × Arm Speak Out in Unison: The PC Industry Is About to Change!

Micron earnings Thursday. Memory stocks been on fire. Please keep running 🔥

Two new data centers announced, powered by renewables, and Wells Fargo just upgraded with a $387 target. Google is quietly building the AI infrastructure moat while everyone stares at the GPU shortage narrative. $Alphabet(GOOGL.US) $Alphabet - C(GOOG.US) 

Micron's HBM4 win with Nvidia is the real story. Memory cycle now driven by AI, not consumer — multiples can expand

🚨 This is big, and a lot of people are reading it as just another green day. It is not. Nvidia (NVDA) ripped more than 6% and Arm (ARM) ran almost 16% in a single session. Here is what actually happened: Nvidia is pushing deeper into PCs with a new lineup of AI chips, and the new superchip is aimed straight at Intel and AMD. Look, that is Nvidia walking into the CPU conversation. For years the story was GPUs for training. Now they want a piece of the chip inside your laptop too. Why did Arm fly 16%? Because that PC push runs on Arm-based designs. When Nvidia expands the tent, Arm collects rent on the architecture. That is the trade the market made today. My take: I think this is the market repricing Nvidia from "the AI training company" into "the company that wants every compute socket." That is a far bigger story. One honest warning: a 16% day on ARM is enormous. Moves like this round-trip fast if the product timeline slips or the launch underdelivers. I am bullish on the direction, not on chasing a 16% candle. 📈

$Circle(CRCL.US) down again today. Value trap or buying opportunity?🤔

INTC just went nuts this week 🚀 Joined Musk's Terafab thing with Tesla/xAI — stock shot up like 15%, hit almost $59. Some call it hype, some call it a turnaround 🤷‍♂️ 18A foundry finally got a real customer. Also bought back full control of Fab 34 from Apollo for $14B. Advanced packaging deals with Google/Amazon rumored. But uh… foundry still lost $2.5B last quarter. Yields apparently only ~60% and needs 80%+ for Tesla orders. Earnings on Apr 23 expected to be negative. Analysts can't decide: KeyBanc says $70, HSBC says Sell. Classic. Musk narrative doing all the heavy lifting right now. Anyone actually buying this or just watching the show?🤔

Marvell Technology (MRVL) reports after the close today. The stock is sitting at USD 196.33, having nearly doubled in 2026. Options-implied volatility is elevated, as expected pre-earnings. Consensus is modeling USD 0.80 EPS and USD 2.40 billion in revenue for Q1 FY2027. For context, full fiscal 2026 revenue was USD 8.195 billion, up 42% year-over-year. The market has a lot priced in. The setup is straightforward: if custom silicon (ASIC) numbers come in above USD 600 million and the data centre networking segment confirms sequential acceleration, gamma exposure flips and dealers begin buying the underlying to hedge their short-call books. That mechanical dynamic can push MRVL past USD 205 in the session following results. A guidance miss or any softness on AI ASIC delivery timelines brings it back to USD 175 quickly. High variance, high vol, watch the options flow in the hour before close.On TSMC: a 15% price increase on 3nm wafers in H2 2026, with another 10% pencilled in for 2027, is not complicated to interpret. Demand exceeds supply and TSMC knows it. Current 3nm wafer costs sit around USD 20,000 each. The customers absorbing this — Apple, NVIDIA, AMD, Marvell — are paying because there is no alternative at this node geometry. That is pricing power at its most fundamental. Every MRVL custom ASIC chip is fabbed at TSMC. If management comments on wafer cost pass-through in tonight's call, that is worth paying close attention to for margin assumptions going into FY2027.Those views can change at a moment's notice if the results or guidance shift the picture materially.

COIN and CRCL still haven’t caught up with Bitcoin — they’ve got some catching up to do.