Alex

$NVDA's Q3 earnings prove that they are now UNSTOPPABLE: • Overall revenue: $18.1 Billion, up 206% y/y • Datacenter revenue: $14.5 Billion, up 279% y/y • Earnings per share: $4.02, up 593% y/y 🤯 • $NVDA's P/E now lower than $TSLA or $PLTR And their run has just begun:

$NVDA @NVIDIADC just shipped the biggest jump in AI yet. I sat down with @nvidia Vice President of DGX Systems and he had a few surprising things to say. Don't say I didn't warn you:

I WAS WRONG. This is a HISTORIC buying opportunity:

Great post. Minor flag: 12,000 GPUs in an NVL72 configuration = ~167 racks. Each Vera CPU rack delivers 22,528 cores / 45,056 threads. 400,000 cores ÷ 22,528 = ~18 Vera CPU racks alongside 167 GPU racks — roughly a 9:1 GPU rack to CPU rack ratio In chip terms: 400,000 CPU cores ÷ 88 cores per Vera CPU chip = ~4,545 Vera CPU chips alongside 12,000 GPUs — roughly a 2.6:1 GPU chip to CPU chip ratio That 33-to-1 ratio only looks so skewed toward CPUs because of mismatched units on Jensen's part.

Rene Haas just confirmed the Vera CPU thesis on yesterday’s Arm Q4 call. He didn’t mean to His framing: GPUs are reticle-limited. CPUs are not. The ratio shift is happening in core count, not chip count His exact words: “256 Vera CPU chips, 88 cores per chip, a 200-kilowatt liquid-cooled rack designed to sit in a data center adjacent to a Vera Rubin system” That is not a host CPU. That is a dedicated agentic orchestration Two days ago NVIDIA’s own engineers published the receipt. They traced a real 33-minute Claude Code session: 283 inference requests 58 main-agent turns coordinating 225 sub-agent invocations Context grew from 15K to 156K tokens before compaction dropped it to 20K Main agent alone processed ~3.5 million input tokens in the first 40 turns Anthropic’s own number: agentic systems consume up to 15x more tokens than chat. Coding agents sustain 95 to 98 percent prompt cache hit rates. Without caching, costs would be 6x higher This is what’s happening between GPU calls. File reads. Tool invocations. Sub-agent spawns. Compaction. KV cache management. None of it runs on the GPU That’s why 12,000 GPUs need 400,000 CPU cores. The 33-to-1 ratio isn’t a forecast. It’s a measurement NVIDIA states it in the blog directly: this won’t be resolved by adding more compute FLOPs and memory capacity Translation: the GPU-only path is exhausted. The agentic chapter requires a platform, not a chip Their seven-chip answer: Vera Rubin NVL72 —capacity and prefill Vera CPU — tool execution, KV cache offload Groq 3 LPX — SRAM-first decode, low-jitter generation NVLink 6, ConnectX-9, BlueField-4, Spectrum-X — fabric Result they claim: 400+ tokens per second per user on trillion-parameter MoE at 400K context. Vera spec: 88 Olympus cores, 176 threads, 1.8 TB/s NVLink-C2C, 1.2 TB/s LPDDR5X, 227 billion transistors. A 256-CPU rack delivers 45,056 threads and 400 TB of memory One detail nobody is talking about. The blog’s second author was previously Head of Agents at Groq. The third was previously at @GroqInc and Intel. NVIDIA didn’t license the LPX architecture. They absorbed the team that built it Haas isn’t pitching a competing thesis. He’s confirming this one from the other side of the table. Arm data center royalties doubled year-on-year. He expects them to double again Things feel slow right now because we’re between platforms. The speedup ships in H2 2026. The architectural argument is over. Deployment is the only variable left I cover this in The Quiet Architect and The Fourth Piece $arm $NVDA

Holy cow, MASSIVE if true:

@BigLehmanski 🙏🍻

@TickerSymbolYOU Congratz, great work man!

YOU make the long nights and weekends spent researching, writing, and editing worth it. Thank you for spending some of your precious time with me each week! 🍻

@alexkoh Thanks Alex! 🍻

@TickerSymbolYOU Totally deserve it Alex! Thanks for all you do!

I'm buying every single share I can:

@jessa_91 @NVIDIADC Just normal price action. Not worried.

@NVIDIADC @TickerSymbolYOU Are you still holding, Alex? Why is it crashing? Market manipulation?

💡 Building and scaling AI requires infrastructure that generates more revenue per watt and drives the lowest token cost. NVIDIA Vera Rubin + LPX unlocks up to a 10x annual revenue opportunity from the same gigawatt-scale data center. More tokens per watt. Higher value tokens = more intelligence + more interactivity.

I got rich off $NVDA. This is even bigger:

@JamieMcullough @Tesla_Tizzler Betting against history is exactly how you win big.

You’re betting against history. It’s always been a cyclical product. Every time memory prices spiked they have made full and complete retracements in price. Shortages always lead to gluts. If you want to say, “this time is different” fine, but statistics are not on your side and you’re betting against history. Prices likely have already plateau’d as DRAM prices have been dropping.

Don’t say I didn’t warn you months ago 😏

@Tesla_Tizzler Confidently incorrect

@TickerSymbolYOU It’s a cyclical product & therefore trading expensive.

@stpitts @nvidia It was great meeting you @stpitts! Thanks for taking the time and looking forward to chatting again in the future. The Groq 3 LPU genuinely blew my mind!

fantastic spending time with you @TickerSymbolYOU! can't wait to share more through the year on what we're building @nvidia and what this means for #AIFactories and #AIDevelopers everywhere.

Wall Street has NO IDEA what $NVDA's about to ship Here's an exclusive look at @nvidia's mind-blowing Vera Rubin systems and the new Groq 3 LPU chips Don't say I didn't warn you:

@kakashiii111 I think you have it backwards. He’s not saying these companies need $NVDA to support them. He’s saying neoclouds require a lot of coordination (chips, racks, cooling, facilities, energy, etc.). But now, the utilization numbers justify all that upfront work and capital

And here it goes, Jensen says out loud what I have been saying since neoclouds existed. Another validation of my work.

@MingleSach No? It just wasn’t in the news

@TickerSymbolYOU Did you drop IREN?!

Wall Street just made a HUGE MISTAKE This is a once in a LIFETIME opportunity for stocks like $MSFT $NBIS $MU $MRVL $AVGO $CRWD $PANW Don't say I didn't warn you:

Spent an hour letting @nvidia's L2++ Mercedes drive me through downtown SF. The @NVIDIAAI team really cooked. If this is what “driver assist” looks like, $TSLA's competition might be here sooner than the market thinks. Long $NVDA. youtube.com/watch?v=MSmSA8…