1/11
CXL is becoming one of the hottest terms in the AI memory market.
But many discussions skip the basics and jump directly to “CXL beneficiaries.”
So before discussing stocks, here is a beginner-friendly guide:
If CXL is not memory, why could it reshape the memory market?
I spend half my time in AI investing thinking about “grand strategy”.
Woke up at midnight with a realization. Anthropic might actually be The One. The One to pull away, achieve dominance, regulatory capture and control the AI supply chain.
Imagine the following…
Finally had a chance to dig into why $NVDA is so interested in $NOK and why it invested $1 billion
➡ Nvidia is interested because Nokia has carrier-grade RAN software. “RAN software” hooks your phone up to a cellular network. It includes baseband/radio products, telecom customer access, data-center switching software, network automation, and optical transport.
➡ The biggest strategic product is Nokia anyRAN / Cloud RAN running on Nvidia GPUs.
➡ The most underappreciated product angle is Nokia SR Linux + telemetry/fabric management for Nvidia Spectrum-X and AI data-center networking. This statement means the following in layman’s terms:
➡ Nvidia builds the AI engines. Nokia may help manage the roads, traffic lights, sensors, and control system that let those engines talk to each other inside huge AI data centers.
➡ The long-term dream is this:
Nvidia compute becomes the standard accelerated platform for AI-native 5G/6G networks, and Nokia provides the telecom software, radios, and networking infrastructure that make it deployable.
➡ There are essentially 6 Nokia products Nvidia is interested in… many of these products can help in the data center environment, not just in mobile networks…
Not sure if people realized this but unless a thesis completely breaks, companies like $NBIS can keep growing.
Just look at $AMZN or $GOOGL over the past 15 years.
If people "trim" it often triggers taxes. And a lot of corrections are typically less than those taxes paid.
By the time a "50% crash happens", it's probably already compounded hundreds of even thousands of percent.
If people need to pay expenses, once you hit 7-8-9 figures, you can always borrow against those assets and keep letting them appreciate.
NFA, just personal opinion. You all do you, but it's highly, highly, dependent on the companies you pick.
Can't do this with something trash like $IREN. But I do believe $NBIS is positioned to be the next hyperscaler.
$MELI Q1’26: el mercado se enojo por la compresión de margen (operating bajó de 12.9% a 6.9% YoY) y se perdió el plot.
Lo que parece deterioro es contabilidad. Lo repito como un loro. Bajo US GAAP (ASC 326 / CECL), cuando $MELI emite un préstamo nuevo tiene que provisionar el expected credit loss de por vida el día 1 (antes de cobrar un solo peso de interés). Si el credit book crece +87% YoY (a $14.6B) y el revenue +49%, las provisiones del trimestre se comen una parte desproporcionada del P&L vs. los ingresos que esos mismos préstamos van a generar a lo largo de su vida. Dos tercios de la compresión vienen literal de ahí. No es que el negocio se deteriore, es que aceleraron originación y la contabilidad cobra antes que el cliente pague.
Lo asimétrico está en los cohorts. Cada tarjeta emitida arranca con pérdida contable. Recién entre el mes 12 y 18 la cohorte cruza NIMAL breakeven. Hoy ~75% del book de tarjetas en Brasil ya pasó ese punto y está en modo cobranza pura. Lo que duele en el P&L de Q1’26 es la cosecha 2026 que recién arranca a marcar pérdida, pero esa misma cosecha va a estar pagando spread durante años. Mientras más rápido emitan hoy, más feo se ve el P&L de hoy y más gordo el de 2027-2028. Es el trade-off clásico de un book de crédito que crece: pagás caja primero, cobrás P&L después.
El otro tema pesimamente interpretado: la duration de personal loans en Brasil pasó de 5 a 8 meses. Sí, más riesgo, más provisión upfront.
Pero contexto: los bancos tradicionales operan personal loans hasta 60 meses (Itaú) y 72 meses (Bradesco), y mis bros en $NU (donde también laburé) hoy llega a 48 meses en su Empréstimo Pessoal. $MELI a 8 meses sigue siendo de los plazos más cortos del mercado por mucha distancia. Si Selic se mueve, repricean en meses, no en años. Es duration extension con paracaídas: capturás más yield y subís ARPU sin casarte con un balance largo.
Sumemos: revenue +49% (más rápido en casi 4 años), +17M nuevos active buyers en 12 meses (84M total), TPV +50%, AUM +77%, items sold en Brasil +56% (más del doble que antes del cut de free shipping threshold). NPL 90+ levemente mejor a 8.0%, allowances cubren 149% de los créditos vencidos +15 días.
El mercado está cobrando la compresión de hoy y no está pagando nada por la opcionalidad de mañana. Y acá está el dato que el sell-side todavía no internaliza: $MELI está ganando share en Brasil mientras $SE acelera en su mercado más importante (GMV global de Shopee +28% YoY en Q3’25, el ritmo más rápido desde Q2’24, con Brasil absorbiendo los recursos liberados al cerrar Chile y Colombia).
Aún así, $MELI elevó conversion rates 1pp YoY en Brasil (un movimiento gigante a esta escala) con NPS y market share en récords. Cuando ganás cancha mientras tu principal rival pisa el acelerador, no es momento de ahorrar capital. Es momento de double down.
La factura llega en 2027-2028 en forma de operating leverage IMHO. Y ahi it rips.
No es recomendacion. Yo me tomo el tiempo de tener mi propia opinion en base a hacer la tarea. Abrazo
AI Semiconductor Endgame 2026 (Part 1)
New Token Economics Computing Paradigm Shifts from GPU Compute to HBM
This article starts from the essence of GPU architectural evolution to address a question the market has long worried about:
Why must each GPU's HBM memory demand grow exponentially, and why won't this exponential growth in HBM demand stall?
It then derives the first principle of token economics under the current architecture: token throughput = HBM size × HBM BW (bandwidth)
It also discusses why the GPU ceiling is determined by HBM's two dimensions of progress.
The topic of HBM cyclicality has long been controversial. Optimists argue that AI-driven demand is much greater than before, but the market mainstream still believes that previous up-cycles also saw 20%+ annual demand growth — so what's different this time? AI doesn't change the fact that HBM, like traditional DRAM, has commodity attributes. Once capacity expansion at the demand peak meets a downturn, history will repeat itself. We can take the perspective of compute-chip architecture, start from first principles, and unpack and reason through this question:
why this time is genuinely different.
———————————————————————————————
History: The Era of CPU Compute
For a very long time, we lived in the era of CPU-dominated compute. The CPU's top-level KPI was performance — running faster — and so each generation of CPUs deployed every method imaginable to push benchmark scores higher. First it was rising clock frequencies, then it was architectural evolution: superscalar designs, and so on.
During this period, why didn't DDR need to advance technologically at high speed? DDR3 to DDR5 took a full 15 years.
Because in this era, DDR's role was purely auxiliary — and only weakly so. By industry experience, even doubling DDR speed would generally only raise CPU performance by less than 20%.
Why did improvements in DDR bandwidth and speed matter so little? Two reasons:
1. CPUs designed all kinds of architectural tricks to hide DDR latency — superscalar designs, wider issue widths, massive ROBs and register renaming to extract parallelism and hide latency, L1 caches, L2 caches — all of which weakened the demand for DDR bandwidth and speed.
2. CPU workloads don't have particularly demanding bandwidth requirements. For most everyday workloads — say, opening a webpage — DDR bandwidth is severely overprovisioned. Even cloud workloads often look the same.
In other words, in the CPU era, DDR bandwidth and speed didn't really matter. There was virtually no difference between DDR4 and DDR5 except in a handful of games — and even the JEDEC standard advanced slowly.
On top of that, only a small portion of any given app needs to permanently sit in DDR. Whatever is needed can be paged in from the hard drive on demand. App size grew slowly, and so DDR capacity demand grew slowly as well.
That's why, over the past decade, the average PC went from 7–8GB of DDR to about 23GB — only 3× growth in ten years.
This slow upgrade pace directly affected revenue. Capacity-based pricing was the main way of making money; speed improvements were just a technological upgrade that raised the unit price of capacity. With both of these dimensions advancing slowly, growth could only come from increases in PC/phone unit volumes.
So along both dimensions — bandwidth/speed and capacity — DRAM was always a “nice-to-have” appendage to the chip industry. The marginal utility of DDR upgrades was very low, and almost completely disconnected from the CPU era's top-level KPI.
———————————————————————————————
The Paradigm Shift: GenAI's Top-Level KPI
When we entered the era of GenAI large models, the computing paradigm shifted, and the top-level KPI changed fundamentally.
By the time GPUs evolved into AI inference engines, the top-level KPI was no longer compute alone (TOPS/FLOPS), as it had been for CPUs — it became the cost of a token. Specifically: overall token throughput per unit cost / per unit power.
A close second is token throughput speed — because in the agent era, many tasks have become serial, and token output speed has become a critical bottleneck for user experience.
This is exactly why Jensen invented the concept of the AI factory: to produce the most tokens at the lowest cost, while pushing token throughput speed as high as possible.
In the AI training era, Jensen's economics were TCO (Total Cost of Ownership): the more GPUs you buy, the more you save.
In the inference era, Jensen's token economics flip the logic:
AI inference has very healthy gross margins, so the logic now becomes: the NVIDIA GPU is the GPU that produces the cheapest token in the world, so the more you buy, the more you earn.
The top-level KPI has become a Pareto frontier: along the two dimensions of token throughput and token speed, optimize as far as possible.
Each generation of NVIDIA's token factory is essentially pushing the entire Pareto frontier up and to the right. This is the most important KPI of the AI inference era.
———————————————————————————————
From Token Throughput to HBM: The Core Logic Chain
Below is the most important logical chain of this article: how to start from the exponential growth of token throughput and derive that the ceiling bottleneck lies in the exponential growth of HBM size and HBM speed.
In the era of single-GPU inference with single-thread batch size = 1, token throughput had only one dimension: HBM bandwidth speed. Higher bandwidth = higher token throughput.
But once we entered the NVL72 era, inference is no longer single-GPU. It is a system-level token factory composed of 72 GPUs + 36 CPUs, designed to fully saturate HBM bandwidth and compute simultaneously, in pursuit of the ultimate token throughput.
Token throughput growth depends on two things: the number of requests batched simultaneously × the average token speed per request.
That is: batch size × token speed.
Take Rubin NVL72 as an example. At an average token speed of 100 tokens/s, processing 1,920 simultaneous requests yields a token throughput of 192,000 tokens/s. A Rubin NVL72 draws roughly 120kW (0.12MW), so per MW it can handle 1.6M tokens/s.
So we need to find ways to push both parameters up: batch size and average token speed. Their product is our top-level KPI — token throughput.
Parameter 1: Batch growth — bottleneck is HBM size
Every request in the batch carries its own KV cache, which has to live in HBM, with sizes ranging from a few GB to tens of GB. Because hot KV cache must be read at high frequency and high speed at any moment, it must reside in HBM. For a model with, say, 80 layers, every token generation step requires reading the KV cache 80 times from HBM.
As batch size grows, hot KV cache grows linearly.
And because the hot KV cache for every request in the batch must sit in HBM, HBM size must grow linearly with batch size.
Like an airport shuttle bus: the gate wants to move passengers to the plane as fast as possible. If HBM size is small, the shuttle is small, so you have to make extra trips.
Conclusion: batch size growth bottlenecks on HBM size growth.
Parameter 2: Average token speed per request — bottleneck is HBM bandwidth
The decode-phase speed of a large model bottlenecks on HBM bandwidth, because every token generated requires reading the activated weights and KV cache many times over.
The emergence of LPUs has, in cases where batch size isn't very large, moved the activated weights portion onto SRAM — but every generated token still requires many reads of the KV cache from HBM. The higher the HBM bandwidth, the faster each token is generated, in essentially linear correspondence.
Like the airport shuttle bus: HBM bandwidth is like the width of the door — wider doors mean passengers board faster.
The rest of the GPU's configuration is essentially adapted to support batch growth and to keep token compute speed in step with HBM growth. In some cases the GPU even spends excess compute to recover effective bandwidth (e.g., bandwidth compression techniques).
—-------
To return to the shuttle bus analogy:
• Shuttle bus cabin size = HBM Size (capacity): determines how many passengers can fit at once (i.e., how many requests' KV caches can sit in HBM simultaneously). Bigger cabin = more passengers (higher batch size) per trip. If the bus is too small, moving 100 people takes two trips — and total throughput suffers.
• Shuttle bus door width = HBM Bandwidth: determines how fast passengers get on and off. A wide door, and everyone piles on at once (decode/token generation is fast). A narrow door, and even with a giant cabin, people queue up and most of the time is spent boarding.
• Passenger throughput = cabin size × door-width-determined boarding speed.
—-------
At this point, we've logically derived the first principle of token-economics hardware demand:
Token throughput = HBM size × HBM Bandwidth
The top-level KPI of the AI inference era is highly dependent on progress along both HBM dimensions.
If we want to maintain 2× token throughput growth per generation, that means each generation of single GPU must grow HBM size × HBM BW speed by 2×!
This is the first time in history that HBM memory size can influence the top-level KPI — token throughput.
To validate this thesis, we can put NVIDIA's token throughput from A100 to Rubin Ultra on the same chart as HBM size × HBM BW speed.
What you find is that the two curves track each other startlingly closely on log axes.
HBM size × speed actually grows even faster than token throughput — which makes sense, because HBM defines the ceiling, and in practice utilization of that ceiling is very hard to push to 100%. Even if HBM size × HBM speed grew by 1,000×, with the supporting compute and architecture, it would be very hard to wring out the full 1,000× of headroom.
This curve isn't a coincidence — it's the necessary solution of system optimization.
throughput = batch × speed. This is the unavoidable first principle of token factory economics.
—-------
What about software? Won't software optimization reduce bandwidth demand? Reduce HBM demand?
This is an independent dimension from hardware. It's like asking: if software on a CPU runs faster after optimization, does that mean the CPU doesn't need to advance for ten years? After all, software is faster now.
If that were the case, would CPU vendors still make money? For a CPU vendor to survive, there's only one path: in standardized benchmarks, ignoring software optimization, every new CPU generation must score higher — otherwise it doesn't sell.
GPUs are exactly the same. How well software is optimized, and the requirement that the GPU's own token-throughput KPI must improve dramatically every year, are two separate things.
As long as token demand keeps growing, the pursuit of higher token throughput will not stop — and so neither will the pursuit of higher HBM size × HBM speed.
If HBM size and HBM speed were to slow down, Jensen would personally fly to the Big Three and pressure them to accelerate, because that ishis GPU ceiling. If the ceiling stops rising, can his GPU still sell?
Of course, NVIDIA also needs to wrack its brains to extract performance beyond the HBM ceiling through heterogeneous architectural angles. The LPU is a great example — it improved the Pareto frontier substantially from a different angle (the right-hand high-token-speed portion).
—--------------------
HBM memory has now bid farewell to that old era of drifting with the tide. On this one-way road paved by exponential demand, it has, in something close to a destined fashion, walked onto the central stage of the industry's epic.
When the inference paradigm's first principles evolve to this point, as long as Jensen still wants to sell GPUs, HBM must double — and it must double every generation. This is endogenous pressure from the supply side. It has nothing to do with AI demand, nothing to do with macro cycles, and nothing to do with the moods of the hyperscalers.
The only remaining question is this:
When demand has been physically locked into exponential growth, will the three players on the supply side — like they have for the past thirty years — once again drag themselves back into the mire of the cycle by their own hands?
$CRDO - you guys have been asking a lot for this- so here is a thread on Credo Technologies- the fastest growing 1B scale company I’ve ever seen.
AI networking is changing fast. At the center is Credo’s Linear Pluggable Optics (LPO) — a new way to connect data centers with lower power, lower cost, and lower latency. Here’s the breakdown: 🧵
Dr. Morris Chang(founder of $TSM)'s autobio is published this month but only in Chinese. Instead of learning Chinese I did a brief summary to help you people. Check it out in my bio.
TSM and NVDA up ~50% from the early August Yen carry trade blow up. 2 months and change. The miserables will complain about inefficient markets and how this stuff is a joke, but the opportunities are beautiful 🙏🏼
Repost if you love inefficient markets 😂
Some of my highlights from $AMZN's Jassy's just-released shareholder letter:
- "We have increasing conviction that Prime Video can be a large and profitable business on its own"
-on AWS: "By the end of 2023, we saw cost optimization attenuating, new deals accelerating, customers renewing at larger commitments over longer time periods, and migrations growing again."
- 2Drones will eventually allow us to deliver packages to customers in less than an hour. It won’t start off being available for all sizes of packages and in all locations, but we believe it’ll be pervasive over time2
- on Pharma: "The same is true for Amazon Pharmacy. Need throat lozenges, Advil, an antibiotic, or some other medication? Same-day facilities already deliver many of these items within hours, and that will only get shorter as we launch Prime Air more expansively."
- "However, a question people never ask, and might be even more interesting is what’s the next set of primitives you’re building that enables breakthrough customer experiences? If you asked me today, I’d lead with Generative AI"
- "Bedrock is off to a very strong start with tens of thousands of active customers after just a few months"
- "Similarly, with a cloud computing business at nearly a $100B revenue run rate, more than 85% of the global IT spend is still on-premises."
- "this GenAI revolution will be built from the start on top of the cloud."
“You know what's cooler than 1 Billion… 2 Billion” - Mr. Enclave
Switchboard has breached another milestone: $2B+ in TVS 🏆
📈 Key reasons for growth:
✅ Trust in Switchboard
✅ Anticipation for On-Demand oracles
✅ Strong presence in the Solana ecosystem
"With the On-Demand model, users can choose their gas cost with the lowest latency possible for their transactions" - @DoctorBlocks
Listen to what @DoctorBlocks says about functions and the importance of Switchboard's On-Demand model 👇
🔮 Something BIG is brewing at Switchboard
Operation [X] is evolving... 🌀
Get ready for a revelation that will redefine our journey and your role in it.
Soon, we make history! 🫡
Switchboard On-Demand is here, and it's a game changer!
Switchboard On-Demand is live on Solana Devnet, bringing secure, efficient, and insanely fast data on-chain ⚡