Paradi Lab

$NVIDIA(NVDA.US) Q1 Earnings Preview:

My pre-earnings notes & watchpoints.

Nvidia will have implications across the whole AI ecosystem on names like: $Taiwan Semiconductor(TSM.US), $Lumentum(LITE.US), $Coherent Corp.(COHR.US), $Applied Optoelectronics(AAOI.US), $Aehr Test(AEHR.US), $Corning(GLW.US), $AXT(AXTI.US), FOCI (3363), Browave (3163), $SIVE, Samsung, $Micron Tech(MU.US), SK Hynix, $Vertiv(VRT.US), $Super Micro Computer(SMCI.US) etc.

CPO supply chain:

1. Spectrum-X

- any push from H2 2026 into H1 2027, even one sentence of "later this year" framing, is a deferral.

- Spectrum-6 ASIC will power two devices (SN6810 + SN6800) - are either sampling w/ hyperscaler customers yet?

- ECOC in late Sept is the next public catalyst, so Jensen has incentive to set up narrative on this call.

2. Has Quantum-X actually shipped yet + at what cadence?

- "Early 2026" was the original window.

- If Jensen quantifies units shipped / customers ramped / hyperscaler deployments, that's confirmation the COUPE supply chain is holding.

- Silence could imply yield/integration friction at $Taiwan Semiconductor(TSM.US) CPO node + negative read-thru for H2 photonics line.

3. Rubin SiPh

- Jensen has consistently listed the silicon photonics processor as one of the chips of the Rubin platform.

- Is this discrete from the CPO switches, integrated into NVLink 6? Or part of a separate scale-across fabric?

- Clarity here matters because it determines whether Nvidia is doubling CPO content per system or just moving CPO from switches into the rack.

- Doubling content per NVL576 has direct implications on optical engine demand - and therefore to InP substrate, EML laser and PIC volumes.

---

Other watchpoints:

1. Q2 guide & trajectory of Rubin handoff:

I feel like anything below ~$84B reads as a deceleration into the Rubin transition window.

Anything >$88B forces up sellside numbers across 2026.

- Will GB300 Ultra move from sampling -> production shipments this Q?

- Does Vera Rubin remain "Q3 timeframe" of 2026?

A Rubin pull-in would be v. bullish.

2. Gross margin direction:

Nvidia predicts GM will narrow to ~70% as Blackwell ramp continues during the original B200 ramp before recovery.

Rubin starts a fresh dilution cycle in H2 - if Kress signals Q2 GM below 75% (citing GB300 mix / early Rubin spend), the current multiples look way too high imo.

I feel like: Rubin ramp + HBM4 cost inflation + adv. packaging tightness = real downside risk that's hard to hedge against.

---

Just some condensed notes that I think are most relevant to you guys, esp. since we're all in a lot of the CPO-related names.

I personally care less about the headline numbers, but more so about fwd commentary that impacts the rest of the supply chain.

Ok, this is big for $Penguin Solutions(PENG.US) & High Performance Compute:

NextSilicon just announced that the Spectra supercomputer (the system $Penguin Solutions(PENG.US) built + integrated) has achieved full acceptance under Sandia National Labs’ Vanguard programme.

$Penguin Solutions(PENG.US) worked w/ Sandia + NextSilicon to deploy & support the platform so it could run real national security workloads.

E.g. HPCG, molecular dynamics + MonteCarlo sims.

And since the Vanguard programme is infamously selective, it basically validates $Penguin Solutions(PENG.US)'s integration capabilities in HPC environments.

Which should help with:

1. follow-on govt work

2. commercial HPC/AI customers looking beyond vanilla GPU clusters

So, $Penguin Solutions(PENG.US)'s strength in high-performance memory + systems integration (incl. CXL) is probably an advantage here when balancing new accelerator designs?

Disclosure: I don't have a $Penguin Solutions(PENG.US) position.

The fact that we've got a $Cerebras(CBRS.US) 2x leveraged ETF...

Is exactly why we can't have nice things.

GS data shows that Retail are using more and more leverage.

Which is why I think immediate releases of leveraged ETPs by firms like Leverage Shares is extremely predatory.

Profiting off of Retail mania hours post Cerebras IPO.

From my own personal surveying over the years, I'm confident in saying that 99% of leveraged ETP buyers hold for days/weeks at a time.

And not day trading those products, which is what they're designed for.

Tbh, for Retail, I don't advocate for leverage full stop.

SpaceX are going to raise up to $75B on IPO next month.

Which is 3.5x bigger than $Alibaba(BABA.US), the current biggest U.S. IPO.

Very grateful to have crossed 40K followers this week. Truly blown away by the support.

As you guys know, everything I make from X will be donated to charity.

-> GOSH are a London hospital that cares for kids with rare conditions/diseases. Very important work done by selfless individuals.

Did £4K to kinda match the 40K followers lol. Unfortunately I'm not rich enough yet to donate £40K in one go.

Regardless, a long-term goal of mine is to do monthly 5-figure donations to my chosen charities.

(I didn't make £4K from the rev share programme btw. Just less than £500)

Passing up on $Cerebras(CBRS.US) at ~$100B MC.

The bull case is compelling since they're a genuine architectural-paradigm story in AI compute.

Highlighted by a stellar list of customers using their chips:

- OpenAI: $20B+ contract

- $Amazon(AMZN.US): AWS data centers

- $Meta Platforms(META.US): Llama API for devs

- Perplexity: sonar model

- $GSK(GSK.US): biomed models

- Mistral AI: Le Chat

- Notion: enterprise search

- AlphaSense: enterprise search

All of that diversifies them away from G42 in the UAE which have been the historic rev driver.

Regardless, not too sure when I'll look to initiate a position, maybe like the $200 mark?

Kinda feel there's some downside risk rn when you layer on top:

- $NVIDIA(NVDA.US) Vera Rubin volume ramp (H2 2026)

- potential Nvidia absorption of Groq

- CUDA moat

So current valuation looks like all the execution upside is being aggressively priced-in.

Kioxia (285A) just put out sensational earnings & guidance.

Very bullish validation of the NAND supercycle.

I'm viewing this dip as a perfect buying opportunity to cost-avg up.

Q4 FY25 (Jan-Mar 2026):

- Revenue: ¥1,002.9B (QoQ +85%)

- OP: ¥599.1B (margin 60%; QoQ +314%)

- ASPs >2x'd QoQ; bit shipments down ~10% QoQ as planned for inventory control.

- 8th-gen BiCS production now exceeds 5th-gen levels.

Cash flow and balance sheet are pristine post-refinancing.

Q1 Guidance (Apr-Jun 2026):

- Revenue: +75% QoQ

- OP: +117% QoQ; margin 74%

- NP: +112% QoQ

"AI inference servers driving overall NAND demand growth… interest is intensifying."

"CY27: Demand anticipated to exceed supply."

Memory keeps going up and up with every quarter...

Pls DYOR.