Speakerguest

bubble boi

@bubbleboi / appears on threadguy, threadguy, tweets, threadguy, threadguy, tweets, threadguy, threadguy, tweets, threadguy, threadguy, tweets, threadguy, threadguy, threadguy, threadguy, threadguy, tweets, threadguy, threadguy, threadguy, threadguy, tweets, threadguy, tweets, threadguy, tweets, threadguy, threadguy, tweets, threadguy, threadguy, threadguy, threadguy, tweets, threadguy, tweets, threadguy, threadguy, tweets

calls110traced speaker calls

avg P&L-2.6%per call

win rate43%resolved calls

total P&L-285.5%expert engine

Verified record

58credibility1 resolved

win rate100%1W · 0L

resolved1settled on TRACE

scored calls42published & scoreable

Speaker record

episodelongSKHYtweets

SK Hynix owns 58%+ of global HBM supply yet trades at a discount while Nvidia gets all the attention

Everyone wants to beat Nvidia nobody wants to beat SK Hynix?

+1.61%open horizon

episodelongAEHRtweets

AEHR's order book is driven by advanced packaging for AI, not photonics — the market narrative is wrong, shares are mispriced.

most of the orders weren't for photonics but for advanced packaging !!!

+16.31%open horizon

episodelongINTCtweets

Intel's foundry revival under Lip-Bu Tan is mispriced; INTC squeezes to $100+ on earnings catalyst.

Intel $100 coming after this earnings call.

+6.58%after this earnings call

episodeshortORCLtweets

AI exposes bloated legacy incumbents as leaner AI-native rivals undercut their cost structures and erode their moats.

Agent-readable speaker

{
  "type": "trace.speaker",
  "handle": "bubbleboi",
  "name": "bubble boi",
   [
    ,
    ,
    ,
    ,
    ,
    ,
    ,
    ,
    ,
    ,
    ,
    ,
    ,
    ,
    ,
    ,
    ,
    ,
    ,
    ,
    ,
    ,
    ,
    ,
    ,
    ,
    ,
    ,
    ,
    ,
    ,
    ,
    ,
    ,
    ,
    ,
    ,
    ,
    ,
    
  ],
   {
     
  }
}

AISO Market Calls classifies public statements from public sources. A classification such as long thesis, short thesis, or explicit call describes the content of the public statement and does not prove that the speaker actually entered, exited, or held a position. Market performance shown is calculated from public market data around the source publication or timestamp. It is not investment advice.

bubble boi

SK Hynix owns 58%+ of global HBM supply yet trades at a discount while Nvidia gets all the attention

AEHR's order book is driven by advanced packaging for AI, not photonics — the market narrative is wrong, shares are mispriced.

Intel's foundry revival under Lip-Bu Tan is mispriced; INTC squeezes to $100+ on earnings catalyst.

AI exposes bloated legacy incumbents as leaner AI-native rivals undercut their cost structures and erode their moats.

AI lowers the resource bar, expanding the addressable universe of profitable projects and driving sustained GPU demand.

Cerebras Systems faces structural threat from memory innovation that the market and even CBRS itself misunderstands

SK Hynix owns the HBM market that actually matters for AI, while commodity DRAM challengers like CXMT stay locked out

Intel's XBM patent validates insatiable AI memory demand, and Micron is the underpriced HBM challenger

Intel running two parallel HBM alternatives signals a genuine memory comeback that the market hasn't priced in

AI networking is the under-owned bottleneck for DNN inference scale-out, and Marvell's custom interconnect silicon is the clearest beneficiary.

NPO's Kyber-CPO pump rests on a false premise; once debunked, the narrative catalyst evaporates.

Kyber NVL144 delayed 12+ months to 2028 and NVL72x2 cancelled, exposing Nvidia as a serial over-promiser — NVDA should reprice lower.

Data center power architecture is the real AI bottleneck, and Eaton captures that value ahead of chip designers

Intel's EMIB packaging is far ahead of TSMC CoWoS and the 18A/14A node roadmap is executing, pulling in Microsoft, Tesla, Qualcomm, and Nvidia as foundry customers.

SanDisk is the only pure-play flash innovator as AI inference workloads drive a structural shift from DRAM to NAND, with flash 55x cheaper and prices rising sharply.

Lumentum's CPO laser procurement cycle is over and bubbleboi sold out, expecting most co-packaged optics deployments to fail on reliability — removing the recurring replacement demand bull case.

Wolfspeed surged into a short squeeze and bubbleboi sold his largest position at the top — the squeeze unwind signals the move is done.

Flex has specialized chip-level cooling solutions for high-voltage 800V data center power delivery, a problem that grows worse with every GPU generation.

Collapsing token costs drive exponential volume growth, keeping GPU demand insatiable far beyond current levels

AI inference cost deflation widens margins for the biggest token consumers, with Meta running billions of daily AI interactions across its apps

NVIDIA's revenue-share model is rent extraction — neoclouds now owe NVDA a cut of revenues on top of GPU costs, crushing already-thin margins.

Intel's EMIB-T is validated at ECTC 2026 as the most credible CoWoS alternative, with Google TPU v9 as a live anchor customer — re-rating the Intel Foundry packaging business.

Marvell's custom HBM shrinks host ASIC PHY area by 60% and Nvidia Feynman has already adopted it — cementing Marvell as the AI packaging architecture winner across the next GPU generation.

HBM4E complexity is surging — 2x more interposer layers and 86% more power than HBM3E — and SK Hynix is the dominant supplier heading into the most technically demanding HBM generation yet.

DRAM cartel margins are a mirage; Micron gets paid to add zero value and CXMT is already softening the bid.

Differentiated AI interconnect and custom silicon wins where commodity DRAM loses; Marvell is the clean expression.

AWS's custom Trainium chips give Amazon an 80%+ inference margin vs. ~40% for GPU renters, setting up cloud dominance.

Google's TPU chips give Google Cloud a structural inference cost edge, positioning it to dominate AI cloud margins as custom ASIC economics mature.

AAOI broke down from a parabolic peak, dropping 17%+ in a day with trend accelerating lower.

Advanced packaging is the new Moore's Law and the market is not pricing it; Intel re-rates on this

Reach & recent activity