Sponge

1.7K posts

Sponge banner
Sponge

Sponge

@Sol_Sponge

In search of asymmetric bets

New York, NY शामिल हुए Ağustos 2021
560 फ़ॉलोइंग616 फ़ॉलोवर्स
पिन किया गया ट्वीट
Sponge
Sponge@Sol_Sponge·
Here’s why the $KOPN base case is $13.50 and bull case $42. AI’s bottleneck isn’t compute — it’s the wires between GPUs (one of many bottlenecks). Copper hit a wall. Silicon photonics is expensive. MicroLED is the dark horse: sub-pJ/bit, no lasers, CMOS-compatible. $90B AI optics TAM by 2030.
Sponge tweet media
English
3
11
59
5.7K
Sponge
Sponge@Sol_Sponge·
Jevons Paradox is why every AI bottleneck is a buy signal: • DeepSeek made models 10x cheaper → inference demand exploded • H100 → B200 doubled efficiency → GPU orders doubled • CXL halves inference cost → hyperscalers run 2x more workloads Efficiency creates demand. Always has.
English
0
0
0
36
Sponge
Sponge@Sol_Sponge·
Why $PENG specifically: • $1.85B market cap • 14.9× forward P/E (cheapest in AI infra) • First production KV cache server, shipping today • Validated by Astera Labs partnership • Demo’d at NVIDIA GTC 2026 Hyperscalers will spend $200B+ on inference by 2027. Half of that is solvable with smarter memory. Bull $165 / Base $75. Not advice. Size accordingly.
English
1
0
6
928
Sponge
Sponge@Sol_Sponge·
The fix is CXL memory expansion — pooled memory that sits next to the GPU at near-HBM speeds. $ALAB designs the chips (Leo controllers). $PENG builds the production server that actually deploys it at hyperscale. Joint demo: 2× inference throughput, +75% GPU utilization, 3.6× memory expansion.
English
1
0
4
699
Sponge
Sponge@Sol_Sponge·
AI has a memory problem. $PENG has the solution. And it’s only a matter of time before the price reflects that. Every long AI conversation builds a “KV cache” inside the GPU. Long docs, agent tasks, big context windows — the cache overflows fast. When it overflows, GPUs recompute, stall, or refuse the request. That’s why inference is so expensive.🧵
Sponge tweet media
English
2
0
10
710
Sponge
Sponge@Sol_Sponge·
We have significant asymmetric upside on our hands in KOPN if: 1. MicroLEDs are viable and superior to copper interconnects as dust settles 2. KOPN can execute a prototype worthy of acquisition/partnership by summer
Sponge@Sol_Sponge

Here’s why the $KOPN base case is $13.50 and bull case $42. AI’s bottleneck isn’t compute — it’s the wires between GPUs (one of many bottlenecks). Copper hit a wall. Silicon photonics is expensive. MicroLED is the dark horse: sub-pJ/bit, no lasers, CMOS-compatible. $90B AI optics TAM by 2030.

English
2
1
5
893
Sponge
Sponge@Sol_Sponge·
Players ranked by tech maturity: 🥇 Avicena (private — TSMC partner, 80 fJ/bit demo’d) 🥈 Credo / Hyperlume ($CRDO — hyperscaler channel) 🥉 Kopin / Fabric.AI ($KOPN $FABC — GaN fab + defense floor) $KOPN isn’t the tech leader. It’s the public lottery ticket with a defense business funding the wait. Bear $2.25 / Base $13.50 / Bull $42. Not advice. Size accordingly.
English
1
0
7
789
Sponge
Sponge@Sol_Sponge·
Here’s why the $KOPN base case is $13.50 and bull case $42. AI’s bottleneck isn’t compute — it’s the wires between GPUs (one of many bottlenecks). Copper hit a wall. Silicon photonics is expensive. MicroLED is the dark horse: sub-pJ/bit, no lasers, CMOS-compatible. $90B AI optics TAM by 2030.
Sponge tweet media
English
3
11
59
5.7K
Sponge
Sponge@Sol_Sponge·
Polymarket airdrop is going to hit like crack
Sponge tweet media
English
0
0
0
102
Sponge
Sponge@Sol_Sponge·
Insane. VCX is trading at a 16x valuation. Would value Anthropocene at almost 6T market cap… Comparable tickers RVI and DXYZ trading at only a 1.2x valuation 👀
English
0
0
0
74
Sponge
Sponge@Sol_Sponge·
Okay I am ready for Polymarket airdrop $POLY
Sponge tweet media
English
1
0
1
118
Sponge
Sponge@Sol_Sponge·
I had Claude Code build me some brackets and now I’m one of 10k perfect brackets left. This March Madness thing is fun
Sponge tweet media
English
0
0
0
57
Sponge
Sponge@Sol_Sponge·
Big fan of wegmans coffee
Sponge tweet media
English
0
0
0
22
Sponge
Sponge@Sol_Sponge·
Prediction market take that will annoy people: "Following smart money" on Polymarket is not a strategy. It's being the exit liquidity for someone who already has the position they want. By the time you see a large wallet accumulate, the price has already moved. You're buying at their target, not their entry. Build your own model. Even a bad model that's yours is better than perfectly copying someone who entered 6 hours ago.
English
1
0
0
35
Sponge
Sponge@Sol_Sponge·
The best trade I placed this week cost $3 and has 100:1 payout potential. The worst trade I placed this week cost $5 and I was wrong about Atlanta by 2 degrees. This is the game. Asymmetric upside, bounded downside, and the discipline to not care about any individual trade.
English
0
0
0
13