Seeing tech through a different wavelength. Photonics & semiconductor research https://xitter.cf/t.co/E8DQu9pkCv
San Francisco, CA
Joined April 2025
- Tweets19.8K
- Following2.2K
- Followers70K
- Likes30.4K
Photon Capital retweeted
@labubu_trader @ParadisLabs @damnang2 @PhotonCap @ren_stocks @Mr_Derivatives @SixSigmaCapital @JonahLupton @babyfolio @CKCapitalxx @retail_mourinho @BlackPantherCap @Sandeman52
I’m tagging you because I always value your perspective on where the market and technology are heading. Is there anything you’d add here?
I think it’s critical that we share this with our retail community so they aren't shaken out of what could be the single largest technological transition of our lifetime.
Photon Capital retweeted
Micro-VCSELs vs micro-LEDs
This was a hot topic at OCP 2026
@austinsemis and I sat down on @semidoped right after the show to discuss the future of short reach optics
(even if it means I'm in a garage, because some things are too important to delay 😅)
Enjoy the show! Lots of visuals in there, so make sure you watch on YouTube or right here on X!
🎙️ NEW EPISODE: MicroVCSELs vs MicroLEDs, a MicroExplainer
Austin and Vik host an explainer on the two competing microemitter technologies for AI interconnects, microLEDs and microVCSELs. They break down Credo's surprising pivot at OCP from its microLED acquisition to a microVCSEL product, and the physics that give microVCSELs better speed and reach.
- The new battleground is the 5-50 meter "reach gap," too long for copper limited to ~7 meters and too short for expensive, supply-constrained InP optics
- MicroLEDs are limited to ~3-5 Gbps over 5 meters because their broad light spectrum causes chromatic dispersion that smears the signal
- Hitting 1 Tbps with microLEDs takes ~300 slow lanes, and coupling hundreds of fibers adds a lot of manufacturing complexity
- MicroVCSELs can hit 50 Gbps+ over 50 meters. Stimulated emission gives a narrow beam with less dispersion, for 10x the speed and reach
- MicroVCSELs use a mature Gallium Arsenide (GaAs) process on 8-inch wafers, shared with cell phones, for a cost and scale advantage over InP
- MicroLEDs' niche is heat. Gallium Nitride (GaN) is highly temperature-insensitive, a candidate for short-reach optics inside a hot server tray
This episode is presented by @CrusoeAI. Crusoe's Serverless Fine-Tuning trains an open model like Llama, DeepSeek, or GLM on your own data, and you pay per million tokens, not per GPU hour. About $2 buys roughly 5M training tokens on smaller models. Get $5 in credit to start: semidoped.com/crusoe-x
Chapters:
0:00 The Credo Pivot
3:49 The Talk of OCP
8:13 The 5-50 Meter Reach Gap
12:52 How MicroLEDs Work
21:10 MicroLEDs' Parallelism Problem
22:55 How MicroVCSELs Work
26:57 MicroVCSEL Performance Specs
31:51 The Physics of VCSELs
33:35 The Market Chooses MicroVCSELs
35:45 MicroLEDs' Niche: Heat
38:16 Teasing Future Interviews
Mentioned in this episode:
- PicoJool’s Al Yuen: The Case for GaAs VCSELs in Scale-Up Interconnects (Semi Doped episode): daily.semidoped.com/p/picojo…
- A MicroLED Field Trip to Avicena (Vik's Newsletter, paid): viksnewsletter.com/p/a-micro…
- MicroLEDs vs. Lasers: The Linewidth Tradeoff (Vik's Newsletter, free): viksnewsletter.com/p/microle…
- Understanding Energy Bands in Semiconductors (Vik's Newsletter, free): viksnewsletter.com/p/energy-…
Get more of Austin and Vik daily, free!
Sign up: daily.semidoped.com/
Connect with Vik and Austin:
Vik's Paid Substack: viksnewsletter.com
Austin's Paid Substack: chipstrat.com
@austinsemis @vikramskr
Photon Capital retweeted
- Sivers ($SIVE) received initial production orders from an undisclosed LiDAR customer serving automotive and industrial applications.
- Sivers will supply continuous-wave DFB lasers and optical amplifiers.
- Shipments are scheduled to begin in Q4 2026, with customer programs planned to enter production in 2027.
- The company estimates $53M–$138M in cumulative revenue potential over the product lifecycle; the initial order value was not disclosed.
> The key milestone is a design win converting into actual production orders.
> The $53M–$138M is not firm backlog. Watch how repeat orders and shipment growth translate that potential into revenue.
Photon Capital retweeted
Micron announced that it will provide strategic funding for Indium to expand high-purity materials refining and manufacturing in Rome, New York. The companies also plan a long-term supply agreement.
Potential benefits for $MU Micron:
• Improves access to specialized semiconductor materials.
• Reduces exposure to concentrated overseas refining and export restrictions.
• Supports the buildout of Micron’s U.S. memory-manufacturing footprint.
• May reduce supply-chain disruption risk over time.
Market effect:
Strategically positive, but probably limited near-term earnings impact.
It strengthens Micron’s long-term supply security and could modestly improve investor confidence, but it does not change the immediate memory-demand, pricing, or capex outlook.
The initial market reaction also suggests it was not treated as a major standalone catalyst.
$MU indium!!
- $MU invests in Indium Corporation to expand critical minerals processing and high-purity materials refining in Rome, New York.
- Long-term supply agreement secures $MU's access to semiconductor-grade minerals and reduces reliance on concentrated global refining capacity.
- Investment aligns with $MU's July plan to invest up to $3 billion strengthening U.S. semiconductor supply-chain ecosystem and advancing national security priorities.
> $MU's critical minerals sourcing initiative demonstrates growing recognition that advanced semiconductor manufacturing requires backward integration into upstream materials supply to mitigate geopolitical supply risks and support sustained production scaling for AI-driven demand.
Photon Capital retweeted
Just in!
$SIVE just got production orders!!
sivers-semiconductors.com/pr…
Photon Capital retweeted
The AI Power Shortage Is Overstated
Over the past week the market has started pricing a power shortage as a real threat to the AI trade. I think accelerator shipments can hit their estimates and the US power gap is much smaller than consensus, for two reasons.
First, consensus overstates how much new power the accelerators shipped through 2028 actually need. It ignores that much of the install base will be upgraded to newer, more profitable accelerators rather than added on top of it, and that a meaningful share ships outside the US. Once you normalize for cost and power and back out upgrades and exports, far less new US power is required.
Second, I’m more optimistic than most that new capacity can come online. ERCOT alone is looking at ~10 GW in 2027 and ~17 GW in 2028, ~27 GW total, roughly two thirds of the ~40 GW consensus need. Recip engines, smaller MW sites and crypto miners converting to AI can come online quickly and fill much of the rest.
Photon Capital retweeted
AI 없는 시절로 돌아갈 수 없다.
기술주 가격 눌림은
누군가에겐 또 좋은 entry point 가 된다.
내가 생각하는
사이클이 끝나지 않았다 생각하기에
계속 관련된 글을 쓸것이며
여전히 강세를 바라보고 있다
Photon Capital retweeted
Spent two hours talking with an NVIDIA engineer today.
My biggest takeaway?
DO NOT SHORT $NVDA.
That's it. That's the tweet today.
Photon Capital retweeted
Five semiconductor and optics developments worth watching
- $ASML expects planned supply to fall short of customer demand in 2027–2028.
- $NVDA $CSCO announced USFP plans targeting 400G/lane and expansion to 12.8T.
- $LITE $ALAB $GLW announced a rack-scale PCIe-over-optics proof of concept.
- SK hynix showcased CXL pooled memory and KV-cache sharing and reuse.
- $TER agreed to a $300M equity investment in $FORM, alongside collaboration covering HBM, advanced packaging and silicon photonics.
AI infrastructure is expanding on two fronts, making more chips and keeping those chips connected and productive. Optics, memory and test belong in the same conversation.
Photon Capital retweeted
Humanoid Robots 🤖
- $INTC targets 2028 launch of Razor Lake-AX SoC for robotics and humanoid systems with 700 TOPS AI performance.
- Nova Lake robotics processor scheduled for early 2027 as intermediate step before Razor Lake-AX.
- 700 TOPS would represent roughly 4x the AI performance of current Panther Lake platform used in robotics development.
> $INTC's robotics roadmap indicates a clear strategy to scale AI compute performance for edge robotics applications, with Razor Lake-AX positioned to compete with $NVDA's Jetson ecosystem in the autonomous robotics market segment.
Photon Capital retweeted
- $NVDA presented optics-on-interposer (OOI), placing optical engines on the same interposer as compute chips, at SEMICON West on October 13; THE ELEC reports that a prototype has been developed and remains at the research stage.
Photon Capital retweeted
輝達把光引擎搬上中介層（OOI)
大家以為 CPO 就是把光引擎搬到晶片「旁邊」，已經夠近了。其實 NVIDIA 下一步要把光引擎搬「進」同一片 interposer，跟 HBM 一樣近。
10/13 SEMICON West，NVIDIA 研究員 Nikola Nedovic 講了一個東西叫 OOI，optics-on-interposer。韓國 THE ELEC 10/15 報導，NVIDIA 說已經做出原型。
差在哪？現在的 CPO，訊號從晶片出來，要先過 interposer，再過一層有機載板，才到光引擎。OOI 把有機載板這段拿掉，訊號直接在 interposer 上走，就像 GPU 連 HBM 那樣。
像你住隔壁棟的朋友，以前要下樓、過馬路、再上樓。現在搬到同一層，開門就到。路短，訊號損失少，也比較省電。
同一場還秀了另一招：一條光纖塞 8 種顏色的光，每種 32Gb/s，一條光纖 256Gb/s。
但兩個都還在研究階段。Nedovic 自己說：他們是在評估未來產品的技術，同時處理還不能商業化的問題。
我覺得更值得看的是報導列的量產 CPO 供應鏈：台積電 2330 做 COUPE 矽光子封裝，台積電跟矽品一起代工；外部雷射是 Coherent、Lumentum、住友電工；光纖零件有 Corning、SENKO、波若威 3163、天孚；散熱和最終測試是鴻海 2317 跟 Fabrinet。
Nedovic 還講了一句：要讓光引擎撐過封裝製程，夥伴要做組裝前後的測試、光纖對位、溫控、散熱跟維修。又是測試跟光纖。
$NVDA $TSM $COHR $LITE $GLW #台積電 2330 #波若威 3163 #鴻海 2317 #CPO #OOI
新聞：THE ELEC：Nvidia Unveils Next-Generation CPO Technology
- $NVDA presented optics-on-interposer (OOI), placing optical engines on the same interposer as compute chips, at SEMICON West on October 13; THE ELEC reports that a prototype has been developed and remains at the research stage.
Photon Capital retweeted
I am slowly reviewing slide decks.
Firstly, this is from $ALAB Astera Labs.
1. 144-GPU threshold: NVIDIA's NVL144 uses copper within a single rack, while NVL576 combines copper and direct optical links across eight racks.
2. Copper limitations: Higher SerDes speeds increase insertion loss and signal integrity challenges. AEC retimers extend reach but add power and cost.
3. Architecture distinction: LPO focuses on removing the module DSP, while NPO and CPO move the optical engine closer to the ASIC. These represent different design dimensions.
4. Fiber density: Individual fiber thickness remains constant, but total fiber and connector counts can grow dramatically with bandwidth and network topology.
Photon Capital retweeted
Great summary 👍
quick summary for NPO vs CPO's techinical issues.
Please take a look at it if you are interested in NPO/CPO or investors
- NPO separates the optical engine from the ASIC package, improving thermal isolation and module-level serviceability.
- CPO minimizes electrical path loss but increases thermal coupling and package complexity.
- ASIC and EIC heat can shift microring resonances and degrade optical link performance.
- One published silicon microring example reported a 0.4nm resonance shift over an 8°C temperature change.
- CTE mismatch and thermal cycling can cause package warpage and fiber-to-PIC alignment degradation.
- NVIDIA uses external laser sources to isolate lasers from ASIC heat, but the PIC's resonance drift remains a separate challenge.
- Non-resonant MZMs can offer greater tolerance to wavelength detuning, with area and drive-power trade-offs.
- NVIDIA reports that its collaboration with TSMC has mitigated thermal sensitivity in 200G/lane microring production.
> The critical thermal challenge in CPO is not simply laser wavelength drift. It is maintaining laser-to-resonator alignment, controlling thermal crosstalk, and preserving optical coupling stability.
> The key trade-off is whether microring density and drive-power advantages outweigh the overhead of tuning, feedback control, and cooling.
> NPO prioritizes modularity and serviceability. CPO pursues maximum bandwidth density and power efficiency.
> The NPO vs. CPO debate is not a reason to short the entire optics sector. Both architectures require optical components and connectivity.
> The more compelling investment question is who solves thermal stability, optical coupling yield, and electro-optical testing at scale.
Photon Capital retweeted
Thoughts on Memory
1. I do not agree with the idea of a memory supply glut in 2029. I expect demand growth to accelerate further.
2. Around 2029 to 2030, the memory business will undergo a fundamental transformation. Think of how we no longer view NVIDIA as simply a gaming GPU company.
3. China will remain a competitive threat, but equipment constraints will make capacity expansion harder than many expect. Its impact on the supply and demand balance in 2029 will also be smaller than the market fears.
4. I expect memory stocks to begin rerating within the next six months. This also ties into a broader rerating of the KOSPI.
5. I believe memory needs to rerate first for other AI themes to follow. That view has not changed.
Source: me
Photon Capital retweeted
- Co-packaged optics technology transitioning from R&D phase to manufacturing readiness
- Ayar Labs advancing toward production-scale deployment of integrated photonic solutions
> Ayar Labs is moving co-packaged optics technology beyond prototype validation toward actual production planning, indicating the technology is approaching commercial viability.
> This represents a critical inflection point where technical feasibility translates into manufacturing scalability and market readiness.
eetimes.com/co-packaged-opti…
Photon Capital retweeted
"We need about 10,000 times more AI than we have today." Pat Gelsinger, this week.
Yet the deceleration crowd is reading quarter-on-quarter revenue and calling a top. But QoQ only looks like a peak if you think we're late in the cycle.
Price it against a 10,000x demand ceiling and the same numbers tell the opposite story: we're still early. ~$7.6tn of AI capex is planned for 2026–31, and US data center demand is heading toward ~194 GW by 2035. That doesn't get us close to 10,000x. IMO the supercycle hasn't slowed, however the market is now pricing it like we passed the peak.
- Co-packaged optics technology transitioning from R&D phase to manufacturing readiness
- Ayar Labs advancing toward production-scale deployment of integrated photonic solutions
> Ayar Labs is moving co-packaged optics technology beyond prototype validation toward actual production planning, indicating the technology is approaching commercial viability.
> This represents a critical inflection point where technical feasibility translates into manufacturing scalability and market readiness.
eetimes.com/co-packaged-opti…