Joined May 2018
- Tweets 6,561
- Following 1,845
- Followers 1,562
- Likes 100,375
721 Photos and videos
Digital Optimus
1
7
1,058
Ryan Wang 🇹🇼 retweeted
While semiconductor supply chain analysis has traditionally focused on equipment, materials are becoming a critical area of discussion. For example:
-In the CoWoS supply chain, we see multiple material suppliers involved.
-In Optical Engines, UV resins are being used to connect FAUs (Fiber Array Units) and optical modules.
-In Hybrid Bonding, specialized adhesives are applied at wafer edges to prevent wafer sliding during wafer-on-wafer (WoW) processes.
-In thermal management, the evolution of TIM (Thermal Interface Materials) has introduced new solutions.
This signals a gradual paradigm shift in semiconductors—from equipment-centric to material-centric innovation.
tspasemiconductor.substack.c…
2
18
114
6,171
Will Grok Build integrate with Cursor? @grok
SpaceX Acquires Cursor AI In $60 Billion Deal As Coding Agent Race Heats Up zerohedge.com/ai/spacex-acqu…
2
1
252
Grok 5 comes in soon
1
5
384
Amazing girl
Jun 15
Lots of misinformation being spread about me the last couple days, so some quick facts
- My name is Tina, not Guo Can (or Jessie Anderson). I’m one of many Raptor flight operators on console since flight2. Before that, I wrote control software for the vehicle, and was a stage software operator for flight1
- Been living in Starbase since surborbital days in 2020, absolutely love it down here. The people are wonderful and so so excited about the mission - the lows are lows but the highs are very high. My friends here are the best in the world, and I love them to the moon/mars and back :)
- The reason I decided to say something was because facts matter, but also because wanted to share my real life journey to how I got here. I don’t have a masters or a PhD, I started full time directly after college after 2x internships also at spacex doing software/automation. I was on a couple design teams in college, including Stanford solar car mars rover. When I started spacex as a software engineer, I knew very little about fluids / propulsion engineering - I learned a lot of it on the job with some pretty incredible mentors. Then I swapped over to propulsion about halfway through my career and have been loving it ever since
1
8
1,531
Wife order:
Buy a new AI4 Model Y and transfer my FSD to her.
😅😅😅
Breaking News!
FSD will land in Taiwan soon!
@Tslachan @SawyerMerritt
@Tesla_AI @elonmusk
@Tesla_Taiwan announced that on June 15, it has submitted the application documents for Full Self-Driving (Supervised), or FSD to the Taiwan Vehicle Safety Certification Center (VSCC).
Tesla will actively collaborate with the Ministry of Transportation and Communications and the VSCC to initiate the review process for this new technology driving system, and will continue to push for FSD to become available on Taiwan roads as soon as possible.
At the same time, in accordance with Tesla’s global FSD usage model transition and activation plan, the one-time purchase option for FSD in Taiwan will cease to be available for sale after 23:59 on June 30, 2026. This change does not affect the lifetime ownership rights of vehicles that have already purchased FSD. All current Taiwan owners and customers with pending deliveries can still add FSD via one-time purchase until June 30 (inclusive). Additionally, HW3.0 vehicle owners who have already purchased FSD can enjoy free FSD transfer if they complete a new vehicle order by September 30 .
3
14
1,862
台灣特斯拉正式跟台灣官方申請FSD落地許可
Breaking News!
FSD will land in Taiwan soon!
@Tslachan @SawyerMerritt
@Tesla_AI @elonmusk
@Tesla_Taiwan announced that on June 15, it has submitted the application documents for Full Self-Driving (Supervised), or FSD to the Taiwan Vehicle Safety Certification Center (VSCC).
Tesla will actively collaborate with the Ministry of Transportation and Communications and the VSCC to initiate the review process for this new technology driving system, and will continue to push for FSD to become available on Taiwan roads as soon as possible.
At the same time, in accordance with Tesla’s global FSD usage model transition and activation plan, the one-time purchase option for FSD in Taiwan will cease to be available for sale after 23:59 on June 30, 2026. This change does not affect the lifetime ownership rights of vehicles that have already purchased FSD. All current Taiwan owners and customers with pending deliveries can still add FSD via one-time purchase until June 30 (inclusive). Additionally, HW3.0 vehicle owners who have already purchased FSD can enjoy free FSD transfer if they complete a new vehicle order by September 30 .
1
3
14
1,994
Breaking News!
FSD will land in Taiwan soon!
@Tslachan @SawyerMerritt
@Tesla_AI @elonmusk
@Tesla_Taiwan announced that on June 15, it has submitted the application documents for Full Self-Driving (Supervised), or FSD to the Taiwan Vehicle Safety Certification Center (VSCC).
Tesla will actively collaborate with the Ministry of Transportation and Communications and the VSCC to initiate the review process for this new technology driving system, and will continue to push for FSD to become available on Taiwan roads as soon as possible.
At the same time, in accordance with Tesla’s global FSD usage model transition and activation plan, the one-time purchase option for FSD in Taiwan will cease to be available for sale after 23:59 on June 30, 2026. This change does not affect the lifetime ownership rights of vehicles that have already purchased FSD. All current Taiwan owners and customers with pending deliveries can still add FSD via one-time purchase until June 30 (inclusive). Additionally, HW3.0 vehicle owners who have already purchased FSD can enjoy free FSD transfer if they complete a new vehicle order by September 30 .
6
9
100
75,948
Ryan Wang 🇹🇼 retweeted
Jun 15
1/ Standard transformers have a fundamental topological flaw: they cannot track dynamic states over time without running out of layers.
Once a state representation reaches the top layer of the feedforward stack, the model's ability to update its belief collapses. 🧵
10
87
592
106,909
Cool
我用 AI 制作的 轨道架构 科普：
- 轨道倾角
- 轨道高度
- 轨道面
它们基本定义了太空轨道的特征。
3
443
Ryan Wang 🇹🇼 retweeted
Jun 15
我最新的產業調查顯示，聯發科內部已將 AI 事業的策略定位，從「IC / ASIC 設計」提升至「系統級別設計」，首要目標鎖定 Google TPU 的 PCBA（L6），以及 Elon Musk 旗下公司自研 AI 晶片的 L10 機櫃。
整體而言，此定位轉變符合產業趨勢，若聯發科執行順利，有助於強化客戶關係與長期競爭優勢。
調查與分析：
▎此轉變爲長期規劃，2 年內對基本面的影響可忽視，目標在於掌握新成長契機，並降低潛在風險影響：
1. 機會：伺服器機櫃設計漸趨複雜（導入 CPO、800V HVDC 等），加上與消費電子相當的更新速度，共同推升了系統級設計的附加價值。
2. 風險：ASIC 設計的成長動能，在 2-3 年後可能會因爲 Semi-COT 商業模式而開始趨緩。
3. 聯發科為確保系統級別設計整合的業務毛利率至少能達 40-50%，預期採「主導設計與驗證」的輕資產模式，並善用台灣硬體供應鏈生態優勢，將製造外包。
▎Google TPU 的 PCBA：
1. 聯發科的目標是自 TPU v10（Icefish）開始，並同步爭取導入自家 CPO 方案。
2. Google 硬體組裝生態已完備，聯發科爭取 L10 勝算不高。
▎Musk 旗下公司自研 AI 晶片的 L10 機櫃：
1. 目前 Musk 旗下公司建置的 AI 算力主要採 Nvidia 晶片方案，故自家 AI 晶片機櫃組裝生態尚未完備，這是聯發科的機會。
2. 此業務目前尚缺乏明確時程能見度；長期成敗的關鍵在於，聯發科能否善用台灣硬體供應鏈生態，並借助與 Terafab 的合作關係，拿到 L10 機櫃訂單。
9
44
317
41,734
Ryan Wang 🇹🇼 retweeted
Jun 15
420
1,083
4,847
67,339,222
Jun 15
程正樺這一場還是挺值得的
聽完第一直覺也是MTK跟大立光還不買爆嗎？
1
4
719
Jun 15
Excellent popular science article
흔히 AI 칩 성능 볼 때 연산 속도(FLOPS)만 보는데, 진짜 중요한 건 따로 있음. 반도체 아키텍처 뜯어보면 왜 엔비디아가 괴물인지, 빅테크들이 어디서 뚝배기 깨지고 있는지 보임. 하드웨어 관점에서 핵심만 요약해 드림.
1/ 아무리 엔진이 좋아도 도로가 1차선이면 차가 밀리겠지? 반도체도 똑같음.
연산 장치 아무리 빨라도 데이터 퍼다 나르는 '메모리 대역폭'이 좁으면 노답임. 이걸 '메모리 벽(Memory Wall)' 병목이라고 부름.
2/ 이걸 해결하려고 칩 바로 옆에 메모리를 아파트처럼 높게 쌓아 올린 게 바로 HBM임.
차선을 수천 개로 뚫어준 거임. 반면 칩 안에 들어가는 SRAM은 빛의 속도로 빠르지만,
너무 뚱뚱해서 많이 넣으면 칩 가격이 우주로 감. 결국 이 밀당을 잘해야 함.
3/ 엔비디아 GPU 속을 현미경으로 들여다보면 연산 장치(ALU) 말고 '제어 장치(Control Unit)'가 공간을 엄청 차지함. 엔비디아는 전 세계 누가 어떤 복잡한 명령을 내릴지 모르니까 칩 스스로가 통제하고 판단할 수 있는 '두뇌의 통제력'이 필요하거든.
4/ 근데 테슬라 AI5 같은 전용 ASIC 칩들은 이 제어 장치를 과감하게 칼질함.
왜냐? AI 연산은 결국 '행렬 곱셈'이라는 단순 막노동의 반복이거든. 복잡한 통제관을 다 자르고 그 자리에 순수 노가다꾼(연산 장치)만 빽빽하게 채우니까 칩 크기는 반 토막인데 성능은 똑같이 나오는 거임.
5/ 이제는 모델이 너무 커져서 칩 하나로는 어림도 없음.
수천 수만 개를 묶어서 '하나의 거대한 칩'처럼 돌려야 함.
여기서 엔비디아의 진짜 치트키가 나오는데, 그게 바로 칩끼리 광속으로 소통하게 해주는 고속도로 'NVLink'임.
6/ 경쟁사들이 칩 설계는 따라 해도 이 네트워킹(연결) 기술에서 다 막힘. 그래서 테슬라는 '도조(Dojo)' 컴퓨터를 만들 때 아예 반도체 판때기(웨이퍼)를 쪼개지 않고 통째로 연결하는 무지막지한 방식을 쓰면서 이 장벽을 깨부수려고 몸틀어쥐는 중.
7/ AI 연산은 공부하는 '학습(Training)'이랑 시험 치는 '추론(Inference)'으로 나뉨. 학습은 정답을 찾아가는 과정이라 엄청 정밀한 데이터(FP32/16)랑 대용량 메모리가 필수임.
엔비디아가 독점할 수밖에 없는 영역.
8/ 반면 테슬라 차량이나 로봇에 들어가는 '추론'은 이미 공부 끝난 모델을 굴리는 거라, 데이터 정밀도를 좀 낮춰도(INT8/FP4) 결과에 큰 차이가 없음.
정밀도를 낮추면 칩이 가벼워지고 전력 소모도 확 줄어듦. 테슬라가 '와트당 가성비'에 올인하는 이유임.
9/ 이제 TSMC 3나노, 2나노로 가면서 칩을 더 미세하게 만드는 게 물리적 한계에 부딪힘.
그래서 요즘 대세가 '칩렛(Chiplet)'임. 큰 칩을 통째로 찍으면 불량률이 높으니, 기능별로 작게 쪼개서 만든 뒤 '3D 패키징'으로 레고처럼 입체적으로 쌓아 붙이는 거임.
10/ 엔비디아 블랙웰도 이 기술로 만든 괴물임. 다만 이 기술을 쓰려면 대만 TSMC의 첨단 패키징(CoWoS) 라인을 확보해야 하는데, 지금 전 세계 빅테크들이 이 라인 선점하려고 전쟁 중임.
결국 칩 설계 잘하는 것만큼이나 '누가 TSMC 공장을 쥐고 흔드느냐'가 미래 반도체 패권의 핵심이 될 듯.
3
1,116
Jun 15
Valuing $SPCX at this stage doesn’t make much sense. The success of SpaceX’s whole future business hinges on Starship succeeding.
In many ways, they’re still essentially a startup.
7
1
19
1,334
Jun 15
After reading Citrini Research’s piece on ‘From Tokenmaxxing to Token Panic’ and Satya’s long article, my first reaction was that OpenRouter seems like a promising business.
1
380