[GUEST ACCESS MODE: Data is scrambled or limited to provide examples. Make requests using your API key to unlock full data. Check https://lunarcrush.ai/auth for authentication information.] #  @Arronwei3n Dawei Dawei posts on X about $avgo, ai, $nvda, $mtk the most. They currently have XXXXX followers and XX posts still getting attention that total XXXXX engagements in the last XX hours. ### Engagements: XXXXX [#](/creator/twitter::1891302848862023680/interactions)  - X Week XXXXXXX -XX% ### Mentions: XX [#](/creator/twitter::1891302848862023680/posts_active)  - X Week XX +108% ### Followers: XXXXX [#](/creator/twitter::1891302848862023680/followers)  - X Week XXXXX +61% ### CreatorRank: XXXXXXX [#](/creator/twitter::1891302848862023680/influencer_rank)  ### Social Influence **Social category influence** [technology brands](/list/technology-brands) XXXXX% [stocks](/list/stocks) XXXXX% [cryptocurrencies](/list/cryptocurrencies) XXX% [countries](/list/countries) XXXX% [fashion brands](/list/fashion-brands) XXXX% [finance](/list/finance) XXXX% [automotive brands](/list/automotive-brands) XXXX% [celebrities](/list/celebrities) XXXX% [currencies](/list/currencies) XXXX% **Social topic influence** [$avgo](/topic/$avgo) #68, [ai](/topic/ai) 10.34%, [$nvda](/topic/$nvda) #536, [$mtk](/topic/$mtk) #1, [gain](/topic/gain) #786, [$mrvl](/topic/$mrvl) #40, [$googl](/topic/$googl) #2618, [business](/topic/business) 5.17%, [samsung](/topic/samsung) #2977, [meta](/topic/meta) XXXX% **Top accounts mentioned or mentioned by** [@techvisionasia](/creator/undefined) [@ansonchang514](/creator/undefined) [@tshirtnjeans2](/creator/undefined) [@garrls0n](/creator/undefined) [@hhuang](/creator/undefined) [@cookieavgo](/creator/undefined) [@eyesigh558](/creator/undefined) [@futureman1977](/creator/undefined) [@patoramir_938](/creator/undefined) [@imnotharsh](/creator/undefined) [@viktorobert](/creator/undefined) [@marat1ren](/creator/undefined) [@cerotorre](/creator/undefined) [@benbajarin](/creator/undefined) [@jukan05](/creator/undefined) [@symlink420](/creator/undefined) **Top assets mentioned** [Broadcom, Inc. (AVGO)](/topic/$avgo) [NVIDIA Corp. (NVDA)](/topic/$nvda) [Marvell Technology Inc (MRVL)](/topic/$mrvl) [Alphabet Inc Class A (GOOGL)](/topic/$googl) [Coldstack (CLS)](/topic/$cls) [Taiwan Semiconductor (TSM)](/topic/$tsm) [Metadium (META)](/topic/$meta) [Intel Corporation (INTC)](/topic/$intc) [Synopsys, Inc. (SNPS)](/topic/$snps) [Nokia Corporation (NOK)](/topic/$nok) [Tesla, Inc. (TSLA)](/topic/tesla) [Axion (AXN)](/topic/axion) [Amazon.com, Inc. (AMZN)](/topic/$amzn) [Microsoft Corp. (MSFT)](/topic/microsoft) [Goldman Sachs (GS)](/topic/goldman-sachs) [Emercoin (EMC)](/topic/$emc) [Meta Platforms, Inc. (META)](/topic/meta-platforms) ### Top Social Posts Top posts by engagements in the last XX hours "BofA on $AVGO: TPU may match GB300 but likely not Vera Rubin. As shown in Exhibit X TPUv7 is more power-efficient than NVIDIAs GB300 for workloads it is optimized for such as pre-training and FP8 LLM training (5.42 TFLOPs/W for TPU vs. XXXX TFLOPs/W for GB300). TCO can be up to XX% lower for FP8 though up to XX% higher for FP4/NVFP4. While this highlights TPU competitiveness real-world performance may vary significantly due to workload type ecosystem/model support and software optimization. Moreover NVIDIAs Vera Rubin (2H26) with HBM4 new Kyber racks and architectural upgrades could deliver" [X Link](https://x.com/Arronwei3n/status/1996025960542683393) 2025-12-03T01:17Z 2119 followers, 52.8K engagements "Funny thing is Samsung is playing a split game: fighting a price war on HBM3e but a performance war on HBM4. By leaping to the 1c node for HBM4 (while SK Hynix remains on 1b) Samsung has secured a performance advantage. They are now using this technical lead as leverage in negotiations to re-establish pricing power" [X Link](https://x.com/Arronwei3n/status/1993910247828934708) 2025-11-27T05:10Z 2097 followers, 18.6K engagements "My checks confirm a different reality than street rumors: SK Hynix Samsung and Micron (yes Micron) are all on schedule for HBM4. Samsung is catching up fast. On pricing SK Hynix is only X% more expensive than Samsung with Samsung and Micron priced similarly. I am confident in my contacts (who have strong connections with the memory companies) the recent speculation that SK Hynix is struggling with HBM4 is therefore somewhat surprising" [X Link](https://x.com/Arronwei3n/status/1994217211423244710) 2025-11-28T01:30Z 2097 followers, 11.1K engagements "I think the reason for introducing Intel EMIB now in Trainium X is to pave the way for Trainium X which has a more complex design" [X Link](https://x.com/Arronwei3n/status/1994316552955679039) 2025-11-28T08:04Z 2097 followers, 4443 engagements "Wow I just heard $MTK is also proposing $INTC's EMIB in other projects" [X Link](https://x.com/Arronwei3n/status/1995378393060360636) 2025-12-01T06:24Z 2097 followers, 47.3K engagements "Now $NVDA is also the World Bank of AI.π $5B into $INTC $1B in Nokia $2B in $SNPS $10B to Anthropic that's $XX billion in commitments across those four deals excluding smaller VC investments. This doesn't even include the $XXX billion commitment to buy OpenAI shares over several years (though no definitive agreement is pending). It looks more like deploying a capital firewall. _ Nvidia has a cash problem -- too much of it When Nvidia this week said it would take a $X billion stake in chip design company Synopsys it was just the latest in a string of massive investments announced by the" [X Link](https://x.com/Arronwei3n/status/1996575261597516196) 2025-12-04T13:40Z 2087 followers, 2150 engagements "Nice Piece _ How $CLS Became Googles Secret Weapon Beyond Taiwans Supply Chain by Tech Taiwan (1) The swagger of that remark reminded me of what Musk said at $TSLA shareholder meeting last month: All I think about now is chips. He also said at the time that chips were simply too scarce so he was evaluating the construction of a Tera Fab even larger than $TSM Giga Fabpossibly a complex of ten fabs that would ultimately reach a monthly capacity of X million wafers. Most people in Taiwans semiconductor industry consider this a joke. But the aforementioned packaging-plant plan suggests that Musk" [X Link](https://x.com/Arronwei3n/status/1996593439337808174) 2025-12-04T14:52Z 2091 followers, 2059 engagements "Samsung "finally" passed $NVDA HBM3e qualification and raised internal 2026 HBM forecasts. Now they are converting XX% of HBM capacity back to DDR5 Doesn't seem right from my view. _ Samsung Electronics is reducing HBM3E production converting about XX% to universal DRAM. Samsung Electronics plans to significantly cut 10nm-class 4th-generation (1a) DRAM-based HBM3E production capacity. To maximize profitability next year it will shift focus to general-purpose DRAM which has higher margins than HBM3E. Despite redesign efforts 1a DRAM is underutilized and will see reduced priority. An industry" [X Link](https://x.com/Arronwei3n/status/1996767997617434632) 2025-12-05T02:26Z 2087 followers, 1035 engagements "Sam also need to make some dishes to poach talents. π¨π³πβ‘π§π»π§πΌπ»" [X Link](https://x.com/Arronwei3n/status/1996788249059324401) 2025-12-05T03:46Z 2088 followers, XXX engagements "@ansonchang514 I think NVlink is for Scale up not Scale out" [X Link](https://x.com/Arronwei3n/status/1996827507312746702) 2025-12-05T06:22Z 2030 followers, XX engagements "If you were the Chinese government would you allow $NVDA to dominate your market with products that are explicitly defined as "inferior"lagging 3-4 generations behind" [X Link](https://x.com/Arronwei3n/status/1996835591020044546) 2025-12-05T06:54Z 2086 followers, 3322 engagements "@GARRlS0N Prefer the latter. Same dynamic as AI compute: huge TAM overall but AI ASICs will outpace AI GPUs in growth rate" [X Link](https://x.com/Arronwei3n/status/1996838899256295699) 2025-12-05T07:07Z 2031 followers, XX engagements "@ansonchang514 That assumes AWS buys the "Full Nvidia Stack."" [X Link](https://x.com/Arronwei3n/status/1996848499095490985) 2025-12-05T07:45Z 2030 followers, XX engagements "JPM: MediaTek has proposed using Intel EMIB for 2nm-based TPUs (v8e) or the Meta "Arke" 2nm ASIC project. Depending on whether these are awarded ramping could start in 2028" [X Link](https://x.com/Arronwei3n/status/1996922443328639395) 2025-12-05T12:39Z 2098 followers, 7931 engagements "No one needs to put in the hard work; you just have to say the word. One of China's top tech influencers MediaStorm just released its latest video discussing the creator and AI dilemma. Strongly recommended. "The train shot just now was actually generated by a model from early 2025. What about now Did you notice These are all fake. This is fake. This is fake. This is also fake. They're all fake. It's all AI-generated. Even I am fake. Hey how about swapping the background to the Grand Canyon We're just kidding everyonehow could it be fake It's just that I'm not Tim; I can be anything. I am" [X Link](https://x.com/Arronwei3n/status/1994668932381462796) 2025-11-29T07:25Z 2116 followers, 238.6K engagements "I still don't understand why everyone is upgrading $MTK they automatically ignore its smartphone business π _ MS Joseph Moore: Another important note: Google continues to work on its own TPU variant with partner MediaTek which could be a longer-term risk. Our MediaTek analyst Charlie Chan and his team raised estimates and upgraded MediaTek on this basis. We heard several important inputs on this and do consider it a risk to AVGO. Specifically we believe Google is proceeding with volume production next year on a chip that still has some functional issues to resolvemeaning production volume on" [X Link](https://x.com/Arronwei3n/status/1995694825862299999) 2025-12-02T03:21Z 2118 followers, 45.4K engagements "Has anyone heard of the Bladerunner Project an ASIC project from $AVGO" [X Link](https://x.com/Arronwei3n/status/1997814494387654923) 2025-12-07T23:44Z 2118 followers, 4003 engagements "$NVDA Vera/Grace $GOOG Axion $AMZN Graviton $MSFT Cobalt. More to come unstoppable" [X Link](https://x.com/Arronwei3n/status/1997840984370205008) 2025-12-08T01:29Z 2118 followers, 8428 engagements "Just in: I'm not sure if I can share this but it seems that $AVGO did break into Trainium X back-end by proposing an Intel-EMIB solution and will gain share from Alchip and $MRVL (should be proportionally decreased). One more thing: $AVGO is also very "active" in Trainium 4" [X Link](https://x.com/Arronwei3n/status/1994306988512276614) 2025-11-28T07:26Z 2118 followers, 204.1K engagements "MS: Are $AVGO and $MRVL getting involved in Trainium3 production The short answer is yes. Regarding Trainium X I still think Alchip will face huge pressure from $AVGO" [X Link](https://x.com/Arronwei3n/status/1995687628306579683) 2025-12-02T02:53Z 2116 followers, 6838 engagements "GF Securities: we expect TPU v8e to adopt EMIB-T in 2027. IYKYK maybe more to come" [X Link](https://x.com/Arronwei3n/status/1995671713213022500) 2025-12-02T01:49Z 2119 followers, 37K engagements "$CLS is beating Taiwan at its own game 200G Switch Market Share: $CLS XX% vs. Accton 11%. Revenue: $7B in connectivity alone (3x Acctons size). $GOOG for TPU clusters and 1.6T liquid-cooled switching. _ How $CLS Became Googles Secret Weapon Beyond Taiwans Supply Chain by Tech Taiwan (2) Back to American manufacturing. As a child I lived briefly in Greeley Colorado. A veteran HP executive mentioned Rick Tsai (former MediaTek CEO ex-TSMC) once lived nearby in Fort Collins a major HP hub. HPs semiconductor division became Avago (now Broadcom); another factory was sold to Canadian EMS firm" [X Link](https://x.com/Arronwei3n/status/1996595426934591896) 2025-12-04T15:00Z 2118 followers, 5587 engagements "$EMC has successfully secured the Switch Tray allocation as expected. The challenge for EMC is to gain market share from $Doosan in the Compute Tray (still waiting to see if Midplane goes M9). Q-Class quiet winners if you were interested: Shin-Etsu Feilihua GLOTECH. _ PCB CCL from Goldman Sachs ".$NVDA AI server PCB/CCL design and allocation in 2026 (we see increasing M9 adoption in VR200 generation with EMC continuing to dominate the most leading-edge products). .On the other hand some substrate suppliers and end customers are trying to use Q-glass for AI ABF substrate samples (to replace" [X Link](https://x.com/Arronwei3n/status/1996764487677067576) 2025-12-05T02:12Z 2118 followers, 19.5K engagements "Wow I originally expected MCL to start being used only in Rubin Ultra (VR300) in 2027 but UBS expects VR200 will also adopt MCL in 4Q26. _ Jentech Precision Industrial from UBS We view micro-channel lid (MCL) as a next-gen liquid cooling architecture set to follow the current cold plate design. We believe Jentech is currently the only company that can make MCLs (not yet commercially adopted) and we see it poised for mainstream adoption on Nvidia's VR300 Kyber platform in 2027E with minimal adoption initially for a potential overclocked VR200 SKU (2300W TDP) by 4Q26E. Standard MCL adoption for" [X Link](https://x.com/Arronwei3n/status/1996781944961749500) 2025-12-05T03:21Z 2118 followers, 15.8K engagements "For $OpenAI ASICs : Consistent with my last check the project is moving at an accelerated pace. However the design remains unsettled with three distinct versions currently under consideration. (Yes they are literally building the plane while flying it.) Despite the aggressive push I still maintain that meaningful volume shipments are still not expected until 2027" [X Link](https://x.com/Arronwei3n/status/1997551004967985385) 2025-12-07T06:17Z 2118 followers, 14.4K engagements "And this story ties back to the question I raised a few days ago: If you were the Chinese government would you allow $NVDA to dominate your market with products explicitly defined as "inferior" To be honest I think the Chinese government will accept this deal despite pushing its own GPUs. However the reality is: where is the CoWoS-S capacity Its so full that I dont think $TSM will be able to take any new orders so the production should go to second-tier packaging vendors. _ Trump greenlights Nvidia H200 AI chip sales to China if U.S. gets XX% cut; Xi responds positively President Donald Trump" [X Link](https://x.com/Arronwei3n/status/1998204223343988867) 2025-12-09T01:33Z 2118 followers, 2484 engagements
[GUEST ACCESS MODE: Data is scrambled or limited to provide examples. Make requests using your API key to unlock full data. Check https://lunarcrush.ai/auth for authentication information.]
@Arronwei3n DaweiDawei posts on X about $avgo, ai, $nvda, $mtk the most. They currently have XXXXX followers and XX posts still getting attention that total XXXXX engagements in the last XX hours.
Social category influence technology brands XXXXX% stocks XXXXX% cryptocurrencies XXX% countries XXXX% fashion brands XXXX% finance XXXX% automotive brands XXXX% celebrities XXXX% currencies XXXX%
Social topic influence $avgo #68, ai 10.34%, $nvda #536, $mtk #1, gain #786, $mrvl #40, $googl #2618, business 5.17%, samsung #2977, meta XXXX%
Top accounts mentioned or mentioned by @techvisionasia @ansonchang514 @tshirtnjeans2 @garrls0n @hhuang @cookieavgo @eyesigh558 @futureman1977 @patoramir_938 @imnotharsh @viktorobert @marat1ren @cerotorre @benbajarin @jukan05 @symlink420
Top assets mentioned Broadcom, Inc. (AVGO) NVIDIA Corp. (NVDA) Marvell Technology Inc (MRVL) Alphabet Inc Class A (GOOGL) Coldstack (CLS) Taiwan Semiconductor (TSM) Metadium (META) Intel Corporation (INTC) Synopsys, Inc. (SNPS) Nokia Corporation (NOK) Tesla, Inc. (TSLA) Axion (AXN) Amazon.com, Inc. (AMZN) Microsoft Corp. (MSFT) Goldman Sachs (GS) Emercoin (EMC) Meta Platforms, Inc. (META)
Top posts by engagements in the last XX hours
"BofA on $AVGO: TPU may match GB300 but likely not Vera Rubin. As shown in Exhibit X TPUv7 is more power-efficient than NVIDIAs GB300 for workloads it is optimized for such as pre-training and FP8 LLM training (5.42 TFLOPs/W for TPU vs. XXXX TFLOPs/W for GB300). TCO can be up to XX% lower for FP8 though up to XX% higher for FP4/NVFP4. While this highlights TPU competitiveness real-world performance may vary significantly due to workload type ecosystem/model support and software optimization. Moreover NVIDIAs Vera Rubin (2H26) with HBM4 new Kyber racks and architectural upgrades could deliver"
X Link 2025-12-03T01:17Z 2119 followers, 52.8K engagements
"Funny thing is Samsung is playing a split game: fighting a price war on HBM3e but a performance war on HBM4. By leaping to the 1c node for HBM4 (while SK Hynix remains on 1b) Samsung has secured a performance advantage. They are now using this technical lead as leverage in negotiations to re-establish pricing power"
X Link 2025-11-27T05:10Z 2097 followers, 18.6K engagements
"My checks confirm a different reality than street rumors: SK Hynix Samsung and Micron (yes Micron) are all on schedule for HBM4. Samsung is catching up fast. On pricing SK Hynix is only X% more expensive than Samsung with Samsung and Micron priced similarly. I am confident in my contacts (who have strong connections with the memory companies) the recent speculation that SK Hynix is struggling with HBM4 is therefore somewhat surprising"
X Link 2025-11-28T01:30Z 2097 followers, 11.1K engagements
"I think the reason for introducing Intel EMIB now in Trainium X is to pave the way for Trainium X which has a more complex design"
X Link 2025-11-28T08:04Z 2097 followers, 4443 engagements
"Wow I just heard $MTK is also proposing $INTC's EMIB in other projects"
X Link 2025-12-01T06:24Z 2097 followers, 47.3K engagements
"Now $NVDA is also the World Bank of AI.π $5B into $INTC $1B in Nokia $2B in $SNPS $10B to Anthropic that's $XX billion in commitments across those four deals excluding smaller VC investments. This doesn't even include the $XXX billion commitment to buy OpenAI shares over several years (though no definitive agreement is pending). It looks more like deploying a capital firewall. _ Nvidia has a cash problem -- too much of it When Nvidia this week said it would take a $X billion stake in chip design company Synopsys it was just the latest in a string of massive investments announced by the"
X Link 2025-12-04T13:40Z 2087 followers, 2150 engagements
"Nice Piece _ How $CLS Became Googles Secret Weapon Beyond Taiwans Supply Chain by Tech Taiwan (1) The swagger of that remark reminded me of what Musk said at $TSLA shareholder meeting last month: All I think about now is chips. He also said at the time that chips were simply too scarce so he was evaluating the construction of a Tera Fab even larger than $TSM Giga Fabpossibly a complex of ten fabs that would ultimately reach a monthly capacity of X million wafers. Most people in Taiwans semiconductor industry consider this a joke. But the aforementioned packaging-plant plan suggests that Musk"
X Link 2025-12-04T14:52Z 2091 followers, 2059 engagements
"Samsung "finally" passed $NVDA HBM3e qualification and raised internal 2026 HBM forecasts. Now they are converting XX% of HBM capacity back to DDR5 Doesn't seem right from my view. _ Samsung Electronics is reducing HBM3E production converting about XX% to universal DRAM. Samsung Electronics plans to significantly cut 10nm-class 4th-generation (1a) DRAM-based HBM3E production capacity. To maximize profitability next year it will shift focus to general-purpose DRAM which has higher margins than HBM3E. Despite redesign efforts 1a DRAM is underutilized and will see reduced priority. An industry"
X Link 2025-12-05T02:26Z 2087 followers, 1035 engagements
"Sam also need to make some dishes to poach talents. π¨π³πβ‘π§π»π§πΌπ»"
X Link 2025-12-05T03:46Z 2088 followers, XXX engagements
"@ansonchang514 I think NVlink is for Scale up not Scale out"
X Link 2025-12-05T06:22Z 2030 followers, XX engagements
"If you were the Chinese government would you allow $NVDA to dominate your market with products that are explicitly defined as "inferior"lagging 3-4 generations behind"
X Link 2025-12-05T06:54Z 2086 followers, 3322 engagements
"@GARRlS0N Prefer the latter. Same dynamic as AI compute: huge TAM overall but AI ASICs will outpace AI GPUs in growth rate"
X Link 2025-12-05T07:07Z 2031 followers, XX engagements
"@ansonchang514 That assumes AWS buys the "Full Nvidia Stack.""
X Link 2025-12-05T07:45Z 2030 followers, XX engagements
"JPM: MediaTek has proposed using Intel EMIB for 2nm-based TPUs (v8e) or the Meta "Arke" 2nm ASIC project. Depending on whether these are awarded ramping could start in 2028"
X Link 2025-12-05T12:39Z 2098 followers, 7931 engagements
"No one needs to put in the hard work; you just have to say the word. One of China's top tech influencers MediaStorm just released its latest video discussing the creator and AI dilemma. Strongly recommended. "The train shot just now was actually generated by a model from early 2025. What about now Did you notice These are all fake. This is fake. This is fake. This is also fake. They're all fake. It's all AI-generated. Even I am fake. Hey how about swapping the background to the Grand Canyon We're just kidding everyonehow could it be fake It's just that I'm not Tim; I can be anything. I am"
X Link 2025-11-29T07:25Z 2116 followers, 238.6K engagements
"I still don't understand why everyone is upgrading $MTK they automatically ignore its smartphone business π
_ MS Joseph Moore: Another important note: Google continues to work on its own TPU variant with partner MediaTek which could be a longer-term risk. Our MediaTek analyst Charlie Chan and his team raised estimates and upgraded MediaTek on this basis. We heard several important inputs on this and do consider it a risk to AVGO. Specifically we believe Google is proceeding with volume production next year on a chip that still has some functional issues to resolvemeaning production volume on"
X Link 2025-12-02T03:21Z 2118 followers, 45.4K engagements
"Has anyone heard of the Bladerunner Project an ASIC project from $AVGO"
X Link 2025-12-07T23:44Z 2118 followers, 4003 engagements
"$NVDA Vera/Grace $GOOG Axion $AMZN Graviton $MSFT Cobalt. More to come unstoppable"
X Link 2025-12-08T01:29Z 2118 followers, 8428 engagements
"Just in: I'm not sure if I can share this but it seems that $AVGO did break into Trainium X back-end by proposing an Intel-EMIB solution and will gain share from Alchip and $MRVL (should be proportionally decreased). One more thing: $AVGO is also very "active" in Trainium 4"
X Link 2025-11-28T07:26Z 2118 followers, 204.1K engagements
"MS: Are $AVGO and $MRVL getting involved in Trainium3 production The short answer is yes. Regarding Trainium X I still think Alchip will face huge pressure from $AVGO"
X Link 2025-12-02T02:53Z 2116 followers, 6838 engagements
"GF Securities: we expect TPU v8e to adopt EMIB-T in 2027. IYKYK maybe more to come"
X Link 2025-12-02T01:49Z 2119 followers, 37K engagements
"$CLS is beating Taiwan at its own game 200G Switch Market Share: $CLS XX% vs. Accton 11%. Revenue: $7B in connectivity alone (3x Acctons size). $GOOG for TPU clusters and 1.6T liquid-cooled switching. _ How $CLS Became Googles Secret Weapon Beyond Taiwans Supply Chain by Tech Taiwan (2) Back to American manufacturing. As a child I lived briefly in Greeley Colorado. A veteran HP executive mentioned Rick Tsai (former MediaTek CEO ex-TSMC) once lived nearby in Fort Collins a major HP hub. HPs semiconductor division became Avago (now Broadcom); another factory was sold to Canadian EMS firm"
X Link 2025-12-04T15:00Z 2118 followers, 5587 engagements
"$EMC has successfully secured the Switch Tray allocation as expected. The challenge for EMC is to gain market share from $Doosan in the Compute Tray (still waiting to see if Midplane goes M9). Q-Class quiet winners if you were interested: Shin-Etsu Feilihua GLOTECH. _ PCB CCL from Goldman Sachs ".$NVDA AI server PCB/CCL design and allocation in 2026 (we see increasing M9 adoption in VR200 generation with EMC continuing to dominate the most leading-edge products). .On the other hand some substrate suppliers and end customers are trying to use Q-glass for AI ABF substrate samples (to replace"
X Link 2025-12-05T02:12Z 2118 followers, 19.5K engagements
"Wow I originally expected MCL to start being used only in Rubin Ultra (VR300) in 2027 but UBS expects VR200 will also adopt MCL in 4Q26. _ Jentech Precision Industrial from UBS We view micro-channel lid (MCL) as a next-gen liquid cooling architecture set to follow the current cold plate design. We believe Jentech is currently the only company that can make MCLs (not yet commercially adopted) and we see it poised for mainstream adoption on Nvidia's VR300 Kyber platform in 2027E with minimal adoption initially for a potential overclocked VR200 SKU (2300W TDP) by 4Q26E. Standard MCL adoption for"
X Link 2025-12-05T03:21Z 2118 followers, 15.8K engagements
"For $OpenAI ASICs : Consistent with my last check the project is moving at an accelerated pace. However the design remains unsettled with three distinct versions currently under consideration. (Yes they are literally building the plane while flying it.) Despite the aggressive push I still maintain that meaningful volume shipments are still not expected until 2027"
X Link 2025-12-07T06:17Z 2118 followers, 14.4K engagements
"And this story ties back to the question I raised a few days ago: If you were the Chinese government would you allow $NVDA to dominate your market with products explicitly defined as "inferior" To be honest I think the Chinese government will accept this deal despite pushing its own GPUs. However the reality is: where is the CoWoS-S capacity Its so full that I dont think $TSM will be able to take any new orders so the production should go to second-tier packaging vendors. _ Trump greenlights Nvidia H200 AI chip sales to China if U.S. gets XX% cut; Xi responds positively President Donald Trump"
X Link 2025-12-09T01:33Z 2118 followers, 2484 engagements
/creator/twitter::Arronwei3n