[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, $nvda, ai, gain the most. They currently have XXXXX followers and XX posts still getting attention that total XXXXXX engagements in the last XX hours. ### Engagements: XXXXXX [#](/creator/twitter::1891302848862023680/interactions)  - X Week XXXXXXX -XX% ### Mentions: XX [#](/creator/twitter::1891302848862023680/posts_active)  - X Week XX +48% ### Followers: XXXXX [#](/creator/twitter::1891302848862023680/followers)  - X Week XXXXX +24% ### CreatorRank: XXXXXXX [#](/creator/twitter::1891302848862023680/influencer_rank)  ### Social Influence **Social category influence** [stocks](/list/stocks) #4901 [technology brands](/list/technology-brands) XXXXX% [cryptocurrencies](/list/cryptocurrencies) XXXX% [fashion brands](/list/fashion-brands) XXXX% [finance](/list/finance) XXXX% [countries](/list/countries) XXXX% **Social topic influence** [$avgo](/topic/$avgo) #97, [$nvda](/topic/$nvda) #149, [ai](/topic/ai) #7041, [gain](/topic/gain) 7.35%, [meta](/topic/meta) #4011, [$mrvl](/topic/$mrvl) 7.35%, [$mtk](/topic/$mtk) #1, [$tsm](/topic/$tsm) #39, [avgo](/topic/avgo) #119, [$googl](/topic/$googl) #707 **Top accounts mentioned or mentioned by** [@tshirtnjeans2](/creator/undefined) [@techvisionasia](/creator/undefined) [@ansonchang514](/creator/undefined) [@garrls0n](/creator/undefined) [@krokodilv](/creator/undefined) [@agt199](/creator/undefined) [@grok](/creator/undefined) [@krokodil_v](/creator/undefined) [@inovrmihd](/creator/undefined) [@patoramir_938](/creator/undefined) [@hhuang](/creator/undefined) [@lordhaartjp](/creator/undefined) [@cookieavgo](/creator/undefined) [@imnotharsh](/creator/undefined) [@futureman1977](/creator/undefined) [@martijnde_boer](/creator/undefined) [@eyesigh558](/creator/undefined) [@lord_haart_jp](/creator/undefined) [@viktorobert](/creator/undefined) [@cerotorre](/creator/undefined) **Top assets mentioned** [Broadcom, Inc. (AVGO)](/topic/$avgo) [NVIDIA Corp. (NVDA)](/topic/$nvda) [Marvell Technology Inc (MRVL)](/topic/$mrvl) [Taiwan Semiconductor (TSM)](/topic/$tsm) [Alphabet Inc Class A (GOOGL)](/topic/$googl) [Intel Corporation (INTC)](/topic/$intc) [Microsoft Corp. (MSFT)](/topic/microsoft) [Metadium (META)](/topic/$meta) [Goldman Sachs (GS)](/topic/goldman-sachs) [Emercoin (EMC)](/topic/$emc) [Morgan Stanley (MS)](/topic/morgan-stanley) [Synopsys, Inc. (SNPS)](/topic/$snps) [Nokia Corporation (NOK)](/topic/$nok) [Coldstack (CLS)](/topic/$cls) [Meta Platforms, Inc. (META)](/topic/meta-platforms) [Axion (AXN)](/topic/axion) [Amazon.com, Inc. (AMZN)](/topic/$amzn) [Oracle Corporation (ORCL)](/topic/$orcl) ### Top Social Posts Top posts by engagements in the last XX hours "$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 2293 followers, 19.8K engagements "Jentech Precision Industrial from Daiwa What is a microchannel lid (MCL) solution MCL integrates the functions of a traditional heat spreader and cold plate into a single component directly attached to the chip package. It consists of a copper base plate with engraved microchannels topped by a cover plate. Manifolds distribute liquid coolant into and out of the microchannels. This shortens the heat dissipation path from die TIM1 lid TIM2 cold plate to just die TIM1 MCL. By eliminating TIM2 and the heat spreader/cold plate interface MCL minimizes thermal resistance and improves cooling" [X Link](https://x.com/Arronwei3n/status/1998568252390388211) 2025-12-10T01:39Z 2293 followers, 1823 engagements "NVIDIA: CoWoS-L demand upgraded from 590000 to 700000 wafers; B40 downgraded to X million units. Broadcom: CoWoS-S up from 155000 to 200000 wafers; CoWoS-L reduced by 20000 wafers due to Meta ASIC design changes but overall from 210000 to 230000 wafers. MediaTek: CoWoS-S wafers up from 10000 to 20000" [X Link](https://x.com/Arronwei3n/status/1995667324213625192) 2025-12-02T01:32Z 2293 followers, 4012 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 2293 followers, 17.7K engagements "I still stand by my word. From my checks $AVGO will have a big role and is "actively" involved in Trainium 4" [X Link](https://x.com/Arronwei3n/status/1998213419879825779) 2025-12-09T02:09Z 2293 followers, 9319 engagements "While this move reopens the door for U.S. revenue the strategic train has already left the station Neil Shah partner at Counterpoint Research told CNBC on Tuesday. Huang said in a Bloomberg interview in May that Chinese tech giant Huaweis semiconductor products for AI are probably comparable to Nvidias H200. Alibaba and Baidu as well as other Chinese startups are also racing to bring competitive products to Nvidia to the market." _ Nvidia can sell the more advanced H200 AI chip to China but will Beijing want them Nvidia has U.S. government approval to sell its advanced H200 AI chips to China." [X Link](https://x.com/Arronwei3n/status/1998579588520448074) 2025-12-10T02:24Z 2293 followers, 1085 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 2148 followers, 11.1K 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 2197 followers, 204.4K 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 2184 followers, 4449 engagements "Please note that this post was not made in bad faith. I was skeptical (please check my previous posts) when I first saw AYZs post about $AVGO role in the TR supply chain. Then Morgan Stanley and Fubon Research denied the rumors which initially reaffirmed my view that AYZ was wrong this time. However an analyst from Nomura (one of the best semiconductor analysts in my view) also mentioned seeing $AVGO involvement. This made me want to dig deeper and my contact just shared the latest developments. Of course my contact could be wrongit might just be a pilot run or something else (perhaps" [X Link](https://x.com/Arronwei3n/status/1994393982944546966) 2025-11-28T13:12Z 2198 followers, 9826 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 2213 followers, 6978 engagements "Now $NVDA is also the World Bank of AI.😆 $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 $X billion stake in chip design company Synopsys it was just the latest in a string of massive investments announced by the" [X Link](https://x.com/Arronwei3n/status/1996575261597516196) 2025-12-04T13:40Z 2177 followers, 2167 engagements "$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 2234 followers, 5634 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 2186 followers, 8037 engagements "@Krokodil_V Dual vendor model" [X Link](https://x.com/Arronwei3n/status/1998389903886110867) 2025-12-09T13:50Z 2234 followers, XXX engagements "@Lord_haart_jp Ive been checking it every day and $AVGO involvement doesnt imply that Alchip will be excluded" [X Link](https://x.com/Arronwei3n/status/1998303079331180661) 2025-12-09T08:05Z 2244 followers, XXX engagements "Is $MSFT Griffin endangered too If Maia 300(Griffin) face any delayed issues I am inclined to believe $MSFT will buy more $NVDA GPUs rather then putting resources in ASICs instead" [X Link](https://x.com/Arronwei3n/status/1997181484126687742) 2025-12-06T05:49Z 2278 followers, 6438 engagements "For $OpenAI ASICs : Consistent with my last check the project is moving at an accelerated pace. However the design remains unsettled with three distinct versions currently under consideration. (Yes they are literally building the plane while flying it.) Despite the aggressive push I still maintain that meaningful volume shipments are still not expected until 2027" [X Link](https://x.com/Arronwei3n/status/1997551004967985385) 2025-12-07T06:17Z 2261 followers, 14.6K engagements "$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 2281 followers, 8626 engagements "@Krokodil_V Please check my above comment that involvement of $AVGO does that mean the rule out of Alchip" [X Link](https://x.com/Arronwei3n/status/1998389708754395240) 2025-12-09T13:50Z 2262 followers, XXX 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 2292 followers, 238.7K 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 2289 followers, 47.5K 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 2292 followers, 37.3K engagements "I still don't understand why everyone is upgrading $MTK they automatically ignore its smartphone business 😅 _ 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 2289 followers, 46.3K engagements "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 2287 followers, 53K engagements "I was not aware of this report. Turns out my contact is accurate. CoPoS is coming stay tuned _ JPM: Panel-level / CoPoS $TSM in the drivers seat ASE and PTIs progress uncertain. Panel-level pilot lines at Visera. We believe TSMC should complete its engineering/pilot line for 310x310mm panel-level packaging (CoPoS) by early 2026 with more concrete adoption timeline decisions in 2H26. NVDA is likely the first customer to adopt CoPoS in 2028 (with Feynman) assuming technology maturity and manufacturing hurdles are overcome. Equipment vendors should start receiving CoPoS capacity buildout orders" [X Link](https://x.com/Arronwei3n/status/1996925104547389743) 2025-12-05T12:50Z 2293 followers, 53.2K engagements "Quick question: If $TSM CoPoS is on track to MP in 2027 or EMIB fall short of expectations will those ASICs proposal to $INTC EMIB be switch back to $TSM CoPoS I will keep monitoring the dynamics.👀" [X Link](https://x.com/Arronwei3n/status/1997237858210631874) 2025-12-06T09:33Z 2293 followers, 28.5K 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 2292 followers, 4098 engagements "@TShirtnJeans2 The first thing you have to understand is that it is $TSM not $NVDA that allocates CoWoS capacity to each client and they know the market better than anyone. ( $NVDA always overbooks and $TSM will decide the amount it granted to $NVDA )" [X Link](https://x.com/Arronwei3n/status/1998201476930470251) 2025-12-09T01:22Z 2293 followers, XX engagements "From my recent supply chain check $AVGO will secure 230K CoWoS wafers from $TSM (in line with my earlier forecast check the comment). I respectfully disagree with Fubon Researchs lower estimate. I believe 3.8M TPU units in 2026 is highly achievable. Of the 230K 3040K will go to merchant networking (Tomahawk/Jericho) leaving 190200K for TPU 3M units plus 0.4M units from $MTK v7e (see pinned post) plus additional non-TSM CoWoS supply. This puts 3.8M well within reach. The 3.8M target is realistic if second-source packaging capacity scales reliably in 2026 thats the main debate. Happy to hear" [X Link](https://x.com/Arronwei3n/status/1998211466927935893) 2025-12-09T02:01Z 2293 followers, 5647 engagements "GF now expects $NVDA to incorporate CPO into the scale-up networking of one version of its 2027 Rubin Ultra architecture. _ NVIDIAs scale-up optical solutions could arrive earlier than expected: As mentioned in our previous OCP report on Oct XX and our series of AI networking reports Google will interconnect over 9000 chips via OCS for its TPU v7 exceeding the XXX dies NVIDIA is projected to deploy in 2027. We expect NVIDIA may have begun considering CPO introduction starting with Rubin Ultra in 2H27 specifically for scale-up interconnect within the 576-die architecture. Within this" [X Link](https://x.com/Arronwei3n/status/1998570417624322397) 2025-12-10T01:48Z 2293 followers, 81.4K engagements "@AGT199 @grok lol. From GF: LITE (CW laser) Sumitomo (CW laser) and Browave (shuffle). We also believe China local FAU makers will lead and expect optical module/shuffle suppliers to benefit" [X Link](https://x.com/Arronwei3n/status/1998593840803098924) 2025-12-10T03:21Z 2293 followers, 1332 engagements "It should be good news that $ORCL is diversifying its client base. Why is everyone selling "Up XXX% from last year and up $XX billion since the end of August driven by contracts signed with $META $NVDA and others as we continue to diversify our customer backlog."" [X Link](https://x.com/Arronwei3n/status/1998936064107622783) 2025-12-11T02:01Z 2293 followers, 2831 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 2293 followers, 5322 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, $nvda, ai, gain the most. They currently have XXXXX followers and XX posts still getting attention that total XXXXXX engagements in the last XX hours.
Social category influence stocks #4901 technology brands XXXXX% cryptocurrencies XXXX% fashion brands XXXX% finance XXXX% countries XXXX%
Social topic influence $avgo #97, $nvda #149, ai #7041, gain 7.35%, meta #4011, $mrvl 7.35%, $mtk #1, $tsm #39, avgo #119, $googl #707
Top accounts mentioned or mentioned by @tshirtnjeans2 @techvisionasia @ansonchang514 @garrls0n @krokodilv @agt199 @grok @krokodil_v @inovrmihd @patoramir_938 @hhuang @lordhaartjp @cookieavgo @imnotharsh @futureman1977 @martijnde_boer @eyesigh558 @lord_haart_jp @viktorobert @cerotorre
Top assets mentioned Broadcom, Inc. (AVGO) NVIDIA Corp. (NVDA) Marvell Technology Inc (MRVL) Taiwan Semiconductor (TSM) Alphabet Inc Class A (GOOGL) Intel Corporation (INTC) Microsoft Corp. (MSFT) Metadium (META) Goldman Sachs (GS) Emercoin (EMC) Morgan Stanley (MS) Synopsys, Inc. (SNPS) Nokia Corporation (NOK) Coldstack (CLS) Meta Platforms, Inc. (META) Axion (AXN) Amazon.com, Inc. (AMZN) Oracle Corporation (ORCL)
Top posts by engagements in the last XX hours
"$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 2293 followers, 19.8K engagements
"Jentech Precision Industrial from Daiwa What is a microchannel lid (MCL) solution MCL integrates the functions of a traditional heat spreader and cold plate into a single component directly attached to the chip package. It consists of a copper base plate with engraved microchannels topped by a cover plate. Manifolds distribute liquid coolant into and out of the microchannels. This shortens the heat dissipation path from die TIM1 lid TIM2 cold plate to just die TIM1 MCL. By eliminating TIM2 and the heat spreader/cold plate interface MCL minimizes thermal resistance and improves cooling"
X Link 2025-12-10T01:39Z 2293 followers, 1823 engagements
"NVIDIA: CoWoS-L demand upgraded from 590000 to 700000 wafers; B40 downgraded to X million units. Broadcom: CoWoS-S up from 155000 to 200000 wafers; CoWoS-L reduced by 20000 wafers due to Meta ASIC design changes but overall from 210000 to 230000 wafers. MediaTek: CoWoS-S wafers up from 10000 to 20000"
X Link 2025-12-02T01:32Z 2293 followers, 4012 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 2293 followers, 17.7K engagements
"I still stand by my word. From my checks $AVGO will have a big role and is "actively" involved in Trainium 4"
X Link 2025-12-09T02:09Z 2293 followers, 9319 engagements
"While this move reopens the door for U.S. revenue the strategic train has already left the station Neil Shah partner at Counterpoint Research told CNBC on Tuesday. Huang said in a Bloomberg interview in May that Chinese tech giant Huaweis semiconductor products for AI are probably comparable to Nvidias H200. Alibaba and Baidu as well as other Chinese startups are also racing to bring competitive products to Nvidia to the market." _ Nvidia can sell the more advanced H200 AI chip to China but will Beijing want them Nvidia has U.S. government approval to sell its advanced H200 AI chips to China."
X Link 2025-12-10T02:24Z 2293 followers, 1085 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 2148 followers, 11.1K 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 2197 followers, 204.4K 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 2184 followers, 4449 engagements
"Please note that this post was not made in bad faith. I was skeptical (please check my previous posts) when I first saw AYZs post about $AVGO role in the TR supply chain. Then Morgan Stanley and Fubon Research denied the rumors which initially reaffirmed my view that AYZ was wrong this time. However an analyst from Nomura (one of the best semiconductor analysts in my view) also mentioned seeing $AVGO involvement. This made me want to dig deeper and my contact just shared the latest developments. Of course my contact could be wrongit might just be a pilot run or something else (perhaps"
X Link 2025-11-28T13:12Z 2198 followers, 9826 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 2213 followers, 6978 engagements
"Now $NVDA is also the World Bank of AI.😆 $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 $X billion stake in chip design company Synopsys it was just the latest in a string of massive investments announced by the"
X Link 2025-12-04T13:40Z 2177 followers, 2167 engagements
"$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 2234 followers, 5634 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 2186 followers, 8037 engagements
"@Krokodil_V Dual vendor model"
X Link 2025-12-09T13:50Z 2234 followers, XXX engagements
"@Lord_haart_jp Ive been checking it every day and $AVGO involvement doesnt imply that Alchip will be excluded"
X Link 2025-12-09T08:05Z 2244 followers, XXX engagements
"Is $MSFT Griffin endangered too If Maia 300(Griffin) face any delayed issues I am inclined to believe $MSFT will buy more $NVDA GPUs rather then putting resources in ASICs instead"
X Link 2025-12-06T05:49Z 2278 followers, 6438 engagements
"For $OpenAI ASICs : Consistent with my last check the project is moving at an accelerated pace. However the design remains unsettled with three distinct versions currently under consideration. (Yes they are literally building the plane while flying it.) Despite the aggressive push I still maintain that meaningful volume shipments are still not expected until 2027"
X Link 2025-12-07T06:17Z 2261 followers, 14.6K engagements
"$NVDA Vera/Grace $GOOG Axion $AMZN Graviton $MSFT Cobalt. More to come unstoppable"
X Link 2025-12-08T01:29Z 2281 followers, 8626 engagements
"@Krokodil_V Please check my above comment that involvement of $AVGO does that mean the rule out of Alchip"
X Link 2025-12-09T13:50Z 2262 followers, XXX 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 2292 followers, 238.7K engagements
"Wow I just heard $MTK is also proposing $INTC's EMIB in other projects"
X Link 2025-12-01T06:24Z 2289 followers, 47.5K engagements
"GF Securities: we expect TPU v8e to adopt EMIB-T in 2027. IYKYK maybe more to come"
X Link 2025-12-02T01:49Z 2292 followers, 37.3K engagements
"I still don't understand why everyone is upgrading $MTK they automatically ignore its smartphone business 😅 _ 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 2289 followers, 46.3K engagements
"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 2287 followers, 53K engagements
"I was not aware of this report. Turns out my contact is accurate. CoPoS is coming stay tuned _ JPM: Panel-level / CoPoS $TSM in the drivers seat ASE and PTIs progress uncertain. Panel-level pilot lines at Visera. We believe TSMC should complete its engineering/pilot line for 310x310mm panel-level packaging (CoPoS) by early 2026 with more concrete adoption timeline decisions in 2H26. NVDA is likely the first customer to adopt CoPoS in 2028 (with Feynman) assuming technology maturity and manufacturing hurdles are overcome. Equipment vendors should start receiving CoPoS capacity buildout orders"
X Link 2025-12-05T12:50Z 2293 followers, 53.2K engagements
"Quick question: If $TSM CoPoS is on track to MP in 2027 or EMIB fall short of expectations will those ASICs proposal to $INTC EMIB be switch back to $TSM CoPoS I will keep monitoring the dynamics.👀"
X Link 2025-12-06T09:33Z 2293 followers, 28.5K engagements
"Has anyone heard of the Bladerunner Project an ASIC project from $AVGO"
X Link 2025-12-07T23:44Z 2292 followers, 4098 engagements
"@TShirtnJeans2 The first thing you have to understand is that it is $TSM not $NVDA that allocates CoWoS capacity to each client and they know the market better than anyone. ( $NVDA always overbooks and $TSM will decide the amount it granted to $NVDA )"
X Link 2025-12-09T01:22Z 2293 followers, XX engagements
"From my recent supply chain check $AVGO will secure 230K CoWoS wafers from $TSM (in line with my earlier forecast check the comment). I respectfully disagree with Fubon Researchs lower estimate. I believe 3.8M TPU units in 2026 is highly achievable. Of the 230K 3040K will go to merchant networking (Tomahawk/Jericho) leaving 190200K for TPU 3M units plus 0.4M units from $MTK v7e (see pinned post) plus additional non-TSM CoWoS supply. This puts 3.8M well within reach. The 3.8M target is realistic if second-source packaging capacity scales reliably in 2026 thats the main debate. Happy to hear"
X Link 2025-12-09T02:01Z 2293 followers, 5647 engagements
"GF now expects $NVDA to incorporate CPO into the scale-up networking of one version of its 2027 Rubin Ultra architecture. _ NVIDIAs scale-up optical solutions could arrive earlier than expected: As mentioned in our previous OCP report on Oct XX and our series of AI networking reports Google will interconnect over 9000 chips via OCS for its TPU v7 exceeding the XXX dies NVIDIA is projected to deploy in 2027. We expect NVIDIA may have begun considering CPO introduction starting with Rubin Ultra in 2H27 specifically for scale-up interconnect within the 576-die architecture. Within this"
X Link 2025-12-10T01:48Z 2293 followers, 81.4K engagements
"@AGT199 @grok lol. From GF: LITE (CW laser) Sumitomo (CW laser) and Browave (shuffle). We also believe China local FAU makers will lead and expect optical module/shuffle suppliers to benefit"
X Link 2025-12-10T03:21Z 2293 followers, 1332 engagements
"It should be good news that $ORCL is diversifying its client base. Why is everyone selling "Up XXX% from last year and up $XX billion since the end of August driven by contracts signed with $META $NVDA and others as we continue to diversify our customer backlog.""
X Link 2025-12-11T02:01Z 2293 followers, 2831 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 2293 followers, 5322 engagements
/creator/x::Arronwei3n