NeutralAVGOBroadcomBroadcom000660SK hynixSK hynix005930Samsung ElectronicsSamsung ElectronicsASICAtegrity SpecialtyAtegrity SpecialtyGOOGLAlphabetAlphabetNVDANVIDIANVIDIAORCLOracleOracle

Today on

- $GOOGL has expanded its TPU customer base to nearly 100. - $ORCL plans to offer Google TPUs through OCI starting in 2027. - Counterpoint Research forecasts a 35-fold increase in HBM bit demand for AI ASICs from 2024 to 2028. - Average HBM capacity per AI ASIC is projected to increase approximately fivefold over the same period. - Samsung Electronics ($005930) and SK hynix ($000660) reportedly supply HBM for Google TPUs. - $AVGO maintains a long-term agreement to develop and supply Google's next-generation TPUs through 2031. - Reports suggest $GOOGL could expand direct HBM procurement, but the scale of the transition and a 15–20% cost reduction target remain unconfirmed. > TPU commercialization diversifies HBM demand beyond $NVDA, creating new growth opportunities for $005930 and $000660. > Direct HBM procurement could strengthen $GOOGL's purchasing leverage while putting pressure on memory suppliers' ASPs and margins. > The key distinction is between reducing $AVGO's procurement role and replacing its TPU design partnership. > Investors should monitor not only TPU shipments but also HBM capacity per TPU and changes in supplier market share.

AVGOBroadcomNeutralBroadcom

Today on

- $AVGO showcased AI networking solutions at SC26 conference - Focus on open standards and high-performance connectivity for data center workloads > $AVGO is positioning its networking portfolio as a strategic enabler for AI infrastructure, emphasizing interoperability and performance. > This reflects the company's effort to capture demand in the rapidly growing AI data center market segment.

NeutralAVGOBroadcomBroadcomAMDGOOGLAlphabetAlphabetMRVLMarvellMarvellNVDANVIDIANVIDIA

Yesterday on

Google is becoming a serious AI infrastructure contender. $GOOGL Today, $AMD announced that its EPYC CPUs will power Google's TPU 8tx host infrastructure. But what about $AVGO? Broadcom remains a key partner, with agreements covering future TPU generations and networking components through 2031. And $MRVL is joining the broader custom silicon ecosystem. Google isn't just building an alternative to NVIDIA. It's building an entire AI semiconductor ecosystem. $NVDA's most interesting long-term competitor might be $GOOGL itself.

NeutralAVGOBroadcomBroadcomANETArista NetworksArista NetworksCOHRCoherent Corp.Coherent Corp.LITELumentum Holdings Inc.Lumentum Holdings Inc.MRVLMarvellMarvellNVDANVIDIANVIDIA

3d ago on

NVIDIA is reportedly exploring old telecom buildings as AI data centers. Think about what that means. The bottleneck is no longer just GPU performance. It's power, permitting, cooling, connectivity, and time to deployment. Repurposing existing telecom infrastructure could reduce some of those barriers. Now imagine AI capacity spreading across smaller, distributed facilities instead of only massive campuses. What happens to optical transport, networking architectures, and interconnect demand? $NVDA $ANET $AVGO $MRVL $LITE $COHR The infrastructure surrounding the GPUs may be just as interesting as the GPUs themselves.