dot61

Sharing thoughts on tech/biotech with focus on long-term strategy and competitive dynamics. Former tech equity analyst on Wall St. EE & MBA. Free to Follow.

[dot61 @dot_61]

$META Tech designed to make human beings lazier usually wins.

https://t.co/eMGNSSPz7M

[dot61 @dot_61]

SaaS down day coming? Until the debate is well settled (i.e., AI models saturate) stick with compute and AI models names.

"OpenAI’s president, Greg Brockman, said the new model “represents a generational leap in capability” and that it could be defined as AGI."

$NOW $CRM $IGV $AVGO $GOOGL $META

[dot61 @dot_61]

$AVGO Narratives come and go, but share price will follow EPS eventually. Multiple undemanding at $30+ EPS. The notion of higher competition is not well founded. Broadcom has the key components to scale AI compute and memory at rack and cluster level, which no other vendor comes close to.

[dot61 @dot_61]

$AVGO Time to market advantage over other ASIC houses: "Just to reiterate the major technology challenges of developing these complex accelerators. Broadcom is now shipping the TPU version 8i ahead of the MediaTek version v8t, which in fact was initiated earlier."

[dot61 @dot_61]

$AVGO $30+ EPS in F28. 15x-$450, 20x-$600, ..

Wont be a surprise it it runs up 30-35% in the next 1-2 months.

[dot61 @dot_61]

$AVGO Guide is breaking bear thesis on margin compression. AI rev growth higher than $NVDA!

Q4 Guide:
Total Semi Rev: ~$26.1B vs St $25.6B
AI Semi Rev: $21.7B vs St $20.9B
Infra Software $8.7B vs St $9.1B
GM ~73% vs St 72.9% (despite weak SW)

FY27: AI revenue to double to ~$115B

FY28: AI revenue to double again to ~$230B

[dot61 @dot_61]

@sundarpichai Implications are huge: Cheaper iso-intelligence tokens running on vertical TPU and Cloud infra! Not fully reflected in the stock $GOOGL

[dot61 @dot_61]

@semodough Could be a multibagger for patient investors $CTMX.

[dot61 @dot_61]

$AVGO St modeling strong rev growth inflection ~80% yoy in current qtr and ~90% yoy for the guide. That said, the stock is not at all pricing this inflection. If Broadcom reports anywhere close to it, stock likely takes off.

[dot61 @dot_61]

Implications on $GOOGL from OPENAI: CHATGPT ADS HAS REACHED $1B IN ANNUALIZED REV RUN RATE

As a comparison - Google's Q2-26 Search and other rev was $63B!

Despite being the fastest adopted tech ever and ~1B MAU, ChatGPT has limitations: Difficult to change established habits in Search. Lacks compute scale of Google

[dot61 @dot_61]

Is $META's change of stance from open-weights Llama to closed Muse Spark is just temporary? At 75% discount for its developer API vs OAI/Anth, it wont even cover Meta's development cost. Mark Z has already said that Muse 1.2 would be an open weights model. Eventual strategy is to offer license fee free. open models if customers use Meta's AIaaS. This could reduce friction (enterprises are currently being charged by both AI labs and compute providers).

[dot61 @dot_61]

$GOOGL $META Owning your own AI models can drive down costs for 1P workloads and open new revenue streams that is not getting modeled properly. Optionality of a competitive AI model models can drive huge returns down the line ($META, $GOOGL) and that combined with internal custom silicon/compute capacity to serve those models is going to be a very strong moat.

[dot61 @dot_61]

When Anthropic goes public, seems like SaaS names are not the source of funds to chase the IPO. $AMZN and $GOOGL both have a meaningful stake in Anthropic. In a tech-dominated basket, what is the source of funds? $AAPL, $MSFT?

[dot61 @dot_61]

$TEAM when every enterprise starts coding using AI coding agents, vibe coding Jira and Confluence is not on their minds first! Plus, AI is seeding multiple new project in all companies (offsetting seat TAM concerns)

[dot61 @dot_61]

What #OpenAI did with Jalapeno is hugely consequential to ASIC design cos as custom chip design appetite may may increase. ++ $AVGO

[dot61 @dot_61]

#OpenAI Jalapeno inference chip uses Ethernet for scale-out (Broadcom's Ethernet switches) yet seems to have lower latency vs other solutions! Inference chips optimized on latency (bandwidth/W) may not need a propreitary or a new low-latency fabric. $AVGO

[dot61 @dot_61]

If #Anthropic's TAM is $30T, what should $GOOGL and $META with their own models and Si be valued at?

[dot61 @dot_61]

$AVGO Broadcom - Too much pessimism. $GOOGL TPU demand expansion and #OpenAI Jalapeno ramp are not well understood.