Sabitlenmiş Tweet
Make this man CEO $NOW
- Carnegie Mellon CS
- Former VP Google Cloud
- 25 years at Oracle
The time for BS sales guys at the helm is over.
@fred_luddy
English
Action Jackson
4.9K posts
Action Jackson
@AJacksonHole
Fundamentally focused investor. Hold no dogmas. Dad of two boys. Nothing I say is financial advice. https://t.co/t6JeatgLq3
San Francisco, CA Katılım Kasım 2022
1.1K Takip Edilen677 Takipçiler
Every single AI startup with $10B+ valuation and $100M+ revenue run rate:
Crusoe - $10B
Mercor - $10B
ElevenLabs - $11B
Baseten - $11B*
Harvey - $11B
Lovable - $12B*
OpenEvidence - $12B
Mistral - $14B
Nscale - $14.6B
Fireworks - $15B*
Sierra - $15.8B
Moonshot - $20B
Perplexity - $22.6B
Cognition - $26B
Scale - $29B**
DeepSeek - $52-59B*
Cursor - $60B*
Waymo - $126B**
xAI - $250B**
OpenAI - $852B
Anthropic - $965B
21 total companies.
*rumored / unannounced
**not fully independent
English
@TheStalwart S&P earnings growth expectation is $400B this year. Is there enough GDP pie to absorb that?
English
I like having a job. So consider this take to be drenched in cope.
But as of right now, I think that: coding being a relatively “easy” thing for AI to learn + the existence of many currently employed coders, implies that we’re a long way off from mass while collar disruption.
English
SpaceX will acquire enough of $TSLA to give Elon control.
Mark it.
English
$AVGO Analysts asked management about rack revenues and margin dilution. Hock Tan emphatically said only chips, no racks.
I guess Broadcom is not doing racks? Seems like this was up in the air before. I wouldn't be surprised if it didn't pass Hock's margin threshold.
English
Oh wow, so not good for $MU right now, and probably a win for $AMD.
Jukan@jukan05
SemiAnalysis: NVIDIA is reducing the SOCAMM capacity for Rubin. via @TMTBreakout
English
We should administer IQ tests to every employed American not yet working at a lab and have the govt 5x salary match the ten thousand highest scorers and just have them all work on solving the outstanding bottlenecks between RSI and material abundance.
Every month lag because the labs are overwhelmed is costing us trillions in lost potential GDP and lives saved.
English
@TMTLongShort Game theory makes sense for sure. But people are still making buying decisions. And none of them trust the AI labs. I agree that a lot of software will get eaten, but I think there will be a lot more resiliency in what’s left standing.
English
My base case is model capability improvements will far outpace enterprise ability to adapt and instead of patiently waiting the labs will verticalize, gatekeep and kingmake with revenue share agreements for frontier model access.
A recurring theme in all my posts across geopol and tech is the default assumption that most people are retarded, game theory is under utilized, and you have to be Machiavellian if you want to get shit done.
In this case you need to put the fear of god into the board members of every public company.
If you work at a lab and don’t understand how to go about doing this my DMs are open.
English
@svarncapital @PatrickMoorhead What multiple would you pay for AVGO then? Given NVDA trades in mid 20s NTM p/e
English
How I’m baselining
Just AI revenue alone (ignoring non-AI semi & software)..
2026: $56B
2027: $100B
2028: $125B (substantially more than 2027)
2029+: more imo (bases upon time taken to setup multi GW data centers)
That’s 50% CAGR for fy27 & 28 alone and their ebitda/profit margins & CFs are very very high.
Anyway that’s my back of the envelope math.
imho $AVGO is extremely cheap at these levels, given the expected growth. Market is overreacting and while we wait for the saner heads to prevail, I’m looking into increasing my share count 😎
(NFA)
English
A few things from the $AVGO call I thought were interesting:
1.AI semiconductor bookings were over $30B in the quarter, versus $10.8B shipped. Demand is running far ahead of recognized revenue.
2.Expects FY26 AI semiconductor revenue of $56B, up about 180% from FY25.
3.Reaffirmed 2027 AI semiconductor revenue at more than $100B, but did not raise it even when pressed.
4.Networking was almost 40% of Q2 AI revenue, but said the more normal level is probably closer to 30%.
5.Customer commitments much more concrete: Meta’s initial 1GW order has been received, and two other customers have $6B in purchase orders.
6.Described the AI XPU (TPU) platform with Apollo, Blackstone and others targeting more than 20GW of compute capacity through 2028, with a first $35B tranche.
7.Tomahawk 6 has been shipping for over a year, and taping out its next 200Tb Ethernet switch this quarter.
8.Gross margin pressure is mix-driven, not structural, according to management. Hock also clarified: “No rack. It’s all chips.”
9.Non-AI semiconductor bookings exceeded $6B, suggesting the non-AI chip business is moving toward cyclical recovery.
10. VMware Cloud Foundation 9.1 adds heterogeneous compute support across AMD, Intel and NVIDIA, and VMware ARR growth was cited at 17%.
Patrick Moorhead@PatrickMoorhead
Broadcom $AVGO Q2: -Revenue: BEAT, $22.187B vs $22.13B, +0.3% -Non-GAAP EPS: BEAT, $2.44 vs $2.40, +1.7% -Guide: BEAT, Q3 revenue guide $29.4B vs $28.47B consensus, +3.3 A beat-and-raise where expectations were already high. AI semiconductor revenue $10.8B up 143% YoY, above guide, on custom accelerators and AI networking, and Q3 AI semis guided to $16.0B, over 200% YoY. Adjusted EBITDA 69% of revenue. Stock off about 2.7% AH, so the bar was set high into the print. The insatiable need for AI connectivity and custom compute continues. Will be monitoring the call. That’s where the action is.
English
Action Jackson retweetledi
BREAKING NEWS: according to CloudFlare Radar Data, Agentic traffic has SURPASSED human traffic across the worldwide internet for HTML webpages.
English
@svarncapital @PatrickMoorhead I’m a long term holder as well. But this will go back to being a long term compounder vs asymmetric moves. And I don’t know if the can defend a a 35-40x p/e anymore
English
@AJacksonHole @PatrickMoorhead I disagree with Jensen here. He is trying hard to promote his expensive stuff but so far hyperscalers $$s have called his bluff but I am wrong a lot. 🤷♂️
I am not selling any $avgo & will probably add more 😎
English
English
Now commonly hearing "We want to the Semianalysis for X". Testament to what @dylan522p has built.
English
@svarncapital @PatrickMoorhead ASICS are not* capturing upside
English
@svarncapital @PatrickMoorhead On 1, unfortunately, the ASICs are capturing the upside like Nvidia. Hock was asked about this specifically - will the cost of equipment in a datacenter to serve one GW go up? Hock said no.
Agree on 2, I think we are going see generational hock tan to defend those op margins.
English
Networking is 40% of biz. Mgt said that share is high and will decline. So mix going to get worse. Honestly, this was the worst part of the call. Avgo should be dominating networking.
And we don’t know which way the ASIC margins are trending at Google. It’s reasonable to think that since Google has two suppliers now, they have the leverage negotiate price down.
English
ASIC margins being lower isn’t a secret. Volume & networking silicon should provide relief (whether it fully compensates ASIC margins is yet to be seen).
$AVGO will continue to own the more lucrative & strategic training infra at $GOOG ; Google can’t afford any misstep there.
Mediatek is for commodity inference at volume because it’s cheaper.
English
@mrgoodcat1509 @SamanthaLaDuc Look at the type of software infrastructure business Broadcom has
English
@AJacksonHole @SamanthaLaDuc Isn’t software infrastructure like the one use case for AI currently outside of being a “better search engine”
English
@AJacksonHole @bubbleboi Do you seriously think it will take the hyperscalers 10 years, even when they were developing the ecosystem for decades, albeit slowly, now they can accelerate the programmes pretty easily?
English
I like Broadcom & they do have the best networking hardware, SERDES bar none. But the asic design house business is a race to the bottom.
Fabs like Intel are planning to take a piece of that business soon and there are tons of smaller asic design houses in Taiwan taking market share. Additionally, lot of the hyperscalers are moving to in house custom tooling which is bearish both asic design and EDA.
So what we are seeing isn’t actually just bearish Broadcom it’s bearish Marvel as well. Marvel actually makes more of their revenue as a % from these ASIC design deals than Broadcom does but it’s waaaaaay less revenue, additionally it’s concentrated in just a handful of customers.
Management and investors are anchoring wayyyy too hard on the asic design businesses which are smaller slices of the pie and not the hugely more valuable scale out networking.
English
@R_and_Invest @WisemanCap They are spending 100% of their FCF on buybacks to offset SBC. They need both growth and margins rectify. Management is correct in focusing on both and not going all in on growth.
English
@AJacksonHole @WisemanCap Free cash flow is driven by top line and margins. This trend shouldn’t change. Did you not know the SBC before 1Q27 results? The delta pre 1Q & post or incremental info is the inflection in sales driven by AI. Most cos just say AI benefit them, they are actually showing it. YUGE
English
I would recommend subscribing to and reading @WisemanCap wisemanc.substack.com on $SNOW - banger article post earnings. Very well written and that line about AI workload GM is higher than traditional workload, and why is a banger!
When you read up on new products, they are getting real acceleration due to AI but RayJ poopooed the buybacks! LOL they want them to stop buybacks to grow faster AND also didnt understand the 2028 earnings guide.
IMO analysts there are smoking something. If a co is growing NRR 126% and look at the cohorts/revs growth since 2020, they are already cracking on high FCF $ to return some to shareholders and keep some for growth. To question buybacks / EPS or earnings for 2028 just seems crazy given what they told you here s26.q4cdn.com/463892824/file… that the TAM beyond 2028 is $460B in 2031 & those charts on how AI is driving more adoption of CoCo and CoWork.
This reminds me of a line that Steve Jobs once used in an interview. "You can please some of the people some of the time, but the hardest thing about affecting change is that in some areas these analysts are right...." ........on paper - IF EPS is not high 2 yrs out, then sure valuation should come in. But the hardest thing is to model the top line and margins correctly. You cant make up an EPS number based on management hear say and work backwards to margins and revenues. Thats NOT how it works. You have to start w/core biz + cohort NRR growths + margins and then back into earnings to figure out the valuation. Not to come up with valuation based on starting w/EPS and backing into biz.
Anyways, had they just read the substack and not even gone to the analyst event they would have figured out streets $7.6B revs next year is a joke. Btw, for those who don't know Databricks - mentioned in their note did $3.7B in revs and before the revenues was completed was valued at $134B in their last series round (in 4Q25CY) led by Fidelity and JP Morgan asset.
What should SNOW be worth growing that fast at high margins.
English
@R_and_Invest @WisemanCap % of FCF is what matters. They need growth and profitability.
English
No doubt it was 50% of revs for multiple years when Buffett bought it in 2021, before going down to 40% for few years. Last year was 33% and last quarter was 29% of revs. Revenues are exploding while SBC is coming down. You could see by this time next year it should be ~20% and the year after in teens. 13x on fake revs which they beat and guide higher. Trades at 9-10x real revs what they can do next year. Not cheap but again read his Substack
English