Category: personal_noteS

China data center left behind?

Equinix capex for 2024 is ~$3bn

Digital Reality capex for 2024 is close to $3bn

The largest independent data center GDS operator only has rmb 6.5bn (less than $1bn) capex for 2024, of which rmb 4bn is outside of mainland China.

GPU shortage is the key, which limits the development and usage of AI.

Overall weak demand in China is also the key, as main internet companies need a surge in net income or cash flow first to support a surge in capex.

2025/26 could be interesting. If US feels the lead in AI is firm, and the H800 etc. is more of a legacy chip, and Huawei develops better AI chips, it doesn’t hurt to sell some Nvidia chips to Chinese customers.

$EQIX capex guide

$DLR capex guide

 

 

Growing capital expenditure for AI firms (MSFT + GOOGL + META => $150bn)

MSFT capex saw 41% increase in CY2023 and 79% increase in the first quarter of 2024. And MSFT expects material sequential increase in capex, which could mean 50-60% increase yoy.

Alphabet capex saw 2% increase in 2023 and 91% increase in the first quarter of 2024. The company expects similar amounts in the following quarters, which points to 50% yoy increase in 2024 capex.

Meta capex saw 12% decline in 2023 and 5% decline in 2024q1, but the company upgrades capex guidance for 2024 to be $35-40bn, which indicates a 33% yoy increase at midpoint.

Adding the 3 above would be $150bn combined capex already, up from ~$100bn in 2023 for those three.

Chinese companies share-based compensation (3)

See previous

Chinese companies share-based compensation (1)

Chinese companies share-based compensation (2)

 

Tencent (HK.700)

2023

FCF: 167 bn rmb (company defined)

SBC: 21 bn rmb, or 12.6% of FCF

Note: FCF significantly improved, while buyback is doubling SBC

 

Meituan (HK.3690)

2023

FCF: 33.6 bn rmb (op. cf minus purchases of pp&E)

SBC: 8.4 bn rmb, or 25% of FCF

Note: FCF significantly improved; buyback starting in 2024

US homebuilders in 2008

What did US homebuilders do in 2008? Residential property market was really bad.

D.R. Horton revenue dropped by 41% yoy in 2008; loss of $2.6bn (more than 2x of 2006 net income) was incurred. Book value was only $2.8bn at 2008 YE.

But D.R. Horton maintained positive cash flows, scaling back expansion and selling inventories.

D.R. Horton started to pay down some debt in 2007 and did so in 2008 as well. It continued to do so until 2011. In 2012, it started to take on more debt.

Similarly for Lennar – revenue dropped 55% yoy in 2008 and incurred loss of $1.1bn. Book value was $2.6bn at 2008 YE.

It maintained positive operating cash flow, reducing supply (deliveries dropped 68% from 2006), pausing expansion and selling inventories.

It didn’t take new debts, but focusing on paying back.

 

 

Microplastic (detection & replacement), sounds like a future business

This article is pretty good at summarizing the importance of microplastic to health (downside), including in cardiovascular disease etc..

Global companies are trying to address this issue for a long time. The most recent development is Starbucks’ announcement of redesigned single-use cups with 10-20% less plastic.

EU’s objective: aims to reduce microplastic releases by 30% by 2030.

See a previous article for fast-fashion and microplastic.

 

Chinese companies share-based compensation (2)

See previous Chinese companies share-based compensation (1)

 

Baidu ($BIDU)

2023

FCF: 25,424 mn rmb

SBC: 6,345 mn rmb, or 25% of FCF

Note: FCF improved over the last two years

 

iQiyi ($IQ), a subsidiary of Baidu

2023

FCF: 3,315 mn rmb

SBC: 637 mn rmb, or 19% of FCF

Note: 1) FCF turned positive in 2023. 2) Capex defined by mgmt is narrow – only “fixed assets”; note that acquisition of intangible assets is higher than acquisition of fixed assets.

iQiyi does say that “Capital expenditures are incurred primarily in connection with construction in process, computers and servers.”

Chinese companies share-based compensation (1)

It’s important to calculate FCF ex-SBC.

SBC is a real cost of business and equity holders should be aware of the dilution.

Sometimes, buyback is smaller of SBC so shouldn’t be touted too much. Net buyback is more important.

This is a series of posts. Starting with Kuaishou and Bilibili.

Side note: FCF/Capex is defined by companies if they present, which could be subject.

Kuaishou (HK.1024)

2023

FCF: 15,881 mn rmb

SBC: 3,570 mn rmb, or 22% of FCF

Note: SBC has down as stock prices are down. FCF improved; was negative in 2022

Bilibili ($BILI)

2023

FCF: -1,033 mn rmb (negative FCF)

SBC: 1,133mn rmb (as FCF is negative, SBC is making it more negative)

Note: SBC is flattish over past 3 years. FCF improved over 2022 (-6,611 mn rmb), but still negative in 2023.

Is LLM still on track of its dream value? Note from podcast

Note from this interesting BG2 podcast: AI Demand / Supply – Models, Agents, the $2T Compute Build Out, Need for More Nuclear & More

Very good discussion. And some notes.

Some would argue that while LLMs are indeed good in coding, chatting etc., that “amazingness” was extrapolated too much. In other words, the gap between pre-ChatGPT and ChatGPT-3 is not gonna continue/replicate. The expectation is too high that the room to beat is diminishing.

Some would question the “moat” of most LLMs. The most advanced ones, the “leading edge” is good, but the rest could be commoditized. Then all those R&D dollars is more of less duplicated and return is low.

GTP-5 is undoubtfully powerful and could be more powerful than people think, but what’s next? What cards haven’t been played or thought about after GPT-5?

Besides technology advancement, there seems to be a problem in regulatory environment in nuclear power in the US, which put a lot of restrictions.

France and China have lowered the nuclear power costs over the years.

 

Zillow’s problem?

1/ the $1.8bn lawsuit became the catalyst to lower commission revenues for each home transaction.

Some real estate agents may exit the industry as the pie is smaller; and spending on Zillow per transaction will likely decrease.

Therefore, Zillow’s revenue from home sales shall decrease under the current model.

2/ competition from CoStar is very real while Zillow is not growing

CoStar combined traffic from 23q4 earnings presentation

Home.com (owned by CoStar) had 149 million unique visitors in Feb 2024, with 567% yoy growth.

Zillow’s 2023q4 reported average monthly unique users was 194 million, down 2% year over year. Visits during Q4 were 2.2 billion, up 1% year over year.

Zillow’s DAU trend from Investor Deck February 2024 – not looking amazing.

Their business models are different. CoStar is more SaaS like. And Zillow is more e-commerce.

(above from ChatGPT 4)

3/ of course the current interest environment is not helping.

While it’s not new and Fed may cut this year, marginal benefits of lowering from 6.5% to 6% won’t move the needle.