Blog

ROIC of Nvidia data center built in 2024 for 2026 inference

Assume you invested in 2024 for Nvidia chips based data centers.

Back then, you used H200.

An 8-GPU HGX H200 server is closer to roughly 10–12kW at the server level.

1 MW IT load → ~80–90 eight-GPU servers
× 8 GPUs/server → ~640–720 H200s/MW
1 GW → ~640k–720k H200s

So 680 H200 per MW is probably more reasonable and thus cost ~$35mn per 1MW, or $35bn per GW.

 

 

1 GW IT load 2024 H200-era capex
GPU/HGX servers ~$18–24bn
Networking/storage ~$3–5bn
Facility/power/cooling ~$8–12bn
All-in ~$30–38bn/GW

Source: ChatGPT

How much revenue it can generate today?

Using DeepSeek V4-Pro pricing and token per day estimates like this give you ~$33.3k revenue/day/MW, or ~$12.2bn revenue/year/GW.

 

Per MW/day Tokens
Input ~51.0bn
Output ~12.8bn
Total ~63.8bn

Source: ChatGPT

Assume cash ebitda margin is 80%.

So the 2024 H200 gives you $10bn per year.

After tax (depreciation deducts income), that is ~30% roic over the $35bn capex in 2024.

HK record financing

Alibaba now tops the list with 80bn HKD financing.

排名 公司 / 时间 规模 关键 Terms HKEX 官方公告
1 阿里巴巴 9988 8/23交易，8/24公告 HK$80.0bn 710m新股；HK$112.70；较参考股价折价约3.6%；扩大后占股本3.57%；net约HK$79.7bn；预计8/26交割。 HKEX 配售条款公告
2 中际旭创 3308 7/30上市 HK$53.410bn IPO 54.5m股 × HK$980；net HK$52.891bn；另有8.175m股绿鞋，但该HK$53.41bn数字未计绿鞋。 HKEX 最终发售价及配发结果
3 宁德时代 3750 4/28 HK$39.190bn 配售 62.385m H股 × HK$628.20；较前收折价7.0%，较5日均价折价10.46%；扩大后稀释约1.36%。 HKEX 配售条款公告
4 智谱 2513 7/9 HK$31.411bn 最多 19.78m H股 × HK$1,588；较前收折价12.99%，较5日均价折价6.72%；扩大后约4.25%。 HKEX 配售条款公告
5 立讯精密 2475 7/9上市；8/5绿鞋 ≈HK$25.060bn 基础 383.473m股 × HK$63.28 = HK$24.266bn；8/5部分绿鞋 12.542m股 × HK$63.28 ≈ HK$0.794bn。 IPO最终结果 · 绿鞋公告
6 胜宏科技 2476 4/21上市；4/22绿鞋 ≈HK$23.135bn IPO最终95.85m股、HK$209.88，基础gross HK$20.117bn；随后全额绿鞋14.3775m股，同价发行。(HKEX News) IPO最终结果 · 全额绿鞋公告
7 美的集团 0300 5/6定价；5/7公告 HK$17.248bn 两档各HK$8.624bn、均zero coupon、100%发行。2027档：到期11 May 2027、转股价HK$96.82；2033档：到期13 May 2033、转股价HK$115.76，到期赎回114.95%。 HKEX 双档CB完整条款
8 MiniMax 0100 7/10 HK$16.041bn 配售：35.6m股 × HK$268 = HK$9.541bn；另发 HK$6.5bn zero-coupon CB，100%发行，约7/14/27到期，初始转股价HK$335，较前收premium 12.64%，到期赎回102.75%。 HKEX 配售＋CB完整条款
9 联想集团 0992 6/17交易；6/18公告 US$2.0bn zero coupon、100%发行、7年期，到期25 Jun 2033；初始转股价HK$36.70，较签约日收盘价premium约47.5%；net约US$1.979bn。 HKEX CB完整条款
10 MMG 1208 6/16 HK$6.268bn + US$0.8bn CB 配股 705.892m股 × HK$8.88；另发US$800m zero-coupon CB，发行价102%，所以CB gross cash为US$816m；21 Jun 2027到期；初始转股价HK$10.21。 HKEX 配售＋CB完整条款
11 牧原股份 2714 2/6上市；3/5绿鞋 ≈HK$12.099bn 基础 273.9514m股 × HK$39 = HK$10.684bn；3/5部分绿鞋 36.2717m股 × HK$39 ≈ HK$1.415bn。 IPO最终结果 · 绿鞋公告
12 紫金矿业 2899 1/29交易；1/30公告 US$1.5bn face / US$1.5375bn gross zero coupon；发行价102.5%；5 Feb 2031到期；初始转股价HK$63.30，较前收premium 37.19%、5日均价premium 45.92%。(HKEX News) HKEX CB完整条款
13 中国宏桥 1378 4/28 RMB10.2bn USD-settled zero coupon CB；100%发行；3 May 2027到期；初始转股价HK$43.90。 HKEX CB完整条款
14 东鹏饮料 9980 2/3上市；2/28绿鞋 ≈HK$11.099bn 基础 40.8899m股 × HK$248 = HK$10.141bn；部分绿鞋 3.8658m股 × HK$248 ≈ HK$0.959bn。 IPO最终结果 · 绿鞋公告
15 华泰证券 6886 2/3 HK$10.0bn zero coupon、100%发行；8 Feb 2027到期；初始转股价HK$19.70，较前收premium 6.78%，较5日均价premium 5.09%；net HK$9.925bn。 HKEX CB完整条款
16 洛阳钼业 / CMOC 3993 1/19交易；1/20公告 US$1.2bn zero coupon、100%发行；24 Jan 2027到期；初始转股价HK$28.03，较前收premium 28.70%，较5日均价premium 26.17%。 HKEX CB完整条款

Source: ChatGPT

Evidence of deflation in China – %Arabica coffee

I met %Arabica Coffee in Hong Kong.

I liked it, especially the design of the coffee shop; I still kept its physical customer loyalty program card.

The coffee chain expanded in mainland China in 2018.

I have been to several of its stores in mainland China and it’s one of my favorites.

In mainland China, the %Arabica coffee loyalty program is more digital (I am using its mini-program on WeChat) and ran separately from the HK one.

The mini-program kept my previous purchases and the Grande Ice Americano is 35 rmb.

You know what, I am buying the %Arabica Grande Ice Americano with delivery to my door for 24 rmb (on weekend) !

As Meituan should also take a cut, the net price for %Arabica is much lower than 23 rmb.

That’s 1/3 cut for consumers and 40-50% cut for %Arabica!

 

That even makes the housing price decline in China “normal” and “not sufficient”.

Best AI

The best AI is the AI that can learn the best after being created.

Education is college is pre-training.

Humans with the best grades in college is probably not the best humans afterwards.

Even you add post-training, which is like internship or few years of work experience, that is still not the best part.

The best AI/human can keep compound on itself.

That is especially true for AI as it can live much longer.

Alibaba’s 3-year payback period

In Alibaba’s earnings call, it touted 3 years of payback and could improve that to 2.5 years or even 2 years.

That sounds strong.. but if you compare, that is nothing.

Even Luckin claims its franchisees have 1.5 – 2 years of payback periods.

And SpaceX said during recent 2q26 earnings that it has <1 year of payback period.

Nebius’s payback period is now 1 year and 10 months, down from 2-3 years, disclosed in 2q26 earnings.

 

Bond yield becomes a topic? looks fine for Europe

Benchmark 10-year yield ranges:

Period Germany Italy Spain Background
Late 1970s–1980s 6–11% 10–14% 11–14% High inflation and nominal growth
Early 1990s peak ~9% 14.2% 14.0% Currency/ERM crisis
2000–07 3–5% 3.5–5.5% 3.5–5.5% Euro convergence
2011–12 crisis peak ~2% 7.3% ~7.6% Euro break-up/default risk
2020–21 low −0.8% 0.45% Around 0% ECB QE and pandemic policy
August 2026 3.26% 4.08% 3.72% Fiscal expansion, inflation and supply

How did European stocks perform?

 

Year STOXX Europe 600 price return Total return Broad environment
1988 +22.2% +25.4% Growth and yields rising
1989 +24.5% +28.0% Reunification optimism
1990 −17.8% −15.3% Bund yields/rates peak; recession fears
1991 +12.5% +15.8% Partial recovery
1992 +1.6% +5.8% ERM currency crisis
1993 +35.9% +40.7% Rate cuts and collapsing yields

 

Index 2011 2012
STOXX Europe 600 price −11.3% +14.4%
STOXX Europe 600 total return −8.6% +18.2%
Germany DAX −14.7% +29.1%
Italy −25.2% +7.8%
Spain IBEX −13.1% −4.7%

Note that 2011-12 is different – one key evidence is German yield yield was falling, while now it’s rising in 2026. Note that 2012 stock increased due to euro-breakup crisis being contained.

Germany’s 10-year yield has increased from about 2.86% at end-2025 to 3.26%, or roughly +40 bp YTD.

Year-end Germany Italy Spain STOXX 600 price return
2008 3.05% 4.47% 3.86% −45.6%
2009 3.14% 4.01% 3.80% +28.0%
2010 2.91% 4.60% 5.37% +8.6%
2011 1.93% 6.81% 5.50% −11.3%
Jul-12 1.24% 6.00% 6.80% Crisis peak
2012 year-end 1.30% 4.54% 5.34% +14.4%

 

Diverging path: Baidu vs. Google

In Q2 2026, Baidu online marketing revenue fell 19% YoY to RMB13.1bn, after falling 22% in Q1.

And AI-native ads doesn’t seem impressive.

Meanwhile Google Search & Other revenue grew 17% YoY to $63.3bn in 2q26.

Google search revenue didn’t feel much impact from growing usage of chat AI.

One difference, probably unrelated to AI-era, is that content sits inside giant closed ecosystems, much of which Baidu either can’t index properly or isn’t the natural starting point for.

Consumers can go directly to vertical apps. This is especially true when mobile is more important than web in China.

Baidu is simply not the end of a discovery journal, while Google still is.

Even if you chat with AI, you might still go to Google to do final checks or to look for places to do the transaction.

Another issue is that Google AI Overview seems more successful in monetization.

Baidu management says it is “deliberately holding back” AI-search monetization.

Google previously disclosed that queries showing AI Overviews monetized at approximately the same rate as traditional Search, and in Q2’26 said it continued to be encouraged by AI Overview monetization even as it expanded into more commercial queries.

 

Governments, tax, shareholders

Isn’t governments the invisible shareholder of all corporations?

Say tax is 25% , essentially gov will take 25 of 100 profit before tax, and the rest $75 are called profit after tax and is distributed to shareholders.

Does it sound odd?

Isn’t governments essentially a permanent 25% shareholder of the company, and the ratio is net by itself.

How convenient.

Btw, shareholders are should give their shares a 25% hair cut.. if they own 50% of a company, they just own 37.5% in case of 25% tax.