After >20 years working as an investor, I learned one big lesson. Before looking at a company financials, valuations or business model, before everything, ask yourself, do you trust the key decision makers? If the answer is NO, run away! I feel same should apply to countries..
@pequityresearch What you call demand is actually future supply or work in progress (WIP).
Exhibit 11 is also puzzling. Why should hyperscalers order chips that require 60GW if they can realistically deliver only 10-15 GW per year?
If token and computing pricing will stop to be subsidised we should finally see if there is any demand elasticity . We already had a sense of that since Anthropic increasingly moved pricing to token consumption
I don’t understand why supporters of Chinese open source keep pretending not to see what is happening right now: even Chinese “open-source” players are raising prices or gradually shifting toward more closed models.
I think China’s token dumping has already bottomed.
Just look at Zhipu, the company behind GLM-5.2. Even Zhipu has raised prices several times this year.
Chinese LLM developers cannot ignore ROI forever.
Open source is not the same thing as cheap API pricing.
@SouthernValue95 Is 20bn per GB Sustainable? There could be around 80GW or more coming on stream in US alone in the next in the next 4 -5 years . Do you see 1.6 trillion of additional demand ?
@jukan05 what your are actually saying is that the two markets are not separate. if there will ever be overcapacity in DDR you should expect it to spill over on HBM too through the channel you described
@zephyr_z9 What do you think it’s gonna happen to data centre cost per GW when memory market will be more balanced? Data centre cost deflation (better chips and lower hardware margins) in the long term seems unavoidable…
@jukan05 It may take few more years but it seems to me that competition is increasing in semi manufacturing too. Couple of years ago it was just TSMC, now Samsung and intel are starting to get some orders. In memory Chinese are increasing market share too
@jukan05 They may be 2 years behind but CXMT Market share went from basically zero 5 years ago to 8% now and 14% by 2027 according to Yole. They won’t be competitive for leading edge but with the current bull market even commodity DRAM margin are > 80%
@robin_j_brooks Robin, the CNY did appreciate in real terms from 2005 till 2022 then we had the property market crash. Maybe your timeframe is a bit narrow… below the CNY real effective exchange rate according to Citi
@arafayzafar@JosephJacks_ i.e. if an opensource model will give you similar capabilities, the TAM will be probably be closer the cost incurred to generate the tokens necessary to solve the task and the margins could be pretty low
@arafayzafar@JosephJacks_ Even if AI could replace 40% of human activities, it doesn't mean the salaries saved are the TAM. AI lab revenues will depend on their competitive position and their ability to charge. 1/2
@Brad_Setser@AngelUbide@cryptotranny@arvindsubraman A lot of Chinese households wealth is allocated to property. It’s difficult to increase spending when you feel poorer. It will probably take 1 to 3 years before property prices bottom..
@Trinhnomics you could integrate your chart with real rates. what's unique of Philippines and Indonesia is the disconnect from the equity valuations and govt bonds
@upanizza and it's not just cars... whole sectors once dominated by European companies are facing similar challenges. Look at medical devices for example. Even among luxury goods there are Chinese brands emerging...
@gianlucac1bis Sul punto 1 e 2 immagino che la creazione di soluzioni logistiche alternative (a prova di conflitto) sia la priorità anche se ci vorranno 10 anni o più. Per i GCC non può che essere una soluzione temporanea. Questo ragionamento immagino lo facciano anche in Iran…
@jasonschips I think the reply is in your post. You don't get most of those earnings, hynix does. it will depend on what CFROI they will get on their investments in the medium term. only if you think they will be able to generate decent returns beyond 2027 (>10%) you should probably buy
@GiZollino sarebbe interessante fare lo stesso esercizio utilizzando i costi attesi di produzione di solare , eolico e ESS fra 5 o 10 anni utilizzando i costi di produzione attesi + ROIC anziche' tariffe attuali.