As predictions are only worth anything, if they are made ahead of time...I predict their share price will lose 25% of its value,over the next 3 months. Drop to 40 to 45% of IPO price, after their first public quarter release.
and has hard financial commitments 25x times, the money they just raised, over the next four years,
They have about $1 trillion commitment in compute over the next 5 - 15 years. I'm not sure where you got the 25x$81b from.They're at $86 annualized revenue right now. If they grow at 0% for next 5 years, they will generate $430b, which is enough to cover all compute costs.
No, predictions are only qworth something if you are willing to put money behind them.
cross shareholdings, agreements to buy, technology access, barriers to entry, assimilation of smaller upstarts into the chaebol, bonds, lending....
Fully open models will be more hobbyist, I expect. Open weights will continue as a strategy but probably not as prominently as today.
Training is very expensive and open weight models don’t generate revenue comparable to frontier labs.
The reasons they exist are mostly “commodify your complement” and (in my, apparently controversial opinion) the Chinese govt trying to take down the American economy and undermine the US frontier labs in order to compete in the AI race.
Possible, as long as everything goes well for them. Their debt deals don't care how much of a transformational company they are. If they get less demand than they expect, or slower demand than they expect, they could end up bankrupt.
The answer is that business is not just technology.
On the technical side, it's also important to see beyond the current market, what's currently possible, and broaden to - ok, many innovations are coming in AI, who is most well-positioned to be the centre of that?
The faith/hype vs objective data doesn’t support the valuation.
It is entirely dependent on having faith that this will work as expected.
Google could "Internet Explorer" Anthropic to death with a superintelligent ad stuffed free Gemini. shivers
> Google could "Internet Explorer" Anthropic to death with a superintelligent ad stuffed free Gemini. shivers
Hell, Microsoft could do the same with a cheap inferencing model; if people simply use copilot because it's already embedded into all of their applications because it's embedded into the OS (Office, Windows, anything that runs on Windows), only a few dedicated people will continue using the expensive paid-for options.
You can say the same about the Company of Scotland.
If their bussiness case is not true, we will be paying for this bubble in higher interest rates, for the next 25 years.
Now tell me: Which scenario do you think is more likely until 2030?
A - We cure cancer, create self improving models of hyper intelligence, that will solve Fusion and bring us to the stars, create a CPU, an OS and a 3D printed Car and deliver it to you within the same afternoon, solve Earth warming issues...
OR
B - This biggest credit bubble in history, will put the sovereign debit of the US and 30 year Treasury to 8%, force the US government to default, cut spending, increase taxes and bring about a recession that will take 25 years to recover from?
Maybe. Or maybe not.
It has already defined the future in some ways. Claude code has already part determined the future of software engineering, for example.
And yet.