Hacker News new | past | comments | ask | show | jobs | submit
Lets assume the best Chinese AI lab is an entire year behind. These companies are valued below 20B, while Anthropic is valued above 1T.

Assume nobody ever catches up to Anthropic and everyone will always be 1 year behind. Is the 50x valuation still reasonable? Is the money really in doing PhD level math, or simpler stuff like summerizing coporate emails?

And lets be honest, some US and Chineese labs are only months behind and the only reason they have not fully caught up is limited compute. For China this could change quickly (e.g. Huawei gets their yield issues under control and suddenly Chinese labs have cheaper compute than west). People who invested in Anthropic at 50x valuation must belive this scenario is impossible in the next 5-10 years, but that is a huge bet.

My view is that Anthropic will be the Microsoft of the AI era. The only moat Microsoft has is that it's the default for many enterprise companies. AI needs access to a lot of confidential data to be useful in a large business. If Anthropic convinces enough customers to use Claude Meet, Claude Docs and Claude Sheets they win. Not because their models are so much better, but because all your business data is with them.
Do we have any numbers on who's actually using those ancillary products? All the non-tech folks I know don't touch them. If they want Claude in their docs, they'll just use the plugin for Google Docs or Word.
that's today. they might get used to it when their workplaces start pushing claude sheets/docs/etc on them
I think Google Sheets,etc is just too entrenched, I'm at a small shop of heavy Claude Code users but I haven't heard a whisper of even trying those other products.
Why is this more likely than firms "choosing" Microsoft's equivalent to Claude Code, once they inevitably improve the harness around copilot?

This is not a trick question or anything, I'm genuinely wondering whether stickiness (MS) or product quality/iteration speed (Ant\) would succeed.

It’s not going to happen at an enterprise level, Microsoft is so intertwined within every part of any large risk-averse organisation that copilot will be the default.
I'd never even heard of those offerings. Not sure they will take off.

How much actual research is Anthropic doing ? How about hardware research ?

Those apps are not any better than alternatives. We are trying to use them as experiment.

Also didnt these companies say no one will need saas apps anymore. Now its their main moat?

We're moving into a world where the model's intelligence is one checkbox among many in the product. As you point out, many companies are able to provide competent models. But that's not sufficient anymore. Buyers want features around collaboration, sharing, permissions, etc.

My analogues for this market are OSes, office software, and CRMs. Windows + Office remain revenue juggernauts 20+ years after fully free and open alternatives got good enough. Salesforce is worth over $100B, even though there have been good FOSS CRMs for literally decades.

IMHO it's difficult to predict how the AI market will evolve without understanding why people continue to choose relatively expensive alternatives that functionally check the same boxes.

Many American companies will preferentially select American providers, so I expect them to have very high retention and also high net revenue retention

Even if new customer growth falls apart, if they can hold 30% growth while public it will be a good buy today.

Why would a more competitive frontier necessarily mean a significantly lower Anthropic market share and not just higher market share for Deepseek etc?

Super intelligence threatens the marketshare of nearly all other companies.

Where is the compute to mass serve these Chinese models going to come from?
> Is the 50x valuation still reasonable? Is the money really in doing PhD level math, or simpler stuff like summerizing coporate emails?

No, and obviously not. Much more going on with this IPO than common sense investing.