I think they have realized that even if OpenAI is first, it won't last long so really its just compute at scale, which is something they already do themselves.
There is a moat in infra (hyperscalers, Azure, CoreWeave).
There is a moat in compute platform (Nvidia, Cuda).
Maybe there's a moat with good execution and product, but it isn't showing yet. We haven't seen real break out successes. (I don't think you can call ChatGPT a product. It has zero switching cost.)
>There is a moat in compute platform (Nvidia, Cuda).
Ironically if AI companies are actually able to deliver in terms of SWE agents, Nvidia's moat could start to disappear. I believe Nvidia's moat is basically in the form of software which can be automatically verified.
I sold my Nvidia stock when I realized this. The bull case for Nvidia is ultimately a bear case.
Coca Cola only has a moat because soft drinks and junk food are mostly a done product without much space for innovation left.
AI is still not there yet, and if any model becomes significantly better than ChatGPT people will flock over to use it despite the branding. It's only when nobody can make better models, then people will just stick to the known brands.
What moat does Nvidia have. AMD could have ROCm perfected if they really want to. Also most of pytorch, specially those relevant to transformers runs perfectly on Apple Silicon and TPUs and probably other hardware as well.
If anyone has moat related to Gen AI, I would say it is the data(Google, Meta).
> AMD could have ROCm perfected if they really want to.
It's not an act of will or CEO dictat. It's about hiring and incentivising the right people, putting the right structures in place etc all in the face of competing demands.
Nvidia have a huge head start and by the time AMD have 'caught up' Nvidia with it's greater resources will have moved further ahead.
Because we already see firms competing effectively with OpenAI.
There is as yet no indication that AMD can match Nvidia's execution for the very good reason that doing so is extremely difficult. The head start is just the icing on the cake.
I see pre AGI AI to be superset of search and its hard to argue that google isn't ahead technically for decades and what they are doing is extremely difficult.
A head start is a moat iff you can't move easily from the leader to a competitor that catches up qualitatively; nvidia’s headstart against AMD is a moat to the extent that you can't just take the software written against NVidia GPUs and run it on AMD if AMD catches up. (That is, being currently ahead isn't a moat, but it can impose switching costs which are.)
Taking code that runs against one hosted LLM and running it against a different backend LLM is... not generally a big deal. So OpenAI being ahead—in the core model, at least—is just being ahead, its not a moat.
ROCM is close to being perfected and I think in few years and some investment you can use it directly to run 99% software written in CUDA with similar performance.
It's hard but not $3T hard. AMD has consistently underinvested in their software. If they invest double digit million to make installation and usage of drivers and ROCM smoother I highly doubt they can't acheive it.
There are open source projects volunteer run projects[1] that are better than official AMD implementation in many ways.
Because they have to rethink a lot organizationaly to put more effort into software. If installing their drivers is so flaky you can't argue they really are trying or putting their best folks in this. Single click installation on top 10 linux distro is something even one (top in the industry)person could achieve.
Not all industries or product segments are equal is the obvious answer. The point here whether one agrees or not is models are easier to catch up to than GPUs
> I don't think you can call ChatGPT a product. It has zero switching cost.
In consumer markets the moat is habits. The switching cost for Google Search is zero. The switching cost for Coke is zero. The switching cost for Crest toothpaste is zero. Yet nobody switches.
1. The switching cost from Google Search is certainly not zero, it implies switching from Google, which is virtually impossible because it's tied to Chrome, YouTube, Android and Gmail
2. I don't know many people who are dedicated "Pepsi" fans, they just grab whatever drink is available Coke/Pepsi..
3. I've also not heard many people who are never willing to switch from "Crest".. People will just grab the next available option if Crest is not on shelf. No one is pre-ordering Crest.
> 1. The switching cost from Google Search is certainly not zero, it implies switching from Google, which is virtually impossible because it's tied to Chrome, YouTube, Android and Gmail
Google Search is a product. Not the whole company. Switching to most other search engines is $0. Naturally no one is honor bound to use anything else you listed either.
You’re standing in the aisle at the supermarket, looking at the soda. Everything is in stock. What do you grab? Most people make the same decision for their entire life: Coke. That is the moat that Pepsi has been trying to overcome for decades.
According to this[0], Coke (the cola brand, not the company) has a ~19% soft drink market share in the US vs. Pepsi at ~8%. This[1] has Coke (the company, not the cola brand) at a 69% market share of the US soft drink market vs. Pepsi 15 27%. Pepsi is a larger company, though, because they own a bunch of snack food brands whereas Coke is beverages only.
I used to work for Coke — they've been making a lot of money for a very long time.
The size of a moat befits the size of a castle it protects. Coke absolutely has a moat, but it's not big enough to defend Coke as a trillion dollar company.
The question isn't whether OpenAI has a moat or not, it's if its current moat is big enough to protect a trillion-dollar company.
The person I'm responding to just said there was no moat because switching costs were zero. They didn't say anything about a trillion dollar valuation. But, while we're on the topic, Google's market cap is $2T because people can't be bothered to change the default search engine in their browser.
(Never mind I was wrong -- deleting my comment to not spread misinformation. The definition of moat is wider than I was familiar with, thanks for the correction!)
I'd say it is because the $ it takes to build out even a small gpu data center is still way, way more than most small cos can do. It's not an impenetrable moat, but it is pretty insulating against startups. Still have a threat from big tech, though I think that will always be true for almost everything
I don't think the hardware is that easy to source just yet. Musk pulled some strings and redirected existing inventory and orders from his other companies, namely Tesla, to accelerate delivery.
xAI does not have infra to sell the service and integrations of it to enterprises and such. It's an open question if "models" alone and simple consumer products that use them are profitable. So, probably hyperscale cloud platform infra is a moat yes. Microsoft has Semantic Kernel, Microsoft.Extensions.AI, various RAG and search services, and an entire ecosystem and platform around using LLM's to build with that xAI does not have. Just having a chat app as interface to one's model is part of the discussion here about models as commodities. xAI does have X/Twitter data which is a constantly updating source of information so in that aspect they themselves do have something unique.
