Seems like an odd thing to do when all we’ve been hearing about for the last 10 years is how great Google’s AI capabilities are (“but you’re not allowed to actually see or use them”) - even more so when Amazon just invested $1.5B in Anthropic a few weeks ago.
It's not odd, this is how big capital extends itself into the future. Alphabet and Amazon are giant tech conglomerates that have been doing this for years. Alphabet bought Youtube, Android, DoubleClick, DeepMind, Maps, etc. Plenty of things, and maybe the best things people think of when they think of Google are actually bought from the marketplace and integrated into the giant.
They buy whatever they want that might help their cause. This quarter alone Alphabet's free cash flow is $22B. This is a small bet for them. Why wouldn't they spend a small fraction in a key area for them?
But this investment seems different from all the other ones as Google isn't actually taking over Anthropic. Amazon is an even larger shareholder and I think Anthropic has made a commitment to prioritise AWS. So what is Google trying to do here? Offer Anthropic to GCP customers as yet another option?
Enemy of my enemy? Google and Microsoft have much greater competitive surface area than either of them have with Amazon. It would make sense for G and A to both prop up an alternative to OpenAI. If Google could integrate quality LLMs into their assistant, that would boost their position relative to Apple too.
That would make sense if Google didn't have its own AI offering and research that is supposed to be first rate and a strategic priority of the company.
Of course Google would prefer Anthropic to beat (or at least be competitive with) OpenAI, but only if they can't do it themselves.
Has Google management lost confidence in its own AI capability? Or are there so many leading AI researchers that refuse to work for Google?
To me this looks like Google could be in big trouble.
Google has astronomical sums of cash and so many hands that the left hand often doesn’t know what the hundreds of right hands know. My point is $2B can be considered a side bet to make sure if their own efforts fail there is a serious competitor to open ai to buy or license by finance people without knowledge of how their own internal modeling efforts are going.
>...a side bet to make sure if their own efforts fail there is a serious competitor...
My point is that this is not how a healthy company can think about its most strategic activities. This sort of failure is not something you can hedge against.
It's like Apple hedging against the risk of failing to keep iPhone competitive by investing in some other device maker that might be able to compete with Samsung.
I don’t think it’s anything like that example, google is an advertising conglomerate not an LLM api company. Anthropic is like a camera company when apple first started making phones in your analogy. Going to become a very important feature to be sure, but not a competing device maker.
By failure I don’t mean they can’t make an LLM product, I mean the LLM product doesn’t have majority market share to be used as an ad surface (or other monetization strategy). That’s only partly dependent on their research and modeling efforts. The revenues they could make from an LLM api business are too small to register for them right now, the market share and product work is also important.
I disagree. Google's entire advertising business depends on them being the best at "organising the world's information". (Apple says Google is the best, so it must be true, right? :)
If Google fails to compete in AI, they will lose search and some competitor will siphon off all their advertisig revenue.
This is all coinciding with regulators taking a closer look at all the other ways in which Google protects its search monopoly. You know, the ones that are not so much based on merit.
Google is at risk of losing both its technology leadership and its grip on distribution channels at the same time.
I don’t see the risk at all. Google has a LLM product with much more reach and better monetization than anthropic already. If you think the DOJ is going to break up google or otherwise destroy their market share in search that’s another story altogether, but if they keep it they’ll be well positioned to make a lot of money buying anthropic and getting their models onto their infrastructure and in their channels. It’d be like buying YouTube, which had absolutely nothing to do with needing their video hosting infrastructure but because they were winning market share with their product decisions.
Google probably knows how to do pretty good AI powered replacement search.
It doesn't know how to keep new information flowing into the AI.
If AI replaces the need to visit the original website and everyone just stays on google.com then a great deal of the web will just stop being updated because nobody is reading it and nobody will read it.
That's googles problem. Frankly it's a problem for every company who wants to try and supplant google search with AI.
I was also thinking about that; in that scenario, for example instead of going to amazon.com to search and buy something you would go to a chat app like ChatGPT or Google Bard and you would interreact with Amazon's chat bot to search and buy something. Think about Chinese super app WeChat but on steroids. This is what Elon is probably dreaming about. But before that vision comes true, we firstly need protocols and standards for building chat bots inside super chat apps. If this comes true, super chat apps would be gigantic walled gardens, even worse than Facebook is today. I think the same conflicting thought process was behind Steve Jobs' push for iPhone web apps whereas web apps are more open and free for users to use than native mobile apps but native mobile apps enable more richer and better user experience. Sadly Apple decided to lock iPhone apps inside iOS and App Store.
When first LLMs and ChatGPT came about, I thought it was just another hype but the web, and the web search industry hangs in a balance (Google in particular).
P.S.
>If AI replaces the need to visit the original website and everyone just stays on google.com then a great deal of the web will just stop being updated because nobody is reading it and nobody will read it.
But even today and for a very long time as a matter of fact, you can use RSS for website updates and read them in your RSS reader and yet classic web still didn't fade away.
Google isn't a focused company, they typically run many parallel bets that often competes with each other. See all their chat apps etc. Betting both on an inhouse and an external solution is normal for them.
Name one other example of Google betting on an external solution in a key strategic area where that bet is a minority interest in a company co-owned by a big competitor.
It makes far more sense for Amazon to prop up an OpenAI competitor to secure access to state of the art models for AWS. Amazon doesn't compete on excelling in AI.
AWS and Google Cloud invested in it for exactly the same reason, to enlarge their cloud businesses. Googles previous investment came with a cloud partnership, then AWS also gave them a cloud partnership, and now Google strengthened their cloud partnership to ensure AWS doesn't take it all.
So this bet doesn't have anything to do with AI, it is a competition between AWS and Google cloud. They are competing over who gets to sell shovels in a gold rush.
But Google is very fragmented as I said, what cloud does has little to do with what other parts of Google does. This investment makes sense for cloud so they make it. It is a weakness and a strength of the company, Amazon is much more top down and that leads to other problems.
Money shouldn’t sit around? It should be invested to make more money. Perhaps Google just thought this was a good investment that would likely provide a good return? It wouldn’t be the first time a company did that.
Maybe someone can explain this, because I never understood it. When a company sits on so much cash, I guess it doesn't mean cash which is liquid, but rather a variety of assets, right?
So when they just have to pull x billions out, it's not just liquid assets, but they will have stage and sell assets representing those funds.
So since it's not just cash, how does a company of this size, then determine what assets to sell? And if those assets are actually invested in something, or representing an entity of some sorts, how do they assess whether or not it will cause any damage or loss of profitability? The crux is, is the risk two fold? First let go of whatever the assets were invested in (one), and then buy a new company and hope it has ROI (two).
Or is it actually possible to have 1.5 billion dollars laying around in cash somehow? I know money is a made up idea, but that is still a big number for a bank/banks/asset holding company to just say good for and expect some kind of real monetary tangible value behind the symbolic currency.
> When a company sits on so much cash, I guess it doesn't mean cash which is liquid, but rather a variety of assets, right?
Cash is a very specific thing on a balance sheet. It has to be cash or very close to cash. "Equivalent", something like a <90d treasury that has virtually zero interest rate risk.
So when someone says "Google has 100B cash" it would mean literally cash or close enough to cash that it doesn't matter.
You'll note, if you read the 10Q, it's also wrong. Google has 30B in cash and cash equivalents, and an additional 90B in marketable securities - stocks, and bonds with >90d maturity.
That said, "marketable securities" are extremely liquid.
> Or is it actually possible to have 1.5 billion dollars laying around in cash somehow?
Yes? Depending on what you mean by "laying around in cash". It's not literal physical dollar bills, it's numbers in a computer.
1.5B is not much for a company that size. I'd image that is payroll and accounts payable for like a week or two?
> I know money is a made up idea, but that is still a big number for a bank/banks/asset holding company to just say good for and expect some kind of real monetary tangible value behind the symbolic currency.
Bank of America alone has like 2 trillion in US deposits.
Most larger companies will use treasury management software within their finance organization to handle these issues (asset mix, risk, prediction, transfer, etc...).
Scanning the players will show you how some of them solve some of the problems you mention.
>Or is it actually possible to have 1.5 billion dollars laying around in cash somehow?
Yes. 1.5 billion is less than Google's weekly operating expense.
>So since it's not just cash, how does a company of this size, then determine what assets to sell? And if those assets are actually invested in something, or representing an entity of some sorts, how do they assess whether or not it will cause any damage or loss of profitability?
There's a lot of smart people under CFO that determine that.
It’s described in various articles as $121bn of cash, cash equivalents (debt instruments and marketable securities with maturities of < 90 days) and other short-term investments (which I think includes government bonds.) Short term investments make up more than 80% of it.
Alphabet has 190k employees. Let's assume the average salary is 5k/month, that's already close to a billion just going to employees accounts each month.
Maybe it’s semantics, but generally with fund allocation when you’re deploying more than 1% of your AUM towards a single thesis that’s a significant bet.
The closest equivalent to "AUM" for a company like that would be the market capitalization, not cash reserves, so over 1.5 trillion.
Then again, shareholders usually have less risk tolerance with companies than investors with hedge funds, so the two aren't really comparable either. (But I assume your 1% number is for more risk-averse funds? Hedge funds regularly make much bigger bets than that)
Do you ever think how much better the world would be off if Alphabet wasn't allowed to buy all those things? I do. Antitrust has to come back in a big way.
Those parts would still end up bought by Blackrock, Vanguard or whatever index fund is in fashion and steered by these funds that have the easiest business model but no capacity to innovate.
Why? "Owned by Vanguard" is the closest possible thing to "owned by the American public". What do you think Vanguard is and where do you think their money comes from?
Maybe just restrict index funds from interfering with whatever business they own? Index funds are ultimately a dumb momentum-based business model that just buys whatever companies are doing well at the moment and sells whichever aren't, but bear no risk themselves outside the economy shrinking. Not sure why they should have any say in how the companies they own operate, it seems like a complete competence mismatch (dumb ones with little risk controlling the smart ones with the skin in the game).
I don't feel that capitalism is an end to be pursued, but I agree that these conglomerates should be broken, for the sake of avoiding irreversible concentration of power.
Not really irreversible unless they get a monopoly on violence, so as long as the overton window from the publics perspective shifts dominantly towards breaking them up in the future we'll be able to get a better situation.
When you can reliably tell they have a monopoly on violence, it's far to late to do something about it.
However, Very Bad Things™ can happen long before they have monopoly on violence. And violence where? Large companies operate in weak jurisdictions, not only in their home states of Maryland and Ireland.
Well, historically, there have been a number of companies with the right to violence. The East Indies and West Indies Companies, for instance, or more recently United Fruit. I haven't bothered to check, but I'd be surprised if Big Oil didn't have mercenaries on call in hot zones, with little to no oversight.
Nobody (company, nation or gang) needs a monopoly on violence to become dangerous.
Those companies didn't have the right to violence in their home countries, they waged wars in other countries instead. Companies overthrowing their home country hasn't really been a thing.
True. However, please don't forget that the conglomerates being discussed initially do exist and have considerable influence in other countries than the US, so my remark remains :)
Also, the Medici family was initially a wool company, then grabbed power in their homeland, kept it for about three centuries and somewhere along the way produced descendants that ruled over much of Europe. Similarly, the fascist uprising that gave power to Franco over Spain was largely privately funded by a bank [1] and I seem to remember that the German Nazi party was largely funded by industrialists and bankers [2] until reached power.
So, I'd say that companies overthrowing their home country's government has unfortunately been a thing for quite some time.
They (search, social media, advertising companies) are gaining a monopoly on truth. With that they indirectly control the government, which is the one with the monopoly on violence.
To emphasize, all human productions are biased, that's human nature, and news media are not exempt, despite many outlets making serious attempts to produce unbiased news.
Part of the job of being a consumer of media is understanding the bias of what we're consuming, determining whether it skews the news, and possibly counter-balancing by consuming other media with different bias. And yes, it's lots of work and most people aren't willing to spend the time doing that.
Google Cloud and Google DeepMind have different goals.
The goal of Cloud is to sell as much compute as possible to the rest of the world. So, they want as many foundational models that are running on their cloud.
Since OpenAI is married to Azure, Google and Amazon are trying hard to get the remaining. Of course GCP can just bet on Gemini, but if you are the head of GCP you wouldn't put all your eggs in one basket. I mean the essence of cloud is redundancy and load distribution. It applies to business strategies too
I was analyzing this as not putting all your eggs in one basket in terms of AI capabilities, but you are right that it's probably also true in terms of cloud sales.
Anthropic, like OpenAI, has a big lead in Google in terms of productization, but lags behind OpenAI in visibility, posing a big risk of OpenAI getting so much mindshare that even if Google were to catch up in productization, everyone in the market would already be tied ti what was visibly the only game in town.
Helping fund Anthropic and assuring that there is visible competition also makes it more likely that LLM-backend-agnostic services will exist and a culture of evaluating offerings rather than just automatically going to OpenAI is established, which puts Google in a better position if they really are ahead in basic science and just suffering from a past lack of commeecialization focus that is remedied by the recent reorg and refocusing of Google’s AI efforts.
(Also, it may help avoid the situation where Google’s lack of product and and the financial relationship between other cloud providers and the AI vendors that have competitive product means that Azure is the favored enterprise platform for OpenAI, Amazon works out something similar with Anthropic, and Google Cloud loses competitive position because Google AI isn’t competitive and Google doesn't have the right partnerships.)
Anthropic lacks more than visibility. I don’t know if it’s an issue with compute or leadership but it’s been really difficult getting access. Of course I guess you can run it on bedrock now? I signed for api access on Anthropics site maybe 6 months ago. Never any follow up, not even to tell me I can use bedrock to run it.
They are keeping the landscape from becoming too centralized around another direct competitor.
Open AI is now basically Microsoft. Copilot in all your docs, etc.
If only Amazon was the large investor in Anthropic it encourages a similar exclusive partnership to develop where Anthropic ends up powering Alexa with proprietary access to its SotA models.
Google throwing money that way keeps Anthropic from becoming too siloed into a partnership with one cloud direct competitor, and also helps fund another competitor to the company in the lead who is partnered with a different cloud direct competitor.
I wouldn't read into it beyond that, and certainly not in thinking this reflects product shortcomings or advantages.
1. OpenAI is too large a supplier for LLMs, and they are trying to create supplier diversity
2. $2B is a drop in the bucket for Google. Even after the severe fall this month, the Google Market Cap is $1.536 Trillion. Consider the investment as a % of market cap
3. It is undeniable that small players can move fast, so even if Google's AI is amazing, the slope is probably not as high as a fast moving startup
4. These products have 2nd order revenue boosters -- you might use Anthropic's marketplace app, as a theoretical example, but you end up spending much more on compute/storage/cloud in doing so, helping the cloud majors
I would not consider the investment as a % of market cap since Google doesn't own all that market cap...
I would consider the non float shares + short term investments + cash on hand or Revenue/Profit better
> I would not consider the investment as a % of market cap since Google doesn't own all that market cap.
Market cap does matter because you can use it to gauge how easily Google could hypothetically raise the same amount[1] by issuing additional shares without a shareholder revolt.
Not seeing how that distinction matters. Google’s owners (shareholders) do own all that. And google represents the collective financial interests of the shareholders.
Isn't this just laundering money into the cloud division? GOOG was decimated after the last earnings call by poor cloud results. This is at least a billion in revenue for Google Cloud, right?
I think their primary goal is to make life difficult for OpenAI by investing in a competent competitor that brings a level of agility that Google just can't deliver.
A common failure mode for startups and a reminder of how universal those lessons are. Whether it would be the death knell of Google we'll have to wait a bit longer.
> Whether it would be the death knell of Google we'll have to wait a bit longer.
I don't understand the trope on HN that OpenAI is a competitor to Google. I do not think Google has the appetite or DNA to be purely an AI API vendor; I don't think selling APIs is as profitable as Google current business. Google investing in Anthropic is a good, cheap hedge against the low-probability (IMO) event that LLMs somehow displace the search results page and display ads.
Google engineers invented most of the primary innovations in NLP in the last decade, such as transformers and word2vec which started it all, that led to the current stack that OpenAI is built on. Their best public models certainly lag OpenAI, but it certainly isn't vaporware.
This is the only thing they can do. They don’t produce anything. They buy. Their AI is a joke and all they earn money on is ads. MS is beating them on the AI front, hard, and they have multiple interesting avenues like having an actual OS, basically all the world’s code and all the world’s business.
2B is too little too late as well. It’s almost embarrassing. MS dumped 10B into OpenAI. 2B is a lukewarm, visionless move.
Why do you say that MS beating Google on AI? OpenAI is not part of MS (49% ownership is nice but it doesn't give MS control of the company). MS by itself doesn't own any of foundational models competitive with Google. MS has a solid AI research team, but it's not ahead of Google DeepMind. Microsoft Azure beats Google Cloud, but I don't think you count GPU rentals as AI?
If MS actually bought OpenAI, it would be a different story.
Looks like Google has better hardware for AI than anyone else, and they are the only company having both AI research and AI hardware. OpenAI+Microsoft and Anthropic+Amazon are much less integrated, working at arms length.
> Google will have multiple very large clusters across their infrastructure for training and by far the lowest cost per inference, but this won’t automatically grant them the keys to the kingdom. If the battle is just access to compute resources, Google would crush both OpenAI and Anthropic.
> Being “GPU-rich” alone does not mean the battle is over. Google will have multiple different clusters larger than their competitors, so they can afford to make mistakes with pretraining and trying more differing architectures. What OpenAI and Anthropic lack in compute, they have to make up in research efficiency, focus, and execution.
> Looks like Google has better hardware for AI than anyone else, and they are the only company having both AI research and AI hardware
On top of that they have lots of training data: youtube, gmail, google docs, google drive, indexed www for google search, lots of data from those fancy cars driving around for google street project and probably still some data from google+
I don’t think 2B is too little. Anthropic already has a competitive LLM they lack a competitive dataset I think and need to scale. 2B can allow them to do this. I don’t know that Google wants Claude to be theirs they just don’t want Microsoft to own the generative AI market
They are buying upside and they are funding a competitor to OpenAI and Microsoft. The enemy of my enemy is my friend. There are only so many groups in AI that have the chops to take on OpenAI and Anthropic is one of them - if not the most promising.
In addition to what others have said about them having so much capital to invest, another angle is that AI is a hot topic right now. It is a much worse PR headline for it to come out that Google’s AI did something bad. If a headline comes out that Anthropic’s model did something bad, very few people will connect the dots to Anthropic’s investors.
