Investors throw another $13B on the Anthropic cash bonfire Anthropic has just pocketed another $13 billion, pushing its valuation to a staggering $183 billion – fresh proof that investors still can't kick their AI habit. The Series F round, led by ICONIQ and Fidelity Management, "reflects Anthropic's continued momentum," according to the Claude chatbot maker. It could also be seen as …
I'm thinking there's soon going to be some shareholders somewhere who are going to raise a stink when they're presented with yet another bag of billions thrown to something that has no ROI.
This is one case where I will accept that the selfish cunts put a stop to the project.
Perhaps more accurately "the technology really has become too big for failure to be recognised or admitted."
But fail it will. It's now approaching a scale where its disturbance of finance and economics globally could undermine it's raison d'être and a great deal more.
Top-tier Banks are "too big to fail" due to their systemic interconnections and a government could step in to guarantee debts to prevent a catastrophic meltdown of the monetary system as happened with Lloyds in the UK.
These AI ventures are generally invested in by direct private capital, investment banks, and pension funds - all of which are perfectly capable of failing, and some of which could potentially be propped up if their investment decisions proved to be poor, but AI companies themselves are not "to big to fail" - the market correction is coming, and these tech firms will take a hammering - many will say about time, as to be honest despite their insane valuations, technology companies globally are currently providing very little that is genuinely "new" - the valuations are created by clever marketing, not underlying value.
The investors will get their fingers burned, many will cry foul, as with HP Autonomy & Theranos in the past. Some might be propped up/bailed out depending on who they are, some will learn their lessons (for a while at least), and some will go under - thus has it ever been.
"Top-tier Banks are "too big to fail" due to their systemic interconnections and a government could step in to guarantee debts to prevent a catastrophic meltdown of the monetary system as happened with Lloyds in the UK."
Ok, fine, but let's make sure when the government has to step in, the C-level execs wind up in jail and/or are ruined for the rest of their professional life and banned from holding the same sort of position anywhere. The vast majority of the executives holdings would be the first funds to "bail-in".
Its investor valuation. I give you 10 guilders for 25% of your company, which means I think your company is worth 40.
Note, too, that these investors aren't collectively crazy. Theyre just gambling on that one of these shops may be the next Google and that is pretty much their job. Place a lot of bets and if only one pays off, you can show your backers a positive return on investment. Also, the FOMO is real. A VC fund won't often be asked why they did invest in something - that's their job - but eyebrows will be raised if an opportunity walked by, they said no, and then that opportunity exits at an eyewatering amount of money.
In other words, things are mostly working as intended.
> Note, too, that these investors aren't collectively crazy.
There's also the fact that it's possible to know something is a bubble and make money off of it anyway--if your timing is right and you get out before the burst. Of course, this is easier said than done, but also of course, most professional moneyfondlers have an unjustifiably high opinion of their own intelligence.
"Theyre just gambling on that one of these shops may be the next Google and that is pretty much their job. Place a lot of bets and if only one pays off, you can show your backers a positive return on investment. "
It is like gambling as without a lot of research and understanding of a business, there's not a good way to asses whether it's solid or not. Pure gambling. When I see terms in the financial news such as "breakout" or "our charts are showing", I run. A company that isn't going well isn't going to "breakout" and charts are often just a guessing machine based on some investment bank's "formula" and, once again, meaningless.
Take something such as humanoid robots. The hype surrounding them is all "Your plastic pal that's fun to be around". It will fold your clothes, feed the cat, water the plants and change the baby's nappy (Yikes!). Do you want to risk your child or have your cat form a bond with the robot while they snub you? House plants might get over-watered or die when the robot "waters" them with an empty watering can. What there hasn't been is solid business use cases that would justify a herd of $100,000 machines that have to be watched like an older relative that's not as sharp as they once were.
Sure - that's exactly the value - anything in the world is worth EXACTLY what people are currently willing to pay for it - be it food on a supermarket shelf or a 60+ year old Ferrari 250 GTO. Shares in AI companies are worth ridiculous amounts simply because of FOMO (which even itself isn't a new invention - we used to call it "Hype"...)
If a company's shares are trading at $100 each and there are 1 billion shares, then it is said to be worth $100 billion. For a publicly traded firm every investor has the opportunity to sell their shares at the current "bid" price, so the market cap (price x outstanding shares) accurately reflects the company's value at that time.
For a privately traded firm like Anthropic not every investor has the ability to trade their shares (the trading rules are - within the limits of the law - up to the company to decide) so there isn't a current "trading" price you can look at. The next best thing is the most recent trade, and that's what this is. If someone gives a company $1 billion and ends up with 10% ownership in the company, it is said to be worth $10 billion. Maybe that's true, maybe it isn't - it depends on whether some of the existing investors would be willing to sell their stakes for less than that amount but are unable to based on the company's trading rules.
I imagine that for a stock in a hot still inflating bubble like AI most current investors would want to continue riding it up for now, so Anthropic's valuation is true today. It might not be a few months from now if the AI world started getting bad news that made people believe it has peaked and the bubble is starting to burst. Or Anthropic might get someone else writing them a big check and claim a valuation 50% higher.
"so the market cap (price x outstanding shares) accurately reflects the company's value at that time."
It represents the current "market capitalization" at that time. Value is a judgement call, not an absolute. If Elon Musk once again sold a large chunk of his Tesla stock, the next day the price of that stock could drop precipitously. The company would be doing exactly the same business as the previous day, but shareholders would place a lower "value" on the company seeing that the largest shareholder is getting out. That they haven't be extremely concerned that the BoD sells off the stock they receive as soon as its allowed, I have to wonder about.
"Are we now "valuing" companies by how much money they've had thrown at them?
What I see is the poor use of the word "value" and "worth". We need a shorter word that means the same as "market capitalization". The best investors peel back the gloss coating and try to understand the fire-sale potential of the company and what the company might bring as a going concern based on its historical income. Anything going forward is a guess, but there's a difference between an informed guess and standing under the kool-aid bong and chugging it down. Our family trust is around 95% invested in companies that pay a good dividend and 5% where we are taking risk that they have a chance of doing a bit better than the dividend returning stocks. A P/E in the altitude of 200 is a no-go, no-way. The bigger the company is that has that sort of ratio, the more likely they are to be a bubble. A very small company might be growing into its shoes, but still a pretty big risk since they'd need to be big enough to make being listed publicly worth the added expense.
This proves yet again that companies and investors will gladly pour billions of dollars into magic GenAI beans that promise to grow into endlessly churning money printers "any day now", all without a pesky employee workforce crying about labor rights or governments imposing regulations on how reckless you can be with it.
We just need all the electricity you can generate, a few hundred more data centers and endless pallets of purpose-built hardware you can *totally* use for other, more productive things once this collective cash-greed hallucination falls apart.
"In the old days, if a company lost a billion dollars, is was finished. Done and closed."
These days the company might be called "too big to fail". I would really like to see that hypothesis tested much more frequently.
If a company fails, if there is anything of value left, the vultures will fight for their place around the carcass. Yes, some people will be hurt, but there is not such thing as "fair".
The IT world has worked to repair its reputation after the Y2K bug did not result in the catastrophic consequences forecast. Techies would argue that the billions spent dealing with the bug were why nothing bad happened. To the general public, though, the panic was unjustified, and the tech sector was subsequently regarded with suspicion for several years.
This always bothers me. The media hyped the bug, then when we successfully dealt with it the papers weren't lauding us as heroes, but questioning whether it was just a scam all along.
In AI land, we're in the hype stage.
At the time the Y2K bug loomed, I had a much upgraded Gateway (remember them?) LP48633SX Computer, originally with a 486SX33 CPU and 8MB RAM, which I had bought in 1994. I ran one Y2K test from a computer magazine, and another from the internet.
The computer passed both. Obviously, it wasn't remotely comparable to the major systems so many people relied on, but I've always wondered......
End of 1996, Tescos received a shipment of baked beans with a selll-by date in Jan. 2000. Their systems couldn’t handle it, so they just changed the date to Dec. 1999, but they knew they had less than two years time to fix this particular problem before it became very inconvenient, and three years until total breakdown.
The media adores fear porn. You may notice each 'crisis' requires our impoverishment but a few's enrichment to solve.
Problem - Reaction - Solution.
For those unfamiliar, you create a problem, wait for the reaction, fear and anxiety, you then profer the solution you want and in fear people grab at it without thought.
"The media hyped the bug, then when we successfully dealt with it the papers weren't lauding us as heroes, but questioning whether it was just a scam all along."
The media only has a passing interest in the truth. Controversy, spectacular wrecks and bodies strewn all over the landscape gets more views/clicks. Write it up with weasel words so it sounds like reportage, but has escape hatches when the lawsuit is filed.
Over time I've had to be very careful when talking with customers as they will remember the bits they like the best and forget that you said "if everything goes perfectly". I quote jobs for much more than I used to. If everything goes perfectly, I can submit an invoice that is under the quote if I like. I have also found out that many times the customer is ok with spending more than I might have quoted in the past.
The long term most sustainable and profitable path for AI development for the big companies lies within sane constraints of cost to revenue. Even if you give it a little leeway that would be revenue > cost * 0.5 or so. Like a bonsai tree? I'm guessing in software algorithms might lead to a 2 - 10x improvement in economics (reduction of energy requirement) for learning, for the same hardware. Hardware is where the real improvement opportunities lie - like orders of magnitude. That kind of improvement can't be ordered on demand, but science funding and gathering of international talent can sure increase the likelihood.
"The long term most sustainable and profitable path for AI development for the big companies lies within sane constraints of cost to revenue."
They are going to need to play with it a lot to see where it has the most real value and hone that rather than having something that does everything.
A legal AI could be a very useful tool if it didn't hallucinate and spit out made up cases and specious arguments. Feeding in SovCit babble by scrapping the whole of the internet is a very bad thing if the system can't be trained to spot it and discard that data. A soft-split from that system might be one that's meant for lawmakers and has been fed different meals that might include more radical ideologies. I enjoy SF novels that explore different societal structures and while most wouldn't work someplace such as the US, tidbits might be worth pondering. For a legislator looking for ways to craft new policy on a subject, it might move them out of their echo chamber.
And the other sneak oil dealers also claim AGI is within their reach. There is not still a single model that were given simple axioms and concluded and come up with proper formal finding like human mind does. Not a single one all are numerical calculations that are being presented to look like human. If not a regular classifier parrots and yet these stupids are burning all resources like in the times of alchemy where some used to sale how to make gold out of dirt. Indeed pathetic. A lot of people will be hurt....
At some point, some people will work out how to do the whole LLM thing at much lower cost. This will cause a humongous bubble pop. Possibly followed by a financial crisis that will make 2008 and possibly 1930 look like a walk in the park! Oh well seems to be a repetative thing in human history so I don't think we can avoid. Just pay off those debts people because one day they will be called in and Blackrock will own your house.
