Venture capital is screwed. A bubble in AI is their last hope, and they’re betting everything on Sam Altman. It’s heartwarming.

Today, we have the Pitchbook-NVCA Venture Monitor for Q1 2025. This is the National Venture Capital Association telling its members how things are going. It’s the view of VC from the inside. The front cover picture is the light at the end of the tunnel — and it’s an oncoming streetcar. [NVCA; NVCA, PDF]

We covered the Q3 2024 report last year. That one tried very hard to present bad news as okay news. The new report can’t even do that.

Here is the state of venture capital in early 2025:

• Venture capital is moribund except AI.
• AI is moribund except OpenAI.
• OpenAI is a weird scam that wants to burn money so fast it summons AI God.
• Nobody can cash out.

The report mostly blames venture’s woes on President Trump and his tariffs — but venture capital’s problems are systemic and you could see them coming since the zero-interest-rate era.

And they have no idea how to fix this. None at all. The report’s only plan for the future is to cross their fingers and hope.

Trump’s tariffs have trashed everything

The Trump economic disaster is the most important material fact for the venture capital sphere. The report goes on and on about the tariffs.

Here’s the first paragraph of the report:

With uncertainty dominating US financial markets throughout the first quarter, our prior expectation that 2025 would see a resurgence in venture capital (VC) liquidity and dealmaking has been dimmed. Should the latest iteration of tariffs stand, we expect significant pressure on fundraising and dealmaking in the near term as investors sit on the sidelines and wait for signs of market stabilization.

That is: nobody is doing deals because business can’t make plans past lunchtime.

The report somehow fails to mention the bit where the Silicon Valley VC and executive crowd worked their backsides off to elect Trump and several of them sat in the front row at his inauguration. Then they were actually surprised when the leopard ate their faces too.

The end of zero interest

VC is hilariously unable to cope with not being in a zero-interest-rate environment.

In 2008, the global financial crisis hit. Central banks use interest rates as the brakes on an economy, so they took the brakes off to make sure money kept flowing around. Interest rates went below 1%. And interest stayed near zero for over a decade. [St Louis Fed]

But zero interest meant that if you had to get returns — say, you’re a pension fund — then there weren’t a lot of places to invest. Venture capital was one place you could put money and get returns.

This meant a flood of money into startups, and that meant all sorts of ridiculous startups got funding. Businesses that lost money on every transaction. SoftBank was setting money on fire as fast as possible, ending in the disaster of WeWork.

Borrowing money was almost free as well. Why not load up on debt?

But then inflation came along, and the Fed put interest rates back up — and suddenly there was less money. Why put money into venture capital when you could get back 5% just buying Treasury bills?

Venture capitalists were around before 2008 — including a lot of the current guys! — but it’s like they completely forgot how to do anything without free money. The party’s over, and now it’s time for the hangover.

So we get hyped up tech bubbles praying for lottery-level returns — because ordinary business returns from, say, making a thing and selling it don’t cut it. Our present hype-fueled AI bubble is a complete worked example.

Everything is OpenAI

This year, the investors are all-in on AI. Crypto’s still dead, quantum isn’t taking off — AI’s the last game in the casino. [TechCrunch]

There’s lots of graphs in this report showing healthy dollar amounts. These are mostly the “$40 billion” OpenAI deal led by Softbank, and then the other AI startup deals. Almost all the graphs and statistics in this report are lies based on the one huge OpenAI outlier.

Andreessen Horowitz is right now raising a fresh $20 billion investment fund to try to keep the AI bubble going.

VCs still believe the marketing brochure version of AI, because they don’t actually know anything about technology:

Just as “internet” evolved from buzzword to business backbone, AI is following the same playbook.

No it isn’t! A completely different thing taking off and working says nothing about whether your personal favourite thing will take off and work! Stop saying dumb things!

Pitchbook says: “57.9% of global VC dollars invested in Q1 went to AI and machine learning startups.” Most of that’s OpenAI. Even Pitchbook calls it “FOMO” — fear of missing out. [Pitchbook]

AI is the only prospect for a lottery-level win. You and I know that’s not going to happen, but the VCs and their customers, the investors, do not know that. They’re still in. Where else are they gonna go?

Everything that isn’t AI is moribund

There’s a world of startups outside AI, and in 2025, they’re not having fun. Investors do not want to put money into early-stage startups with nothing to show as yet:

First-time financing activity remains low, reflecting broader trends across the VC industry. In Q1, we recorded $3.8 billion invested across 892 deals, which was down both QoQ and YoY.

… Only $10 billion was raised across 87 VC funds, setting a pace for the lowest year of fundraising in a decade.

Early-stage firms that do get venture funding tend to get acquired early on, instead of doing more rounds to pump their imaginary value number first.

Even when a fund gets investors, it can hardly find anywhere to put the money. One VC firm, Charles River Ventures, returned $275 million of uncommitted capital to its investors. And wasn’t that a shock to the world of VC. [NYT, archive]

Number must go up

These companies have mostly imaginary valuations — say they pay $10 million to get 1% of a company, that makes the company worth a billion dollars, right? And not the ten million of actual money it’s got.

The VCs jealously guard these imaginary valuations as big fat book entries they can trade with each other. Just don’t ever look inside the box, don’t test what the market with actual dollars thinks the value is.

A pile of startups just ran out of cash and died in 2024 — more than in 2023. And 2025 is expected to be even worse. [TechCrunch]

A lot of these were frankly trash companies funded a few years ago, and they should have died then. But VC still had the startups’ imaginary values on their books and they didn’t want to admit the startups were worthless until they were forced to.

Nobody can cash out

AI is the only thing that’s catching the eye of the investors. The investors don’t want to put their money into other areas because they’re just not getting payouts.

Here’s the simple version of venture capital investing:

• A VC firm starts a fund.
• Big investors — pension funds, insurance companies, sovereign wealth funds, super-rich family offices — put money into the fund.
• The VC firm puts the money into startups that will progress to an exit — an initial public offering or a bigger company buying them out.
• The investors get paid. Hooray!

Most startups fail, but one big winner can make up for a pile of losers.

Unfortunately, the stock market is broken — the Trump tariffs again — so IPOs are not popping on launch, so they’re not happening.

CoreWeave, riding the AI buzz for all it was worth, was the big public exit in first quarter 2025, and that one fell flat because their numbers were objectively terrible and their business plan doesn’t work. Ed Zitron has covered CoreWeave at length. But even with all that, CoreWeave on its own was 40% of all first quarter exits by value. [Where’s Your Ed At]

The other big exit was Wiz Security getting bought by Google. But as the report says, “the rest of the exit market has remained relatively silent.”

Funding rounds are increasingly late-stage — that is, VCs are mostly putting extra money into existing companies, often just to keep them on life support.

Many of these later funding rounds also have secondary sales — where startup employees get to sell some equity to the new investors. Because then the talent might not give up and quit to work at a non-zombie. And the old investors get some money back from a company that’s not doing anything.

The future of venture capital

The report’s theme for the future is: hope really hard! Maybe markets will recover and IPOs will pop again. Maybe Trump will come up with a coherent plan and stop being volatile and weird. Maybe!

That’s it. That’s all the plan they have. These guys haven’t got a clue what to do next. They really don’t. This is the industry talking to itself and saying “well gosh, chaps, looks like we’re screwed.”

It’s important to note something else here: quite a lot of the venture capital class are not actually so bright. But they got lucky once.

That’s a big thing to say. They’re way more rich and successful than you, David. These are intelligent and educated people, they were definitely good at something! A lot of them have Ph.D.s!

But also, it’s horribly clear that right now they’re way out of their depth. Semafor documented last week how these guys enthusiastically joined a long-running billionaire group chat with literal neo-Nazis as their court philosophers. Marc Andreessen got into e/acc — which is literally a neo-Fascist revival movement. [Semafor]

This is not smart behaviour! This is dumbass behaviour!

These are not people who can be trusted with dangerous machinery such as billions of dollars. They were lucky as hell, and now they’re not.

The only future plan in this report is: maybe Trump will stop acting like a bizarre freak. They’ve met the guy and they think that’s a plan!

Maybe they can start a bubble in guillotines. [Mastodon]

Not that anything unpleasant should happen to them. None of the VC class will ever miss a meal. But you can feel a small amount of schadenfreude at their mild discomfort, while everything burns around the rest of us.

• Video version

It was my mom’s birthday this week and we went to see the play “Good Night, and Good Luck” on Broadway. George Clooney, his hair dyed jet black, plays the newsman Edward R. Murrow, who stood up to Joseph McCarthy during the height of McCarthy’s anti-communist witch hunts, helping to precipitate the greasy Senator’s downfall. The parallels to our current time were easy to see. The audience, hungry for hopeful bits of history to buoy their spirits in the age of Trump, cheered wildly at every deadpan line about standing up to bullies. But the underlying message of the story is not the simple good-and-evil story that everyone thought they were applauding.

It did, of course, take courage for Murrow to run stories critical of McCarthy. He knew that McCarthy would retaliate by smearing him as a communist sympathizer, and he did. In response, Murrow gave a speech denying any communist sympathy, and proclaiming his love for America, saying that he, unlike McCarthy, believed that people could engage in civil debate with the reds without become reds themselves. Murrow’s ultimate triumph was one of manners, which the public ultimately preferred over McCarthy’s boorish browbeating. The most important lesson in the play, however, lies in what Murrow didn’t do.

America’s most respected newsman told his audience that McCarthy’s most serious violation of decency was the fact that he flung around false accusations of unpatriotism. In defense of the ACLU, Murrow said, “Twice [McCarthy] said the American Civil Liberties Union was listed as a subversive front. The Attorney General’s list does not and has never listed the ACLU as subversive, nor does the FBI or any other federal government agency. And the American Civil Liberties Union holds in its files letters of commendation from President Truman, President Eisenhower and General MacArthur.” He went on to say, in that famous broadcast, “the line between investigating and persecuting is a very fine one and the junior Senator from Wisconsin has stepped over it repeatedly. His primary achievement has been in confusing the public mind, as between the internal and the external threats of Communism… This is no time for men who oppose Senator McCarthy’s methods to keep silent.”

McCarthy called people and institutions communists. Murrow replied that, in fact, they were not communist, they were upstanding patriotic Americans, and that McCarthy’s methods of accusation were out of line. What Murrow did not say is: “It doesn’t matter if people are communist or not.” He did not say: “The conflation of communism with anti-Americanism is a cheap rhetorical trick.” He did not say: “I reject the implication that communism is a threat to American values.” He did not say: “Perhaps the communists are making some valid points.” Murrow’s bravery was real, but its boundaries stopped at the edge of the stars and stripes. He wanted to contest McCarthy on the field of patriotism. He could not bring himself to peer into the hollow heart of patriotism itself. Thus, Murrow’s victory allowed Americans to sleep soundly in the knowledge that decency had prevailed, without ever peeking under their beds at the enormous pile of skulls.

When one child in a schoolyard tries to insult another by saying “You’re gay!”, the proper response is not to cry, “No I’m not!” It is to say, “So what if I was?” To accept the very premise of the slur is to validate it. Grasping this distinction is a mark of moral development. Yet this rudimentary principle has, from Murrow straight through today, always been a bridge that America’s pillars of liberalism could not bring themselves to cross. What held Murrow back, and what still holds his successors back, is their determination to Put America First. They are Americans and they work for Americans and they want (in the kindest way) for America to flourish and they believe (evidence notwithstanding) that America is good in its heart. It is a quasi-religious belief—by far the most popular religion in our country. You can call this American exceptionalism, but let’s call it patriotism for short. Those who presume themselves to be making a case for liberal morality first bash themselves on the head with the brick of patriotism, and then wonder why they can’t quite think straight.

Capitalism of the sort that America practices naturally generates patriotism for the same reason that jockeys put blinders on horses: It keeps things running smoothly. It precludes questions. It channels vision into a single direction with no distractions. It is good for business. This practice has been spectacularly successful for American capitalism. But there is no reason for the people and institutions who are supposed to promote the common good to hobble themselves with patriotic shackles. And yet they do. Murrow, for all of the praise that we lavish him with, could not wrap his head around the idea that perhaps communists—people who wished for a more fair and equal world!—should not be used as a repugnant counterpoint for all that America represents. The decency that Murrow sought existed in a sharply bounded channel of public discourse in which it was more outrageous to label an American as a communist than it was for the American government to launch a murderous global war against the ideology of communism that would destabilize nations, smash nascent democracies, and cause untold suffering over the next several decades. If you accept that America is good and America is free enterprise and communism is the opposite of free enterprise then communism is the opposite of America which means that communism is bad. This kindergarten-level puzzle was all it took for the right wing, the real army of capitalism, to create a social sanction against straying beyond the bounds of patriotism so strong that it still defines media and politics and popular culture today. Indeed, we are all swimming in its fetid political retaining pond right now.

Liberalism, with its embrace of universal values and the rights of mankind, has never truly prevailed in America because most of its alleged advocates have not been willing to release their grasp on patriotism. For fascists, patriotism is a door to pass through; for liberals, it is a wall. They should walk around it, but instead they continue to bang their heads against it. It makes liberals look pathetic, scared of their own conclusions. “We, ah, don’t support the war, but we support the troops, and we certainly support the pilot of the bomber, as an American, but we don’t support the bombs themselves, although we do support the company that makes the bombs, since it’s a pillar of the American economy, but we hope that the bombs don’t kill anyone innocent, but we still hope that American wins the war, though we know the war is unjust, because we love America.”

“We are outraged at January 6th not only because a bunch of poor suckers were dumped by lies but also because breaking into the US Capitol is an assault on democracy, notwithstanding the fact that all attempts to subvert democracy at home and abroad originate there. There are flags there and they must be respected.”

“We believe that Joe Biden is a good man, because he did good things for America, and yes, perhaps he facilitated the violent killings of a few tens of thousands of children, but that is complicated and we are going to therefore act as though it should not be a part of the conversation.”

“We are the free press. We are here to report the truth. In wartime, we run weepy graphics with waving flags and explicitly hope for America to win the war. We are the free press. We do not find it necessary to ask what ‘terrorism’ means, and instead focus our questions on whether, you know, teenage college students actually fit the description. We are the free press. We are here to support freedom and condemn the destruction of property. We call the president ‘sir’ no matter how tyrannical he is. We respect the same institutions that exist to oppress us. We wonder why our truth telling has not caused American democracy to flourish.”

Donate to our reporting fund

Alas for liberalism, it suffers from an acute lack of public figures with the courage of their own convictions. It is not tricky or complicated to believe in universal human rights, but following that belief to its logical conclusions is quite bad for the progression of your career. “Hello, I would like to be a national newsman. I will not wear a suit and call a man ‘Mr. President’ if I know he has committed crimes against humanity. I will not wish success to the troops if I do not believe they are killing people for just reasons. I will not brand people or organizations or entire nations ‘terrorists’ without interrogating what exactly they believe and what they are trying to accomplish and why they are trying to accomplish it that way. Do they really ‘hate our freedom’ or did we, you know, do a lot of evil stuff to them?” This does not fly. This is a nonstarter. This will bring mockery and job loss and marginalization. It is not that this sort of thinking is some esoteric intellectual secret—it is embraced by everyone from rural Buddhist monks to highly educated academics—but it is not the sort of thing that is allowed to flourish on the stage set that is Mainstream American Discourse. A total rejection of patriotism, which is a prerequisite for an honest discussion of national affairs, is a nonstarter for those who want to be a member in good standing of national politics or the national press. As a result, the conversation that flows out from these places is a warped and stunted version of free inquiry, a field that is fertile for thoughtless nationalism.

And here we are! Once you accept the premise of patriotism, you have already lost. There are those who believe that they can call themselves patriots because they yearn for the promise of America, the higher values that the founders vowed to aspire to, even knowing that we have never achieved them. But this, too, is a trap. What these people are embracing is not patriotism, but fandom. They were born in America and they are fans of it because it is their home and they hope that it will be good. Fine. I am a fan of the Jacksonville Jaguars. I cheer for them and hope they win. That is fandom. But if they lose, I do not accuse the Houston Texans of terrorism and communism and raise an army to subjugate the rest of the NFL to serve the interests of the one true and righteous team, the Jaguars. That would be patriotism. Do not call yourself a patriot if the latter version of things makes you uneasy. That is the final outcome that waving that flag leads to. Do not step on that train, and you won’t end up there. Simple.

Free yourself from patriotism’s burden. Breathe the clear air of universal human rights. It is the inability of the alleged liberals to walk away from the fixed game of American exceptionalism that leaves them always battered and bruised by those who don’t give a fuck about universal human rights at all. Once you stand on the field of patriotism, stealing all the world’s wealth and buying more guns than anyone else and using them to keep the whole world working for us makes more sense than anything else. Each year, the Global North uses its might to expropriate over 800 billion hours of labor from the Global South. Is that bad, for humanity and equality? Yes. But what are you gonna do—advocate for a lower standard of living for Americans to make up for it? Ha! Try rolling that one out at the presidential debate. It is out of bounds. It violates the law of American prosperity above all. Discussion of it must remain relegated to theory rather than practice. The wheedling liberals who try to have it both ways, who try to square the circle of American prosperity with the nice desire to be nice to all the nice people of the world, will always end up sputtering uselessly as strongmen vow to do whatever it takes to keep us rich. Patriotism has lured us to a losing game. As we gape and scratch our heads and wonder why the little steps of progress towards a more just nation seem to always be followed by a vicious backlash, and why the Democratic Party always seems to compromise its way to hell, and why an obviously corrupt and dishonest would-be dictator is able to accumulate power with promises of Making American Great Again, just look at that flag on the wall, on the lapel pin, on the football field, at the parade, on the shirts, in the corner of your TV news screen, and you will be able to deduce the answer.

I guess that’s kind of what the communists were talking about the whole time.

Leave a comment

More

• Related reading: Regarding January 6th, American Is Built to Feed Us Poison; Regarding elections, How to Think About Politics Without Wanting to Kill Yourself; Regarding the Democrats, Why Would Dick Cheney Endorse Kamala Harris?; Regarding language, Retire the Word “Terrorism”; Regarding the USA, Nationalism Is Poison.

• Bad shit is going down but it is not hopeless because the people ultimately have the power. There will be many actions on May Day: find one here. Protests are taking down Tesla: find one here. You can organize your workplace and join the labor movement: get help here. I wrote a book about how unions can get us out of the bad place we’re in: you can order it here.

• There’s a lot of crap to look at on your computer so I want to thank you for being here and looking at How Things Work. You might notice that this site is free for everyone to read. I think this is best. The reason I can keep this place free is that a bunch of people make that possible by being paid subscribers of this site. They allow me to do this work and they also allow me to keep this place open for everyone. They are doing a nice thing for everyone. If you like reading How Things Work and want it to continue to exist, you can help by taking a quick second to become a paid subscriber right now. It doesn’t cost too much and it gives you the satisfaction of supporting independent media with no American flag graphics in the corner. I appreciate all of you for making this place work.

Subscribe now

YOU CAN ALSO BUY THE UNION-PRINTED HOW THINGS WORK T-SHIRT AT THIS LINK. FUCK YEAH!!!!!!!!!!!!!!!!!!!!!!! [ORDER HERE]

Can people be persuaded not to believe disinformation?:

Dr [Thomas] Costello believes chatbots work where humans fail because they offer rational responses instead of letting emotions get the better of them. What’s more, they are able to comb through their extensive training data to offer precise counter-arguments, rather than the generalised ones humans often reach for in debates.

Researchers lift the lid on how reasoning models actually “think”:

When Claude itself is asked to reason, printing out the chain of thought that it takes to answer maths questions, the microscope suggests that the way the model says it reached a conclusion, and what it actually thought, might not always be the same thing. Ask the llm a complex maths question that it does not know how to solve and it will “bullshit” its way to an answer: rather than actually trying, it decides to spit out random numbers and move on.
Worse still, ask a leading question — suggesting, for instance, that the answer “might be 4” — and the model still secretly bullshits as part of its answer, but rather than randomly picking numbers, it will specifically insert numbers that ultimately lead it to agree with the question, even if the suggestion is wrong. 

These stories were posted just a few days apart. It’s comical to me how many AI researchers act as though the hallucinations and bullshitting simply don’t exist. Also: LLMs are not rational or irrational or emotional or anything else that human beings are. They are the conduits, thanks to their corpora, of human rationality or irrationality or emotionalism. 

The discourse on “AI” systems, chat bots, “assistants” and “research helpers” is defined by a lot of future promises. Those systems are disfunctional or at least not working great right now but there’s the promise of things getting better in the future.

Which is how we often perceive tech to work: Early versions might be a bit wonky, but there’s constant iteration and work going on to improve systems to be more capable, more robust and maybe even cheaper at some point.

The most pressing problem for many modern “AI” systems, especially the generative systems that are all the rage these days are so-called “hallucinations” which is a term describing when an AI system generates incorrect information. Think a research agent inventing a paper to quote from that doesn’t exist for example (Google’s AI assistant telling you to put glue on pizza is not a hallucination in that regard because that is just regurgitating information from Reddit that every toddler would recognize as a joke). Hallucinations are the big issue that many researchers are trying to address – which mixed results. Methods like RAG are shifting the probabilities a bit but are still not solving the problem: Hallucinations keep happening.

But I think that this discourse misses an important thing: Anything an LLM generates is a hallucination.

That doesn’t mean that everything LLMs generate is incorrect, far from it. What I am referencing is what hallucinations are actually defined as: A hallucination is a perception you have that is not connected to any actual stimulus. You hallucinate when you perceive something in the world that you have no sensor data for.

The term hallucination itself is an anthropomorphization of those statistical systems. They don’t “know”, or “think” or “lie” or do any such things. They iteratively calculate the most probable set of words and characters based on the original data. But if we look at how it is applied to “AI”s I think there is a big misunderstanding because it creates a difference between true and false statements that just isn’t there.

For humans we separate “real perceptions” from hallucinations by the link to sensor data/stimulants: If there is an actual stimulant of you feeling a touch it’s real, if you just think you are being touched, it’s a hallucination. But for LLMs that distinction is meaningless.

A line of text that is true has – for the LLM – absolutely no different quality than one that is false. There is no link to reality, no sensor data or anchoring, there’s just the data one was trained on (that also doesn’t necessarily have any connection to reality). If using the term hallucination is useful to describe LLM output it is to illustrate the quality of all output. Everything an LLM generates is a hallucination, some just might accidentally be true.

And in that understanding the terminology might actually be enlightening, might actually help people understand what those systems are doing and where it might be appropriate to use and – more importantly – where not.