E106 [AI-Translated] Burn Rate Intelligence #2 – AI-Konzentrationsfalle | VC Winner-takes-most | Open Source Risiko

00:00:00: The original podcast was recorded in German.

00:00:03: This podcast was translated using

00:00:05: Artificial

00:00:05: Intelligence, Burnrate the Venture Insider Podcast with Max Meister and Guy Giafredi.

00:00:12: Hello!

00:00:13: And welcome to Burnrate – the VC Insider podcast.

00:00:17: I'm talking here with Guy Giofredi about a startup scene where they focus on VC.

00:00:22: With episode one hundred six we're continuing our new Wednesday format.

00:00:26: It's called burn-rate intelligence.

00:00:29: In these episodes, we take up current and relevant topics from the VC world put them into context.

00:00:35: And share the knowledge that we have built up over more than thirty years of combined experience in the VC environment.

00:00:42: The Episodes are deliberately kept compact as a short impulse food for thought.

00:00:56: What is the topic we selected for today?

00:00:59: Today, we're discussing the question of whether venture capital is now manoeuvring itself into a concentration trap with AI.

00:01:06: So an extremely important topic for us venture capital investors and for the entire ecosystem.

00:01:12: I find that topic really interesting because depending on the outcome it only has negative consequences.

00:01:19: That's kind of crazy right.

00:01:21: Usually when something happens you have good or bad outcomes.

00:01:25: One consequence can still somehow be positive.

00:01:28: But here, depending on what happens over the next few months or years you basically end up with almost only negative consequences if a certain development occurs and that's what we're discussing today.

00:01:40: I find it very fascinating but also a little frightening depending how you look at it.

00:01:45: You could certainly say that

00:01:47: Exactly.

00:01:48: When we talk about the AI boom in venture capital, we usually talked about valuations billion dollar rounds and a question of which company will ultimately become the next open AI or anthropic but perhaps more interesting question by now is different one.

00:02:04: What is this AI boom?

00:02:05: actually doing to venture capital itself, because an increasingly large share of capital is concentrating in fewer and fewer companies.

00:02:14: Axios showed that very clearly this week.

00:02:16: the news outlet OpenAI and Anthropik alone are said.

00:02:27: Now, of course those are two exceptional companies but the figure still shows how much the market has changed.

00:02:34: Venture capital is actually based on the idea of a portfolio.

00:02:38: you invest in many companies some fail Some perform well and one or two exceptional winners ultimately return The fund.

00:02:46: You can also be a little more concentrated like we do at Coyo.

00:02:49: At Coyo We don't have a Portfolio Of thirty Or twenty five Companies But A portfolio of ten.

00:02:55: But we do have a portfolio, and now in the AI boom this logic seems to be partially reversing.

00:03:01: A huge amount of capital is flowing into a small group of companies because investors assume that these exact companies could one day be worth several hundred billion or even more... ...and there's good reason for it!

00:03:14: Many VCs have been waiting for liquidity for years.

00:03:17: Private valuations have risen but LPs, meaning the investors can't do much with paper returns.

00:03:24: Meaning book values.

00:03:25: they finally want the cash back and the IPOs currently being planned Anthropic Open AI and others could deliver exactly that.

00:03:34: And this is where it gets interesting And I'm looking forward to the discussion, Mickey.

00:03:38: A recent crunch-based article argues that successful AI IPOs would not necessarily make the VC industry broader or healthier.

00:03:47: quite the opposite.

00:03:48: if the biggest funds such as Sequoia and Dresden Horowitz, Axel Balderton in Europe Back open AI, Anthropic and other mega winners generate enormous returns.

00:03:59: LP capital may then flow right back to those same funds.

00:04:03: so LPs are invested with Andreessen Horowitz.

00:04:06: they get money back.

00:04:07: And what do the LPs do?

00:04:09: With this profit?

00:04:09: They invest it in Andreessen horowitz's next fund.

00:04:12: that means large funds become larger.

00:04:15: We can already see that today.

00:04:17: they can lead larger rounds pay higher valuations and defend their ownership across several rounds.

00:04:23: And in industry terminology that is called the concentration flywheel, at the same time on the company side there's a risk if open source or open weight models as open-source models are called in AI space become better and cheaper which we're already seeing today with Chinese models.

00:04:41: then question arises of how defensible current valuations of frontier labs meaning large LLMs really.

00:04:50: So the uncomfortable thesis is this.

00:04:52: If the VC's AI bet fails, which is possible they have an extreme concentration risk and if it succeeds the VC market could become even more concentrated.

00:05:03: that's what I meant when i said The outcome is negative.

00:05:06: either way let's get into it Guy!

00:05:09: Is This actually a new problem for you?

00:05:11: A Concentration Problem in Venture Capital?

00:05:13: or is this simply venture Operating the way it always has?

00:05:17: a few big winners generate their returns.

00:05:20: The basic mechanism, the power law is of course not new.

00:05:24: I mean venture as basically always followed this law to power law where if you large returns generated huge amount of money back small share of investments simply produce extremely good outcomes and allowed share of three turns.

00:05:39: The difference is where the concentration takes place today.

00:05:42: In the past, you might have had thirty or forty companies within a fund and hoped that two or three of them would become exceptional.

00:05:50: Today we additionally see Concentration at the level of the entire market.

00:05:56: A lot of funds absolutely want exposure to the same very small number Of companies namely We can name them here open AI, Anthropik and a handful of other AI frontier labs.

00:06:09: And with the biggest funds there's something else as well.

00:06:13: their funds have now become so large that many normal venture outcomes are mathematically no longer interesting for them.

00:06:20: if you know how to fund that is ten or fifteen billion in size just imagine that an enormous sum then accompany it gets sold two billion which would actually be a very good venture outcome if we invested at an early stage, only helps you to a limited extent.

00:06:37: You need exceptionally large outcomes and that creates structural pressure to concentrate on the few companies that could theoretically be worth one hundred three hundred five hundred billion or even trillions of dollars.

00:06:51: so in that sense AI is not necessarily the cause of the problem but it is accelerating it enormously right now.

00:06:58: And that naturally raises the slightly uncomfortable question for me.

00:07:02: Are VCs investing so massively in AI because objectively, the best companies are emerging there or because today's VC model now genuinely needs these mega outcomes?

00:07:13: I

00:07:13: would say it is a bit of both and certainly be wrong to say that they're simply financing dynamic.

00:07:18: behind this.

00:07:19: AI is without doubt currently one of biggest technological platform shifts we have seen in long time.

00:07:24: but problem arises when investment thesis changes.

00:07:28: Axios picked up an interesting thought.

00:07:30: In some of the investments that VCs make in an open AI or an Anthropic, it is less about picking meaning finding a good company and more about do they even have access to this investment?

00:07:40: Can they still invest in the company at

00:07:42: all?".

00:07:43: The shift then is that the investment thesis no longer we as investors recognize We absolutely have to be in these names and will do everything we can to get access, rather than picking the best companies with the greatest potential.

00:08:04: That's an important difference because What does it cost us if we're not part of it?

00:08:12: And this company really does become the next trillion-dollar company.

00:08:16: Of course, that can also be a factor driving valuations higher because the company basically sets the value and is no longer the result.

00:08:22: due diligence but Max... If we look at this so critically isn't it completely rational?

00:08:27: at the same time I mean open AI or Anthropic do become some of the most valuable companies in the world.

00:08:32: then surely its right to compete for exposure.

00:08:36: We discussed last week that we are seeing massive valuation jumps even in the later stages of companies, sometimes six times within just a few months between two rounds like we have seen with Anthropic for example.

00:08:48: And I think that's exactly why i find this situation so interesting.

00:08:51: at The level of an individual fund?

00:08:53: That can be absolutely rational but through aggregation.

00:08:58: So if one hundred investors each rationally decide that they absolutely need exposure to the same three companies, then at market level an extreme concentration emerges and That works brilliantly as long as The underlying assumption is correct.

00:09:14: but that assumption Is not only AI will be huge.

00:09:18: it also says a large share of the economic value Will remain with exactly these frontier labs?

00:09:25: And depending on how things develop That can be problematic because we are seeing how quickly the quality of models is converging.

00:09:33: For example, The open-weight models from China.

00:09:36: they're becoming much better prices are extremely attractive and the models Are becoming interchangeable?

00:09:43: And therefore the decisive question Is not whether AI becomes huge?

00:09:48: AI is huge anyway that is undisputed.

00:09:52: The decisive question, who ultimately captures the value?

00:09:56: Who is investing in right

00:09:58: companies?".

00:09:59: That's incredibly interesting!

00:10:00: Let us go a little deeper into how large is open source or open weight risk today for investment thesis behind OpenAI and Anthropic?

00:10:08: I wouldn't say open-source destroys open AI – it would also be far too simplistic But it obviously does change the economics.

00:10:16: If models become cheaper and cheaper, A short digression.

00:10:28: There's a very interesting interview in the NZZ today with an economist and he says, priced into their share prices and valuations, then there will be an enormous write down.

00:10:53: And because everything is so interconnected I mean companies are giving each other capital.

00:10:58: they hold stakes in each other.

00:11:00: this could theoretically lead to an enormous chain

00:11:02: reaction.".

00:11:04: That's something we need to examine critically.

00:11:07: but perhaps back on the topic... meaning sales.

00:11:19: We saw that OpenAI has just fired its head of sales and then brought in a new Head Of Sales from Wiz, it's about developer ecosystems—meaning who gets the best developers?

00:11:28: Who has the best

00:11:29: daughter?".

00:11:29: And so on…and so forth... That is enormously relevant for a VC —that's where they need to look very closely!

00:11:37: If you want to justify a valuation of several hundred billion dollars today You don't just need enormous growth You also need the expectation that this company can control an exceptionally large share of the future value pool.

00:11:50: And so long story short, open weight at least raises the question how secure this assumption about future profits really is!

00:11:57: Good let's quickly take a positive scenario as well.

00:12:01: we don't want to be only negative Exactly.

00:12:03: OpenAI, Anthropic, Databricks and others achieve gigantic IPOs.

00:12:08: that means LPs the investors finally get back the billions they invested And actually that should be fantastic for venture capital.

00:12:16: So why do many critics argue precisely this liquidity could make industry even more concentrated?

00:12:24: That

00:12:26: was already mentioned at the beginning.

00:12:28: I mean liquidity.

00:12:29: so a cash event like that is not simply neutral, and when LPs receive money back they then ask what's next?

00:12:37: Where do we invest the capital

00:12:39: next?".

00:12:39: And the obvious choice of course is that the manager who just paid out the capital to them and proved that they can produce extraordinary returns simply receives a large share of that capital again if not all of it.

00:12:52: So If The Large Platform funds are the ones that disproportionately hold stakes in the biggest AI winners Then They Don't Just Receive The Returns meaning The managers themselves don't just get rich, but they will very likely also receive a disproportionate share of the next funding cycle.

00:13:09: So their platform becomes much bigger again.

00:13:12: and that is precisely this flywheel.

00:13:14: You have a large fund, which means you can write large checks get access to these highly sought after mega rounds.

00:13:20: And then that produces a large exit.

00:13:23: the distribution follows and the distribution convinces LPs To invest in the same managers even larger next fund and participate again?

00:13:31: Then this cycle simply starts all over again at higher level.

00:13:35: The interesting thing is an IPO boom could solve the VC liquidity crisis Absolutely.

00:13:41: But at the same time, within the industry it could have the effect that a few large players receive even more capital because those few major investments are the ones generating the distributions.

00:13:52: Max what does this mean concretely for smaller funds and founders?

00:13:56: Is is really problem or simply natural professionalization of market?

00:14:01: For founders, it's not negative or not exclusively negative.

00:14:05: I mean a large fund can obviously offer enormous advantages!

00:14:09: It can co-finance several rounds... ...it can support companies for longer and it can provide huge amount of capital in difficult phases.

00:14:18: but there is an important side effect.

00:14:20: A fifteen billion dollar fund looks for different outcomes than the three hundred or five hundred million dollar funds.

00:14:28: so it needs large ownership stakes needs very large companies.

00:14:32: And as a result, more and more capital and increasingly large funds are competing for relatively small number of companies that can satisfy those return requirements at

00:14:44: all.".

00:14:44: That could create a barbell market... A barbell is a dumbbell right?

00:14:49: On one side, a small number start-ups that access practically unlimited capital.

00:14:54: on the other a large number of good, notably good companies for which fundraising becomes more difficult because they may be attractive businesses but are not large enough to really move the needle for a mega fund.

00:15:09: And smaller VCs face a similar problem—they're not only competing for LP capital and also for access.

00:15:18: And if LPs increasingly concentrate their commitments on a few large brands and firms, it can become even harder especially for new managers to build the first or second fund at all.

00:15:30: Good let's summarize the key points again.

00:15:32: Guy could you start please?

00:15:34: Of

00:15:34: course I would say First The concentration in AI venture market is not simply the result of AI being popular right now.

00:15:42: It also connected with the structure of VC industry.

00:15:46: Funds have become larger, LPs are waiting for distributions and large funds need exceptional outcomes meaning mega-outcomes in the trillion dollar range.

00:15:55: And no longer just in the billion dollar range.

00:15:58: that makes companies like OpenAI and Anthropic almost inevitably magnets for capital.

00:16:03: second The central question is not whether AI will become economically huge?

00:16:08: The interesting question Is where the long term economic value ends up If models increasingly become commoditized, open-weight alternatives improve and that then generates price pressure and prices fall.

00:16:21: Then value could shift away from the pure foundational model toward distribution applications data or infrastructure for investors paying enormous valuations today.

00:16:31: That is certainly not a small question.

00:16:33: And third even The best case scenario has a downside If the large AI companies really do produce fantastic IPOs, which is not yet clear.

00:16:43: LPs finally get back the liquidity they have been waiting for years but that money could subsequently flow mainly to funds already largest and strongest today.

00:16:55: In this case AI would create.

00:16:58: a winner takes most market among startups a winner takes most market among VCs and perhaps that is exactly the most interesting question for the coming years.

00:17:10: Does the AI boom make venture capital stronger in the long term or does it mainly makes industry larger, more concentrated?

00:17:22: That was it for Burn Rate Intelligence, the VC Insight podcast.

00:17:26: If you want to support our podcast subscribe to our newsletter and share with your network.

00:17:32: Thank You very much for listening!

00:17:33: We wish a good second half of week.

00:17:35: take care and goodbye.