SPEAKER_00: hey everybody welcome back to twist this is alex and it's wednesday july 8th and that means it's time for yet another venture capital roundtable today we are going to dig into how exposed american startups are to a possible ban on open model exports from china secondary markets and their needed fixes startup m&a and why the ai conversation has re-centered around the value of data and more to help you understand all of this i brought a couple of friends in one corner we have mr nikhil basu triveti of footwork last fund was 225 million dollars he's a backer of companies like gpt zero and windborne nikhil hey glad you're here great uh great to be here alex and great to be with you michael as well speaking of which we also have michael kim of sundana capital a fund of funds investing lp capital into early stage venture funds it also runs a secondary fund and a co-investment fund for series b startups and later michael welcome to the show great to see you guys thanks for having me on so i'm glad we have we have both of you because we have the traditional vc perspective and then we also have kind of the lp side of things and the biggest point of conversation or contention really i think in venture circles for the last week may have been the blow up between usvc the angel list aligned open ish venture fund and also andrel the well-known late-stage american dynamism defense company essentially catching people up who don't know usvc claimed they purchased uh exposure essentially to andrel at its series h price and then mr grimm a co-founder of andrel uh said no you didn't and there was a big blow up on twitter trying to figure out how this all went down uh the gist gentlemen as far as i can tell is that people are figuring out that public markets had reasons for some of their rules about transparency disclosures and so forth and secondary markets are still a bit like the wild west at the same time aren't they supposed to resolve the liquidity issues we've seen in venture recently so first of all uh michael what was your take on the back and forth mess and also was anyone in the wrong per se or is this just kind of a case of everyone's trying to do their

SPEAKER_08: best and ended up at cross purposes yeah i think it's more of that and you know what we see often is uh the playbook where fund managers have access to an interesting asset they'll create an spv and they actually use that to entice lps to actually make a commitment to their fund so you know that's been going on for a while um the issue with spvs of course is the provenance of whether they actually have the shares and then when you start stacking multiple layers you know most investors don't do

David Sacks: the diligence to ensure that each layer is legitimate and that's where the the trouble can start so when

SPEAKER_00: you say that some firm managers are using spvs as a way to get lps to invest in their funds are you essentially saying they're putting together one-off deals to get um i don't know relationships with lps

SPEAKER_07: and then using those to later raise a traditional fund yeah exactly they might be raising a fund right

SPEAKER_08: now and you know in their prior fund they might have a very interesting asset uh a consensus deal that everybody wants to get into they might have some pro rata in it they might create an spv and say hey lp if you come into this spv uh you know you can come into our fund so uh you know that's a dynamic that i think a lot of emerging managers use nikhil i've never heard of that particular um

SPEAKER_07: method before is that common should i be am i behind on this i think what's happened is a few factors

SPEAKER_20: uh that lead up to uh the situation between andrew and usbc so let me try to unpack a few of these the first is that our industry is more power law pilled more power law obsessed than ever before and so there's just a small handful of companies that um you know uh gps and lps want to be a part of um i've heard from some lps it's a hit list of 10 companies i've heard from others it's 20 i've heard from some it's actually just six um but when you have that obsession over a small handful of companies um the the sort of downstream effect is that uh you know people are are tripping over themselves to invest in those and find creative ways to invest in those uh and be a part of them in in some manner and so to michael's point like one of the methods is i have access to you know a couple of those companies in that in that small basket and um and that's a mechanism for me to get uh potentially a fun going uh by enticing lps and so uh we have definitely heard about this but it is for a very small set of companies such as uh anthropic open ai uh angel and again maybe like 10 others uh that this phenomenon is occurring um and then the other sort of factors that lead into this are you know companies want to stay private for longer and so um you know you have to find creative ways to be part of these companies in the private markets versus them just going public uh another is the founders want as much cap table control as possible um and so unfortunately these factors don't fit perfectly together right they're at odds and hence why you get dynamics such as what's unfolded between between angel and angelus okay so here's the problem i see with that because

SPEAKER_00: if you're going to stay private essentially forever and i was just talking to um the ceo of hippocratic ai for our ai show the other day about this and he was like why would i ever go public private markets are super deep and i want to have all this control and not deal with the headaches of being public i guess fair enough but if you also then ban or preclude secondary activity to provide liquidity to your investors you end up essentially locking up people's capital what feels like forever or only michael at the whim of the ceo which seems like a pretty difficult place to put capital under a stewardship model because if you can't guarantee you can get it out

SPEAKER_27: then what are you actually buying other than a stock of monopoly money yeah i mean i think if you're

SPEAKER_08: going to be buying into one of the six companies that nikhil's talking about um you have to have some confidence that ultimately down the road you can actually get liquidity yourself and so you know if you take stripe as an example they have regular uh tender offers if you take spacex and that's probably the best example historically they've had regular tender offers and so um that effectively is uh the you know a public market for a private company and i think that um that dynamic that motion is is a lot better understood now compared to say 10 years ago and it's exactly as nikhil said companies are staying private longer and so um you know i think the uh uh a ceo founder has the understanding that you know you got to provide some liquidity to your people and you know it's great that that's happening the other great thing is that you know secondaries used to be uh almost uh shameful it

SPEAKER_32: was almost a dirty word yes why would you do that and now it's a well well used tool in the toolbox

SPEAKER_08: and i think you know obviously part of that is because of the dearth of distributions over the past few years we're going to be entering in a new phase where there is going to be a lot of distributions you know largely from spacex open ai and anthropic probably databricks um but you know we can talk about this but those the the the beneficiaries of those distributions are a handful of groups uh a handful of bc firms a handful of lps so it's it's also going to be the case of have and have

SPEAKER_35: nots in the lp world back to the point that uh nikhil said about being power law pilled i mean this also applies to exactly what we're talking about here uh i laughed at the databricks comment because having

SPEAKER_00: spoken to ali godzi over the years uh ad nauseam about when he will finally list and being told forever no i've given up on that company i don't think they're ever going to list i think they're going to be the next stripe and just stay private forever uh we're going to get back to information rights and transparency in the secondary markets in a second but first we're going to pay the piper uh dear friends we here at twist are big fans of the plod pin we've talked about it on the show endlessly jason's an enormous fan we wear them all the time as our note-taking devices if you are someone who creates a lot of information goes to a lot of meetings runs a company has a family to organize keeping track of what's going on is incredibly useful and important we recommend the plot note pin s and if you want to get one for yourself you can go to plod dot ai slash twist p-l-a-u-d dot ai slash twist use the code twist save 10 percent be cool like jason or like myself and therefore never having my wife nag me about things that i've forgotten it's aces so on the information point you guys talked about how uh the markets are getting a bit better much more than they were 10 years ago but one thing that uh macrim raised the end rule guy who was mad was that a lot of these investors simply don't have access to any information about how these companies are performing and so to me michael the point that we now have more professional or you know maybe just uh just more liquid secondary markets doesn't solve the information problem and therefore puts a lot of other people that want to get access to these companies at a material disadvantage and that just doesn't seem like a good long-term solution to the lack of ipo problem to me it just seems very uneven and so i'm curious like is that a reasonable perspective because you have a secondary response you're on the buy side of this at times

SPEAKER_37: one of the most important things you can do as a busy founder is stay focused you need to prioritize the decisions and workflows that are going to move your company forward not get lost in extraneous details and minutia that's why the best decision you can make for your business is signing up with every every is the all-in-one platform for incorporation banking payroll benefits taxes basically your entire back office they're going to keep up with all the bookkeeping and paperwork so you can stay laser focused on your next big product or hiring features customers all that important stuff if you're just getting started they're going to handle your delaware c-corp and getting yourself a registered agent all with no legal fees and no delays and every has worked with over 1 000 startups already that's the kind of know-how and experience you can rely on so head over to every.io and stop wasting time on stuff that isn't growing your company that's e-v-e-r-y dot i-o yeah i mean the

SPEAKER_08: asymmetry of information is certainly the dynamic that secondary funds um you know i wouldn't say prey on but take advantage of and you know uh it could go both ways you know someone who's selling may accept a 50 discount because they know that company's actually probably worth a lot less whereas the buyer you know the secondary firm thinks that they have the um information arbitrage that they they have a strong conviction high conviction that the company's actually be worth a lot more so that that's actually the the the price discovery and how transactions get done you know i wouldn't say necessarily that um either side is right more often than not but secondary funds do perform pretty well um they're never going to be a 5 to 10x kind of return but could they deliver you know a solid two two and a half three

SPEAKER_28: times yes and we've seen that time and and again over the past 10 15 years all right so nikil on the

SPEAKER_11: point you made about there only being so many companies that are of note it's 10 20 or six whatever

SPEAKER_00: it is and all the liquidity from the secondary markets pooling over there what does that do to a

SPEAKER_35: fund of your size 225 million in its last case um you're only going to have so many bets in those few companies that have high amounts of secondary liquidity so how do you approach providing what

SPEAKER_00: your lps need if you can't just put all of your funds into let's say anthropic well i mean first of

SPEAKER_46: all we've got a uh our fundamental job is to get into companies at an early stage in our case that end

SPEAKER_20: up being in that power little basket down the road of course um the other interesting thing that is happening is uh those parallel companies uh are acquisitive and so a number of us at the early stage uh have have ended up with shares in those companies by those companies acquiring our uh our portfolio companies um and so then we actually uh you know uh have the chance of liquidity uh um you know this is actually interesting aspect of the information rights conversation as well which is sometimes we we get real access to that information um at a company that's acquired one of our portfolio companies and sometimes we have we have zero um which is another interesting dynamic but i think the fundamental job we all have is to is to back those companies that are parallel companies and then the benefit of having a smaller fund is that it isn't just those companies that can matter for us and so you know uh if you have a 500 million dollar exit let's say in a company where you own 15 and you you get back 75 million that for uh a couple hundred million dollar fund is actually quite meaningful it is really not meaningful though when you have a billion dollar fund um and so um those are

SPEAKER_48: some of the the dynamics that i think are really relevant to fund sizes like ours yeah you know what i

SPEAKER_08: i would also add um it's the stage so like if you're a very early stage investor and you're in a company like cursor 11 labs you can actually do secondaries along the way and no one cares in fact everybody wants you to do secondaries because then you know from the founder perspective just one pocket going to the other um but you know if you're like a lead investor like say a sequoia or an excel or you know whomever founders fund and you own 20 of the company you really can't do secondaries because then aside from what nikil's talking about in terms of the fun math it also sends a negative signal to other investors like oh why is the lead investor you know say sequoia selling right now um and so

SPEAKER_53: they're negative signal risk essentially yeah exactly whereas no one cares about the early stage guys

SPEAKER_00: selling in fact they encourage it make this work for me because on one hand we see andrew stressing founder cap table control and i don't think they're making that point only about super late stage hottest companies and then also what michael's telling us about how early stage investors are almost encouraged to sell some of their holdings so how prevalent is the i don't know how to phrase this politely uh founder iron grip on cap table and disallowing these transactions they often need to

SPEAKER_35: be blessed and also the demand to let people like you cash out i'm just trying to figure out the balance

SPEAKER_20: between the two if that makes sense i think i haven't seen actually um that much conflict uh uh you know from my own primary uh experience with this because the founders ultimately want folks that are long-term uh aligned on the cap table and they understand especially for those investors that believed in them early um i think they're sort of grateful for that belief often and they understand the needs of those early investors and you know transferring positions over to folks that are more long-term aligned is actually in their benefit and so i think that the the situations that are more hairy are when um you know those things are trying to be done more under the table um and then there's a lack of understanding of who actually holds the the shares and so um you know i think as long as

SPEAKER_48: as folks are communicating well and transparent i haven't seen that many problems with misalignments

SPEAKER_35: okay so that's that's encouraging michael one more question about this topic before we move on to to

SPEAKER_00: mna but there's been a a bevy of slightly vague posts from vcs over on x in the last couple of weeks about a rise in fraud in the spv secondary market uh how much of that is smoke and how much of that is fire because based on what nikhil just said it doesn't seem like there's that much conflict which to me

SPEAKER_08: sounds like a low fraud environment yeah you know it it it gets headlines but you know we work very closely with a firm called klein hill so we have a joint venture secondaries fund with klein hill and i was actually just talking to them the other day about you know are you seeing a fraud in these spvs and generally speaking we don't and they've not and they're very active in the market so you know but then again they're not necessarily buying into a bunch of spvs but you know i think they have a very strong network we have a strong network and we don't really hear of of fraud um does it exist i'm sure it does and there are people who've been sent to jail because of it but um i i think they're just more like headline grabbing kind of you know tidbits yeah because when uh spacex was going to list people

SPEAKER_00: were like oh man how many people bought essentially fake spacex shares and i was kind of waiting for those stories to bubble up after the ipo and i didn't see a single one and so to me it seemed like people were really over indexing on a on a concern that wasn't there all right uh let's talk about exits in a different way uh there's a company called superhuman had him on the show a bunch of times they just bought a company called gpt zero which nakil i think you are an investor in that's right

SPEAKER_68: yeah we led the company series a i served on the board yeah so i i well first of all can you tell us

SPEAKER_00: a little bit about how that deal came together and uh why was the right exit for gpt zero and then we're going to talk a little bit about numbers but start there please yeah the the companies have

SPEAKER_20: known each other for a long time and uh there are a lot of parallels in their businesses so uh superhuman um is a combination of a number of companies uh grammarly coda um and now superhuman as well and and and sort of change the name to superhuman of the whole company when grammarly acquired superhuman um and shashir their their ceo has known uh edwin and alex the founders of gpt zero for the last couple years um uh the companies you know have been talking about doing this together for a while because there are a number of parallels in their products and i think what the gpt zero founders got really excited about was the opportunity to have a lot more distribution uh through superhumans distribution from their various products to tens of millions of active users um and so while gpt zero you know it's publicly reported they uh they got to tens of millions in in revenue uh they actually never burned a dollar that we invested uh the company's lifetime profitable has more cash on the balance sheet than it has ever raised so they didn't need to go find home but uh this was a very attractive offer for them uh and and so uh that plus the opportunity to go build something uh even bigger together is what got everyone really excited to do this um and so an amazing outcome for the founders

SPEAKER_32: a great outcome for all the investors so alex i have a couple stories here and you'll i'm a personal

SPEAKER_08: investor in your fund so i'm very happy for for myself and for you so thank you um but you know we're also an investor in uh sandana is an investor in uncork and one of the partners there is trip jones and the funny story i want to tell you is that he learned about the founders of gpt zero because he was reading the princeton alumni magazine and he was hearing about like what these kids were doing because they were i think still at princeton right and nikil went to princeton as well and you know so he actually got on the plane and went and met with them and so now it was coming down and there was a a tier one firm a multi-stage firm that had a term sheet but um the the the the founder of that firm was supposed to get on a call blew off the call so basically trip won the deal agree has an

SPEAKER_37: incredible new offer exclusively for twist fans and family members go sign up at their easy to use all-in-one contract to cash stack and if you tell them jason sent you you're going to get 50 off for life agree is the number one fastest way to go from contract to cash this means gathering all those e-signatures the invoicing the billing the payments and of course revenue recovery no more jumping between four or five different platforms just to write out your contract and get it signed set up the billing and send the invoice and start generating instant invoices payments and automated revenue workflows the moment the contract is signed check out the numbers the average time from contract sent to contract signed is under seven hours with agree.com and the median time from invoice sent to invoice paid is just 36 hours so if you want to stop chasing invoices go right now to agree.com and tell them jason sent you and you'll get 50 off for life and then the other element that i think that's

SPEAKER_08: kind of interesting based on what we were just talking about um the parents of these founders were saying we're pressuring the these kids to like sell the company not now but like this is a while back and what they did ultimately was they did a small secondary uh which got the parents off the backs of these the founders because they're like okay now you have some dough and you're in the

SPEAKER_53: bank account just go execute so secondaries can actually be very very beneficial in terms of

SPEAKER_80: alleviating financial pressure yeah but that's not financial pressure that's familial pressure which

SPEAKER_00: i've never seen come up in a case like this yeah i understand parents wanting their children to take money off the table to avoid owing princeton and their uh lenders more money for a long period

SPEAKER_59: of time but that's that's crazy uh nikil when this deal came together uh were you in favor of it from

SPEAKER_20: the very beginning as we talked about earlier when companies are really working um they have the chance to be you know uh uh part of that power low basket and so my initial reaction uh you know when we first started having conversations superhuman was we absolutely shouldn't do this we should go for it um and i think part of the responsibility that we have as early stage investors and board members is if we feel really strongly that a company has that that chance to be a big independent company you know we have to give that perspective to the founders and so uh i uh you know i i uh i strongly you know voiced that opinion and thankfully we didn't uh pursue an acquisition uh a year ago uh instead we did a secondary as michael as michael referred to but um but this time around it was a very attractive offer and so um my responsibility shifted once the founders decided they really wanted to do it to making sure everyone had the the best outcome uh possible here and so um that's what we

SPEAKER_35: worked uh you know closely together on so i i ask about your relative interest in the deal because we're seeing a little bit more of this we just saw uh versell they bought uh better off which had only raised five million figma i know it's now public but they just bought the team behind bud which i think was a vibe coding platform and if we take a look at the data here uh this is from my friends over at crunchbase showing global venture-backed m&a accounts and exit value uh there's been a pretty

SPEAKER_00: steep ascent in uh total dollar value from kind of the nadir in the late 23 early 24 era to what appears to be a relatively robust amount of money changing hands here the total number of exits doesn't seem to be changing too much uh but i'm curious about what you guys are seeing in the building uh kind of inbound from either the largest private companies the stripes the anthropics and so forth or the public said and who's more interested in picking up um startups today and michael let me

SPEAKER_28: start with you on this yeah i mean i think we're a beneficiary because we're in cursor through neo

David Sacks: and you know obviously a 60 billion dollar exit is is always welcome um they are outliers um but you know it shows how uh you know within with with power law uh a small firm like neo can return

SPEAKER_32: multiples of their fund uh with one investment um i would say that you know given where we are on

SPEAKER_08: valuations and um you know it's basically currency that uh a company can use to acquire other companies and so even though spacex is paying 60 billion for cursor you know when you have a two and a half trillion dollar market cap it's less than you know what five percent it's low single digits to them and you could argue that spacex is overvalued you could argue it's undervalued but they have the currency to do it um and so that i mean obviously spacex is now public but even for private companies if databricks is valued at 150 billion dollars for them to issue a billion or two billion of their stock it's less dilution to them and they can do that and they can actually turbocharge and you saw that in the late 90s with cisco and cisco had acquired over 120 companies people are arguing that that basically was their r d strategy but you know cisco was trading at 100 times pe so they had the currency why not

SPEAKER_35: yeah no there are some advantages to being public i hate to sound like the old broken record but like to me you have a lot more firepower to do stuff because you have the liquid currency that you have

SPEAKER_32: that in the private markets now right i mean open ai anthropic companies like that they they have you

SPEAKER_97: could argue overvalued currency but they're going to use it yeah no no i i i by that point the overvalued

SPEAKER_35: currency part yes less liquid but still very useful um i'm curious about if there's a genre of startup

SPEAKER_00: that's seen the most interest because i was a little surprised to see uh nikhil the uh bud deal be more of a talent acquisition i thought aqua hires were kind of out of vogue given that everyone's trying to produce their overall human headcount so in your portfolio uh who's not who's getting the most kind

SPEAKER_99: of door knocks from other companies i mean i think that there is such a freemium right now on talent

SPEAKER_20: because it is uh it is relatively easy to go start a company and get funded and so you know a lot of really talented people are doing that and um and so and then it is just hard to hire great people um based on a set of dynamics so yeah i mean our teams that that that have a lot of talent are absolutely getting reached out to and then the companies that have momentum in their businesses um you know that that's what a company like superhuman wanted in the case of gpt zero along with of course the team but but you know there there's a real business that they're picking up as well um so it it does feel again if if i uh i'm sure if i uh if i leverage the models to ask about like our m a conversations in our team meetings and how that's trended over time the last couple quarters has probably been the high watermark since we started footwork five years ago in terms of the number of discussions that we've had internally about our own portfolio companies getting reached out to by potential acquirers which is an interesting data point that kind of matches the the crunch based data that you shared yeah are the

SPEAKER_07: prices being discussed attractive because we mentioned earlier that many secondary deals often trade at a

SPEAKER_00: discount to last private prices uh you're holding primary shares so you don't want that so when these acquisition offers come through are they being put forward at a price that you think is fair and

SPEAKER_99: attractive again of course there's uh there's nuance in every situation but um i think the dynamic that

SPEAKER_20: that michael uh and you were talking about is really interesting which is when the consideration is not cash but instead equity then the key question is what is that equity worth and uh and so you can believe that perhaps an offer is unfair if you think the equity is far overvalued relative to the fundamentals of business um on the flip side uh you know if if if anthropic had acquired one of your companies uh a year ago uh you would probably feel terrific about um that equity position today and so um you know i think the the the main conversation that that we're having in each of these situations is okay like what is this offer actually and what is the equity actually worth um whether it's a public company or a private company doing the acquiring because in both cases um that equity may not be really tied to the

SPEAKER_108: fundamentals of the business are you trying to say that the nasdaq keeps going up every time the state of hermos explodes and that's slightly confusing to all of us it is a wild time on how to value

SPEAKER_20: anything what you know i think you know when you look at a company like hubspot in the public markets that's three plus billion of arr and i think valued at less than 10 billion dollars versus some of the companies in the private markets that are valued at 10 plus billion dollars uh with not much revenue

SPEAKER_111: it's just a it's just a crazy time exactly 10 billion uh today actually for i mean we're talking about

SPEAKER_08: anderal you know uh lockheed martin trades at about 125 billion dollar market cap with 75 million billion of revenue and you know anderal shares are trading above that market cap and the private markets uh and you know anderal has a fraction of what lockheed martin has um so one thing i'd also point out alex is i do think that you're going to have a lot more acquisitions going on um partly because of the sas apocalypse right uh the re-rating of enterprise software companies and you know you look at meritex index um it's i think ford revenue multiples are like at 3.6 3.7 times and and to nikhil's point you have good companies that are trading at substantially lower valuations i think it's part of that fear that these legacy sas and enterprise companies aren't going to make it because of ai um you know you have a rule of 40 company like monday monday.com trading at two times i mean that's crazy uh it's legitimately insane every time we see a revolution in how

SPEAKER_117: software is built and used one company ends up owning the infrastructure that everyone else depends on the next great platform is being built right now that company is digital ocean and they just launched their ai native cloud this is not just another place to rent gpus or a hyperscaler overwhelming you with features and services but leaving you on your own to patch it all together no we're talking about a full stack platform that comes pre-assembled and that sends you just one bill at the end of the month work auto runs a trillion automated workloads on digital ocean with 67 percent lower inference cost 79 lower latency and it's two times faster to production if you

Chamath Palihapitiya: want to understand what building on a true ai native platform looks like go to do.co twist that's do.co twist start building on the digital ocean ai native cloud today and cut your ai work low cost by up to 50 percent that's do.co twist yeah but you could also see how ai could wipe them out so like

SPEAKER_08: what we have you know just very recently intercom was acquired it's a portfolio company of freestyle it returned more than the entire fund it's an interesting case study because it's a company that has been around for a while people thought it was dead because of you know sierra and other companies like that um but i think they were they actually had a really good story about implementing an ai native solution and then getting traction on it so they kind of revitalized the company and that's actually what this sort of messy middle of like sass and enterprise companies need to do i also think that salesforce acquired them because they wanted owen as a senior executive yeah and they were not going to buy sierra didn't bring brett taylor back and so you have also companies like air table which were darlings you know they they're they're they're their last round was at 11 billion um and which by the way freestyle was able to top tick as a secondary um but you know a company like air table other legacy companies they have the customer base they have the workflows they have the proprietary data and so i think that's actually what um what's going to drive a lot of acquisitions down the road uh as well as senior executives you know i mean i think mark benioff would love to have uh you know is excited to have owen as as a senior exec he would probably be very excited to have howie from air table as as a senior exec so i think you're gonna see that as well as the the technical acquisition the acquisition of technical people that you know the smaller acquisitions are representing

SPEAKER_35: all right so we talked about workflows there we talked about data let's get into that so nikhil we were talking about this before we jumped on the show there's been a a host of posts from intercapitalists and founders in the last couple of i want to say the last week digging into the concept of data now if you go back to the early ai era you know right after uh chat gpt came out people were saying oh data is so important on the training side of things to create these models

SPEAKER_00: startups were spun up to facilitate data licensing that kind of went okay as far as we can tell then for a while everyone just talked about intelligence and uh reinforcement learning and

SPEAKER_35: so forth and now finally we're back to where we were before which is the value uh of data often

SPEAKER_00: locked behind um corporations just not really accessible to the world so i'm curious if you can tell us why the conversation has shifted from the founder and venture perspective to data being so important today and then also what that does for companies like monday uh which michael just pointed out has a lot of this but is trading out essentially zero for how much revenue it does every few

SPEAKER_20: months it feels like there's a different uh you know substrate to the ai revolution that um the the world gets excited about and so you know when you think about the stack right it is uh it is energy compute uh and data that that are the fundamental substrates for these models um and and the models are sort of the fundamental substrate for ai and it feels like you know the market gets really excited about you know energy and then compute and then data in in different moments and right now it's kind of a loop exactly data is having its moment in in you know in the in the x uh eco chamber uh with a number of people posting about it um you know i think the interesting thing about data is uh whereas in energy and and in compute you have enormous companies enormous public companies right like the biggest public company in the world in the case of nvidia for example um you don't have on the surface uh such enormous companies that are just you know uh in data and so um that is an interesting opportunity now of course you have uh every company uh that that sits on data that that that could be important um and relevant to what's happening in ai so that's just one dynamic to think about which is that the company is like purely focused on data um there aren't that many of those that are huge public companies versus in the other couple key areas and so um you can argue that there's like more white space there than in other areas and then it does feel like what the labs are really focused on is is data at the moment uh you know they've been um they've been uh you know there've been a number of projects uh underway at the labs to solve the energy and compute um challenges and those are those have been more publicly talked about but what's happening on the data side is more murky and therefore i think like interesting for people to be weighing in on um and the companies out there beating their chest about data are particularly the ones in data labeling um macor uh handshake and others and it's interesting that they constantly talk about like their revenue run rate and where that's going but that's not the only piece of data that is interesting um there's also data licensing and the model companies do a tremendous amount of licensing of data um not just labeling uh uh of of data through humans um data licensing applies for example to real world data that they're trying to get access to and then of course the models themselves have data from usage that they're using to improve their models so those are just some of the parts of the stack we have one of our one of our portfolio companies that's growing the fastest is a company called protege which is in that data licensing realm um um and then you know i think again what is intriguing about this whole space is how little we actually

SPEAKER_129: hear about from like how companies and the model companies themselves are leveraging the data to

SPEAKER_35: improve their models because that is their secret sauce so i'm pulling up protege here so how how much

SPEAKER_00: they accelerated in the last six months because if these people talking about data becoming more important again as we talk about the different layers of the stack i would presume that they've seen

SPEAKER_20: a pretty big uplift in their business yeah this is a company that has again not been that public about how well it's doing but in year two of its business is hundreds of millions in revenue and what they do is they enable um data providers folks sitting on data to license their data to the model companies and application layer companies that have a need for that data and a purpose for licensing that data and so it's a very simple idea on the surface but one that makes a ton of sense when you think about the importance of this in this moment in time you know you've heard about there are public deals such as reddit steel or the new york times uh deal with with open ai but you can imagine that there's uh a lot more of those types of deals that that are happening and proteges emerged as the leader in in sort of facilitating those types of um licenses michael there's been a lot of questions

SPEAKER_35: about where the value will accrue in the broader ai race uh will it be people who just sell gpus kind of the picks and troubles argument clearly there's a lot of money to be made in interconnects high bandwidth memory photonics etc uh the app layer the model layer it all kind of predicates on data so i'm curious what your perspective is on how startups are going to be able to to capture

SPEAKER_00: when they often don't have as much historical customer data as your sales forces and hubspots

SPEAKER_08: yeah i mean that's a really good question i i think having proprietary data is actually the the where the economic value is going to accrue um you know in a way you can look at token economics a token as as a unit of human labor actually uh because ultimately you're trying to complete tasks around that and that's the sort of expenditure that you need to to do um yeah but in order to do anything to have any insight to have anything actionable you have to have the data and that's why i think the sas apocalypse is overblown um you know i think it's it's it's almost hubris to dismiss all these companies that have the workflows that have the proprietary data that have the customer relationships and to just assume that you know ai is going to blow them away ai is actually going to help them leverage what they have and actually create more economic value if they do it right and intercom is

SPEAKER_35: a good example of that intercom is a great example of it they did refactor their whole business rebranded as finn and really went all in early i mean frankly i think owen was kind of stuck his neck out frankly on that point and ended up selling for michael backing up 3.6 billion i think 3.3 3.4 awesome yeah but the the inverse of your point oh it's a great outcome especially for a company that had for a while their stagnant arr and was actually shrinking a little bit like that that is i mean that's coming back from the dead and i say that with respect as opposed to being a diss right um but your

SPEAKER_00: point about sas apocalypse being overblown the importance of data the power of workflows i think about a company that as we all know box with aaron levy they have tons of customer data they're

SPEAKER_35: expanding into workflows they seem to be accelerating their revenue a little bit i track their earnings more carefully than i should and as of today they're trading at a 3.4 x trailing price sales

SPEAKER_56: multiple so is the market just systematically undervaluing data and its eventual value today as the private

SPEAKER_08: markets get it right i yeah i i do think that yeah i mean if you actually look at salesforce they they're growing at 10 to 10 to 12 percent so why why is salesforce that has all this amazing data you know it's a database company right ultimately crm company and they're they're only you know they're they're down 40 for the year um and i think you know it really is that investors are chasing after growth fundamentally and you have that with you know uh samsung and sk hynix and and micron you have um you know i think vcs chasing after companies that are going from zero to a hundred million in six months now not not 12 months um and it's to the detriment of the companies that aren't growing that fast but yet are probably you know very well maybe building enduring companies yeah and i think that's that's actually where the alpha is going to be made um yeah you could get on the the bandwagon and go off and have fomo and chase after the high momentum deals but i think there's going to be uh very thoughtful founders and companies that are building enduring businesses that are going to take advantage of this just as well and from a vc perspective to have good ownership in those kind of companies actually will potentially play out better the fear of course is uh from the lp perspective you know i hear this all the time you know how much exposure do you have to spacex open ai and anthropic that's all that people talk about and it's it's uh it's an important point but i i think they're missing the bigger point of what early stage venture is supposed to do yes which is not

SPEAKER_00: back the super late stage bloating since those companies that we already all know i mean that's not

SPEAKER_24: venture that's just basically ipo investing under the mantle of pe with venture capital stencil in front of

SPEAKER_149: the building like it's not yeah i mean i'm sure nicole has a lot of thoughts around this but you

SPEAKER_08: know and from our perspective what we do is we try to back vcs who are finding non-consensus founders building non-consensus companies and that's where the alpha comes from because ultimately if they become consensus then there's unlimited capital coming in right yeah i i posted this a while ago it's like basically zerp 2.0 instead of the us government giving all that money it's actually the the the platform firms that have pretty much unlimited capital um and that's why we're back to 2021

SPEAKER_00: era revenue multiples i mean like how many vcs were on my phone back then be like i just heard about a series a done at you know 10 000 arr and 10 000 x multiple i'm like that's crazy and then it was a crazy and now i think we're right back to it like yeah yeah i mean i i think you know you look at

SPEAKER_06: nvidia you look at their ford p it's like 24 times it's just not crazy it's not like cisco at 100 in

SPEAKER_08: 1999 and then you know the what's happening though is that e is doing a lot of the work right and you're assuming that that e is going to the earnings is continuing to be like amazing right samsung just announced that they had 59 billion dollars of of ebitda and you know it's just like holy where did that come from yeah you know um i think the the biggest risk over the next 12 to 18 maybe 24 months and no one knows is that suddenly that that growth is not going to happen there's a a lot of catalysts in the market that could create major issues and you can easily see nasdaq going

SPEAKER_00: down 20 percent easily okay before we let nikhil drop in i want you to tell me uh i have my own list of catalysts that could lead to such a correction which i don't think actually would be that a miss

SPEAKER_61: but um what are you looking at as possible um tripping points well you know i think some of the

SPEAKER_32: catalyst could be you know clearly like if nvidia or any of these semiconductor companies have a miss

SPEAKER_08: if you have projects like um you know the the the star cluster that oracle's working on with open ai if they you know their bonds are being have even a higher premium now in order to to so it's really the financing risk being able to raise the capital in order to do this because it is a circular economy and um you know if any part of that circle stops you know the the musical chair stops then you can have a serious correction because then all the growth estimates are off the table and that's going to be a catalyst it one one world that i don't know very well is private credit and a lot of that is funding these data center build outs and i think you know if there's any hiccup there the other thing of course people are talking about recently is like these leverage etfs right the three times leverage etfs of these semiconductor companies you can see where if that crashes then retail takes down all the hedge funds that were long and and suddenly you'd have contagion we've seen this movie many

SPEAKER_00: many times over the past you're not supposed to say contagion just like you're not supposed to say recession it's bad luck i mean you're gonna you're gonna curse us all all right nikhil we've been talking for a minute let's bring you back in you know the phrase that i think is on my mind

SPEAKER_20: as i think about what happens in the markets is you know it's not too big to fail as it was in you know some some prior corrections but it's too important to miss like there's a set of companies that just have to uh hit their numbers beat their numbers on both top line and bottom line yeah i actually worry more about uh the bottom line for a set of companies that are just consuming a tremendous amount of capital um and obviously i i think like open ai is high up on that list i worry more about that than i do the top line of a company like nvidia for example which i i think you know has a

SPEAKER_167: lot uh uh of predictability to their their business i think we've also seen predictability baked into

SPEAKER_87: the memory sector for example michael mentioned a couple of memory companies i think micron was

SPEAKER_35: the one in its last earnings report that noted they dropped in like uh multi-year contracts for some

SPEAKER_00: of their biggest buyers a lot of stability there the thing that i'm looking for the most is a decline in the compute crunch at the hyperscaler level so i think that will tell us where capex is going from the biggest buyers it'll tell us how much you know they're spending on their own ai uh projects

SPEAKER_35: and also customer demand so it's kind of a good proxy for overall health and i think we are we are one major like alphabet pulls back from its capex plans away from at least a 10 correction which i think puts a lot of pressure on the upcoming earnings cycle because once again here we are everyone's worried and every single quarter so far every hyperscaler said we're compute constrained it will

SPEAKER_170: be for quarters to come and one day that won't be true and that's i mean alex one one one really

SPEAKER_41: recent example and i mentioned samsung reported like 58 billion dollars of evita yeah up from like

SPEAKER_08: a billion eight or something like some astronomical increase year over year it actually the stock price went down so eight percent seven or eight percent it went down and that's sort of the sentiment you have in this market is like you know that's great but you know uh you you have to keep outperforming and that's that's what's worrisome is that unrealistic expectation um you know every time nvidia reports i'm like very very nervous because even if they have an amazing earnings report and then nikhil correctly you know points out they have really stable growth uh if it's not exceeding you know sort of market

SPEAKER_53: sentiment and expectations you could have a disaster what's the private market version of this because

SPEAKER_00: uh the latest news is that lovable everyone's favorite european hyper corn uh is looking to raise i think it's 300 million more at a 13.2 billion they've crossed 500 billion in annual run rate not annual recurring revenue um to me that's incredibly impressive company but i mean uh nikhil if

SPEAKER_07: they miss two quarters in a row what happens to their valuation i think the thing that worries me

SPEAKER_20: is less um them and other companies missing on the top line it's just a set of companies that require a ton of capital uh because of their their burden levels i think the ones that are scary are like you know reports that open ai needs to go raise another several hundred billion dollars um uh that

SPEAKER_181: is the thing that worries me much more so than okay but but but isn't there a direct connection here

SPEAKER_35: between ability to finance that bill that you're describing and their revenue growth because you know once anthropic hit like 64 billion run rate whatever you can kind of just double that for next

SPEAKER_00: year and then they'll have at least 100 billion in revenue in 27 or you can do fun math that way it's not that hard but that does cover a lot of spend if there's a reasonable amount of margin on those tokens and if they're not making hella margin on fable at uh 50 bucks per million output tokens then

SPEAKER_20: the whole industry just throw in the bin yeah i think that's the key question like how much are these companies over investing in compute to capture demand and how expensive that is and then what their margin is and um and so uh again we we shall see but right now uh the music feels

SPEAKER_35: like it is very much playing music is blaring but it tends to blare before the speakers blow out now

SPEAKER_00: on that point i want to just talk about chinese ai models for a second because i do think they fit into the margin conversation uh going back two months suddenly everyone stopped paying subscription for ai and started to pay on a usage basis everyone freaked out seems to have quieted down a little bit now that everyone realized you shouldn't use opus 4.8 fast for everything you're doing simple enough but a lot of companies did lean into either using off the shelf openweight models usually from china some from france uh or also fine-tune your post training their own versions of them we saw examples like cursor using kimmy k2.5 i think uh airbnb used quinn the list goes on uh recently news was out this week that the chinese government may preclude the release of future openweight models uh essentially kind of the same fight we're seeing in the us about accessibility kind of cutting edge ai and my my question to kill is what does that do to startup margins because a lot of companies were moving their inference off of state-of-the-art closed source to either you know cheaper openweight or self-hosting

SPEAKER_35: them frankly and if they can't do that what happens to their business we have seen uh our own portfolio

SPEAKER_20: companies use a mix of models and you know i think many of them are using the frontier models for coding um and so that is a significant area of spend for them um but but they're also using the open uh the open source openweight models uh for uh elements of their product where the frontier models are just not necessary now the the the good news i think for them is that the non-frontier models at the close source um companies are getting cheaper and so you know you can redirect a lot of stuff to some of the older models that are still very capable and i think that is an option for a set of companies that um that again helps the the margin conversation so i don't you know you know i i think obviously a key thing for for us uh in the us is figuring out how to have a vibrant open source ecosystem open weight uh set of models here i think that's really important um but i think what is plugging the gap is that uh you know the the non-frontier models from the closed source companies are just getting cheaper um and i have to believe that what will happen is uh you know more of a stratification there where of course the frontier capabilities get expensive and uh and will be needed for a set of tasks but i think we can redirect a lot

SPEAKER_88: of toss these these these models that are cheaper michael that doesn't solve the whole problem though i

SPEAKER_35: don't think because there's been a lot of conversation we saw alex gar from palantir and a lot of other people begin to really beat the drum about not using closed source models be they you know at the

SPEAKER_00: absolute cutting edge or a generation behind or just a sonic size model because the major labs are going to train on your workflows train on your data and essentially put you out of business so in a world where we don't have open weight chinese models and you don't want to trust the closed source models from major labs where does that leave startups yeah i mean i think ultimately it gets back to the

David Sacks: my comment about enduring businesses and you know the joke was ebitda plus c you know the cost of compute and uh you know economically speaking you you you want to take advantage of of of that pricing delta

SPEAKER_08: between the open source models and the proprietary models um i think it's really good for the us ai industry i mean i think that kind of pricing pressure is is really good you see a lot of startups working on orchestration uh uh layer and and trying to figure out how to optimize which models to use when and i think that kind of innovation is very important and so i don't have any more insight other than that uh but i do think from an economic you know the the free hand of the the the invisible hand of the free market i think it's really good to have that pricing pressure put on the proprietary

SPEAKER_56: models oh i agree i mean shout out adam smith and all sorts of you know invisible appendages but

SPEAKER_35: i'm just worried uh nikhil that yeah the models from nvidia the nebatron family that the models from uh google's gemma family aren't sufficient to replace what we have from deep seek from zipu from moonshot from quinn and we're going to end up in a place where a lot of people are betting on rolling

SPEAKER_00: their own especially startups that don't want to pay opening eyes marching for them and they're just

SPEAKER_35: not going to have something to fall back on and to me that seems like a an almost structural risk

SPEAKER_20: in the startup market today yep yeah i think you're right to be concerned about it um uh and i think the the other thing here is just uh all of this requires teams that are capable enough of uh of leveraging different models of switching models in and out of you know potentially fine-tuning building their own model and um and i think uh as we discussed earlier like the set of really talented people is finite and and and so there are a set of companies that just don't have another option but to um but to to work with either application layer companies that are using the closed source models or um or just work with the the frontier labs because they they don't have the talent level uh to do

David Sacks: anything else well alex the other thing is that you know ultimately you can see where different industries have you know are leveraging small language models right and so that you know if if there are companies that were helping develop these small language models for an enterprise customer they that can actually take advantage of the data that they have and the workflows that they have that would be probably more cost effective than just you know using anthropic or uh uh or open ai i'm just

SPEAKER_35: disappointed that we've now said several times in this conversation that there's only so many companies that matter so only so many founders that matter only so many researchers it seems

SPEAKER_00: disappointing that we seem now back to nikil's earlier point about being more power law pilled that the number of things that matter in the market companies founders etc is going down it feels like at a time in which it's more easy than ever to build something and to me those seem to be contrasting in a way that doesn't help the current political situation around ai being relatively unpopular so you know is there a way to i mean i'm going to sound like uh the mayor of new york city here but hear me out to uh spread the love a little bit and have more companies become market leading so that way we don't end up with just like six people from andresen and their friends making a bunch of money and everyone else sitting around with a tin can going please lord give me some tokens

SPEAKER_20: i think two things can be true right like there's um there's a concentration of resources and power the bigger companies getting bigger can can coexist with it is easier than ever to start a company and it is uh it is more possible than ever for two people to build a business that is uh that gets big quickly and gets wildly profitable quickly like i think both of those things are happening before our eyes and and so it is an amazing time to go build um and you know it is an amazing time to be one of

SPEAKER_22: these massive companies that is only getting bigger yeah you know alex one of the things that we've had

David Sacks: uh an idea around is this arms race to find younger founders and in an ai native world you know everybody

SPEAKER_08: needs to be the first check and that also could mean younger founders and you look at you know um some of the some of the groups that we work with like josh browder who is a teal fellow he's helps on the selection committee we have corey levy at z fellows we have um the guys from prod you know you look at companies that for example are like neo you know they're spending time with university kids and especially at harvard mit and stanford you have and and you know carnegie mellon and some others you have kids who are uh who actually want to start companies who have are almost fearless like what uh you know one of the companies that we're in is is etched and you know these young people are like hey we're going to totally disrupt nvidia and build it inference specific chips and you know what what 20 year old goes around thinking that but they're unencumbered by the fear and they're not 20 year veterans thinking oh it's an impossible task and so i think that that's actually what's very um you know heartening and very exciting about entrepreneurship so yeah you do have these big incumbents they're getting bigger but you also have really smart kids who are fearless and are trying to build companies

SPEAKER_194: that are totally going to disrupt and that is the beauty of of capitalism uh gavin is the ceo of etched

SPEAKER_35: i believe right yeah yeah we had him on the show one of my favorite conversations i've ever had one of the nicest people i've ever met and also they just came out of stealth announced michael correct me

SPEAKER_00: here 800 million in funding and they're bringing a trip to market soon they're building a test data center of a couple megawatts in taiwan and they're going to bring out rack scale systems yep i forget the time frame but soon which i think is gonna be great for cutting inference prices overall if you're not

SPEAKER_35: into cerebrus or sanvenova which just raised a billion at an 11 billion dollar post uh guys

SPEAKER_00: bringing this to a close i want to do some kind of fun questions here and i love to go through portfolios and pick one out and then ask the uh the venture firm in question about why they picked that one now windborne as far as i can tell is a company that wants to put a bunch of balloons up into the sky and provide essentially a private weather observing network and then sell that data to energy traders and everyone else i i love this idea but nikil it does seem that right now we've seen

SPEAKER_35: headlines about how we've reduced funding at the national level here in the states for our own

SPEAKER_00: weather gathering um uh technology so is this company just nailing the right time right market moment because i feel like they must be just fending off customers yeah it's it's uh it's one of the

SPEAKER_99: companies we're most excited about and um and some of the characteristics here i think are interesting to

SPEAKER_20: come up for founders which is windborne uh collects its own data through its own weather balloons so um it it has proprietary data on the atmosphere in their case they leverage that data to build their own models and and they use ai for those as well and so ai plus their own proprietary data has led to some of the most accurate weather forecasting models in the world now for this company um and this is by the way you know it's a 50 person company based in uh in the bay area um and then on the commercial side uh you're right that unfortunately the cuts in the national weather service in the u.s and um issues with uh um these atmospheric associations around the world have led to uh in a time where like the weather is changing more and where you would think um you know models should get better unfortunately um there there have been a lot of issues with uh with accurate models with accurate weather forecasts and so windborne is one of those companies that's filling that gap and they sell both to governments and to uh and to companies and by the way within the u.s government for example um not only do we work with the national oceanic and atmospheric association noaa we also work with the department of war um and and you know you can imagine there is lots of uh reasons for why weather data is important to both those departments um so really fascinating company it actually is related to the the data conversation we were having earlier because of its own its own data uh edge and moat there um and the one other thing i'll call out about this is the company was actually started back in 2019 um the the four founders all went to stanford together they were part of the stanford space industries group at stanford um so well pre this ai era but it's one of those companies that now feels more interesting than ever based on what's happened in ai the last three and a half years and the company itself is one of the most ai-pilled companies that we work with they've automated everything possible internally they actually have their own ai lead uh software lead at the company that's managing their their fleet of agents and so uh a company that feels very well set up for the

SPEAKER_35: next 10 years you get 10 points for the answer and minus five points for dodging saying climate change during your response but still five points not bad all right michael now over to you i don't i'm not going to press you about a particular company but i am going to say who is your favorite fund manager

SPEAKER_08: in your broader portfolio and you cannot say nikhil i love nikhil um right now i would say josh browder i mentioned him earlier he's a fellow he has a company called do not pay it's very profitable he's and but you know he has hustle his average post money is five million dollars he's on the team his he's on the teal selection committee and so he meets you know these 200 kids who are uh in the final process and that's an amazing pipeline of of potential entrepreneurs so you know i think josh has done a very good job harvesting that um and i feel like uh you know that that he he i wouldn't say he's neurodivergent but he can read people very very well and i think he sees the spark in young people who are actually going to walk through walls to build amazing companies that's the problem right now with these younger founders a lot of them are actually just trying to get the credential you know nothing against yc but you know getting into yc is kind of a credential now or becoming even a teal fellow is kind of a credential so you got to really suss out who's going after uh what for what reason

SPEAKER_35: and i think josh is extremely good at that so essentially in the zerp era we had a lot of tourist

SPEAKER_00: money flowing in and today when you can grow a company from zero to 100 million revenue in a year we have a lot of tourist founders okay well you know what neither will last because if there's one

SPEAKER_35: thing that's true it's building a company it's hard as hell and it takes a really long time uh guys we're gonna leave it there but i really enjoyed this i'm really curious to see in six months how this data conversation changes and where we are in the broader arc of uh going around in a circle

SPEAKER_170: between energy compute and data uh but nikil where can people find your firm online and is there uh

SPEAKER_222: anything else you want to shout out before you go yeah just footwork.vc and i'm at nbt on on x and nbt.substack.com is where i is where i write as well fantastic michael uh where can people find

SPEAKER_08: sindana and anything else you want to add it's uh sindana capital.com and my uh twitter handle is mk rocks which was my gamer tag from the early 90s and i use it for everything mk rocks we're not going to stop now then uh what are you playing lately uh call of duty at one point i was top 10 in the world on that um but uh i don't play as much but yeah i like first person shooter games i i'm an old quake

SPEAKER_35: guy myself a big fan of doom 3 but i've really branched off into factory automation games lately i can't help myself right so good uh nikil do you game oh man i used to but uh now like in in high

SPEAKER_20: school uh i played uh warcroft 3 and oh yeah i was high us west with one of my best high school friends but in this current zone where i have two young kids and started a firm uh it would be a bad situation if i was gaming as well so yeah i'm off you just have to do it after everyone else is asleep

SPEAKER_59: and just cut back on your sleeping now it works it works every time the time to write and think

SPEAKER_160: these days so all right guys this has been an absolute treat this has been this week in startups my name is alex we're back on friday we'll see you all then thanks for watching this week in startups

SPEAKER_240: if you liked this episode check out more if you're a startup founder founder university cohort 13 kicks off this fall it's a 12-week program that provides guidance on building your product launching to real customers and pitching to investors top startups receive 25 000 or 125 000 in investment apply now at founder.university twist already have traction the launch accelerator invests 125 000 and connects you with 500 plus investors to help you raise your next round apply at launchaccelerator.co if you're an accredited investor looking to gain access to quality deal flow apply for jason's angel syndicate at the syndicate.com we find two to three deals a month and check out this week in ai jason's experts only roundtable with top ai founders and operators every week find it this week in ai.ai check out the twist ticker our daily newsletter at this week in startups dot com slash ticker thanks again to our sponsors for making today's show possible follow the show on instagram follow the show on x.com this week in startups publishes three days a week monday wednesday and friday at 5 p.m central time you can submit an audio or video or video file question by emailing it to this week in dot com