Crypto’s Nasty Downturn Worsens, SpaceX IPO Hype Halo Effect, Selling AI in Regulated Industries

30 Mar 2026 · 44 min · 21 chapters

Ask about this episode

Ask anything about it. ChatGPT or Claude reads this page and answers with the times it was said.

Connect VO and ask about every podcast you hear, including the moments you saved. Add to ChatGPT · Add to Claude

In short

Crypto market downturn and pivots; SpaceX IPO “halo effect” lifting space stocks; venture investing themes (token/data acquisitions); and how to sell AI into regulated industries (compliance, trust, FedRAMP).

Guests (backgrounds)

  • Yuechi Yang, crypto reporter at The Information.
  • Theo Waite, Elon Musk reporter at The Information.
  • Ashe Songvi, partner at Haystack (early-stage software); deals include Exa Labs, Ceronic, and Lilac (acquired by Databricks in 2024).
  • Cheryl Cheng, Managing Partner at M12 (Microsoft’s venture fund).

Key claims

  • Crypto is down ~50% from its October peak; ETF inflows reversed; layoffs and pivots underway; biggest crypto risk is a regulatory “swingback” if a crypto market-structure bill fails.
  • SpaceX IPO hype is boosting rocket/satellite stocks; SpaceX may still dominate, squeezing others.
  • AI “big labs” may buy startups for unique data (“unique tokens”).
  • Regulated-industry AI adoption is slowed by data readiness and workflow fit; FedRAMP is a major federal barrier.

Notable examples

  • Crypto pivot: a company moved from Bitcoin yield to an AI-powered crypto trading tool.
  • Prediction markets compete for retail capital; bets can be on Bitcoin price in minutes.
  • Space stocks: Rocket Lab (Electron/Neutron), AST SpaceMobile (hype, low consumer service), EchoStar (HughesNet users down 50%+; sold spectrum to SpaceX), Globalstar (iPhone SOS satellite features; value tied to spectrum/possible acquisition).
  • FedRAMP: Knox Systems can enable FedRAMP in <90 days at ~10% of typical $2–$5M cost; only ~500–600 companies are FedRAMPed.

Written by AI. May contain mistakes. Listen to the episode to check what was said.

Chapters

Tap a time to open that second in VO

The Crypto Market's Prolonged Decline

0:45 to 2:59

Discussion on the ongoing downturn in the crypto market and its causes.

“Our Elon Musk reporter wrote about that possibility.”

Crypto Founders' Strategies

2:59 to 6:36

Insights into how crypto founders are adapting to market challenges.

“One company that I talked to was pivoting just last month from offering Bitcoin holders a yield product to providing an AI treatment trading tool.”

Venture Capital's Role in Crypto Investments

6:36 to 10:24

Exploration of venture capital trends and challenges in the crypto space.

“Now, you spoke with an executive at Coinbase involved in their institutional business.”

Regulatory Risks in the Crypto Market

10:24 to 12:39

Discussion on the potential impact of regulation on the crypto industry.

“I mean, the banks kind of have even more ground to stand on, right?”

SpaceX's IPO and Its Market Impact

12:39 to 14:02

Exploration of SpaceX's IPO and how it affects related companies.

“As SpaceX marches towards its IPO, other public companies that play in its ecosystem are also getting a boost in the stock market.”

SpaceX vs Incumbents: Market Dynamics

14:02 to 15:09

Explore how SpaceX's valuation compares to traditional aerospace giants.

“Their stock prices haven't jumped as much.”

Rocket Lab: A Growing Competitor

15:09 to 16:26

Learn about Rocket Lab's growth trajectory and its competition with SpaceX.

“So let's talk about SpaceX's rocket launch business then.”

Broadband Innovations: Starlink vs HughesNet

16:26 to 18:01

Discover how Starlink is disrupting traditional satellite internet providers.

“And what about in the broadband and in the mobile business for Starlink?”

GlobalStar: Spectrum and Acquisition Potential

18:01 to 19:36

Understand GlobalStar's market position and acquisition interests in telecom.

“GlobalStar, similar name, totally different company.”

Financial Performance vs Investor Interest

19:36 to 21:27

Analyze the financial health of satellite companies amid rising investor hype.

“I mean, some of them are growing, but not a lot of them are profitable or anything like that.”
Show all 21 chapters

SpaceX: Industry Impacts and Future Leaders

21:27 to 22:40

Evaluate whether SpaceX will dominate or if competitors will emerge in the space sector.

“The industry experts that you talk to, and I realize that I say industry expert, I mean, the industry itself is kind of being built as it's flying, as they would say.”

The Big Token Thesis: Startups and Acquisitions

23:32 to 25:15

Learn about the strategic thesis behind acquiring startups for unique data.

“I want to talk about a blog post that you published nearly a year ago now.”

Investing in AI Infrastructure: Insights from Ashe

25:15 to 27:29

Explore Ashe's investment approach in AI infrastructure and its implications.

“the way i was thinking about that thesis in particular when you say unique tokens what do you mean by that um i mean like things like robotics data things like um collecting video game data to help train robotics.”

From Furniture to AI: The Entrepreneurial Journey

27:29 to 28:00

Examine how Ashe's experience in furniture influenced his venture capital approach.

“Which is a big, I mean, furniture to AI infra is, I don't know, I guess building blocks, you could say that much.”

Insights from Building a Furniture Company

28:00 to 29:06

Learn how experiences in a unique industry can translate to AI investment strategies.

“and the business has scaled a ton since then.”

AI Adoption Challenges in Regulated Industries

29:06 to 30:24

Explore the barriers to AI integration in sectors like healthcare and finance.

“That is Asha Songvi from Haystack here on TI-TV.”

Understanding FedRAMP Compliance

30:24 to 33:11

Understand the importance of FedRAMP compliance and its impact on software sales to the government.

“And there are so many complex workflows that they absolutely need AI to help with that.”

Streamlining FedRAMP with AI Solutions

33:11 to 34:30

Discover how AI can expedite the FedRAMP compliance process for startups.

“And that is very important because the federal government is the largest buyer of software in the United States.”

The Risks of Agentic Systems

34:30 to 37:19

Learn about the potential risks of AI agents and the importance of internal controls.

“So are you saying that the, so, I mean, it's cost around$5 million, you said, to get the FedRank compliance.”

Future of AI Compliance and Security

37:19 to 42:00

Understand the ongoing development needed for AI compliance and enterprise readiness.

“So I think you're talking about two very different things.”

Challenges in AI Adoption for Regulated Industries

42:00 to 43:47

Explore the current gaps and challenges in AI adoption within regulated industries.

“But so when is that going to happen, is my question.”
Hear the part that matters, and keep it.Open this episode in VO. Double tap your headphones to save a moment as you listen.
Get VO free

Transcript

Automatic transcript. May contain errors.

0:13Welcome, everyone, to The Information's TI TV. My name is Akash Pasricha. It is Monday, March 30th. We are kicking off today's show with the latest on the crypto winter. Bitcoin has lost half of its value since peaking in October. The information's crypto reporter published a story about how companies in the sector are reacting. Next up, SpaceX's eagerly anticipated IPO is boosting shares of rocket and satellite companies, but those same companies could be also at risk if SpaceX expands its dominance. Our Elon Musk reporter wrote about that possibility. He will join us shortly to discuss that.

0:51We'll then shift over to the venture capital world. The information published our next GP's list. It is our view of the folks best position to lead their firms down the road. Today, I'm speaking with Ashe Songvi, a partner at Haystack, one of the partners we highlighted in the project. And to close out the show, I will speak with an investor from M12, Microsoft's venture fund, about selling AI into some of the most highly regulated industries in the world. It's going to be a fun show. So let's get right on into it. The crypto market is nearing its sixth month of prolonged decline, coming after what initially was an extraordinary moment of euphoria when President Trump was first elected.

1:32Now, the sector is trying to broaden its focus to more than just markets. Yuechi Yang, our crypto reporter, published a story on that today, and I want to bring her on to talk all about it. Yuechi, welcome to the show. It's great to have you here. Hey, Akash. So why is crypto still there? It's been six months. Yes. Well, the sell-off has no ending side. There are macro reasons and there are crypto-specific reasons. Broadly speaking, there is a market sell-off in risky assets, including stocks, because of the war in Iran, which introduced uncertainty with political situation, with potential economic snowdown.

2:13And then within crypto, I would say there's a crisis of faith. People are pretty bearish right now, generally speaking. It seems like the market has run out of positive news to bet on. The Trump-induced bull market has run out of steam. The ETF inflows has reversed. And then across the industry, you're starting to see that layoffs are starting to happen. And then some of these crypto companies are pivoting their business model to try to make money from services outside of crypto trading. So this is what you were talking about last time on the show. I mean, you came on and I asked you, what's the solution?

2:53You said crypto needs some kind of a catalyst. And there's really no catalyst in sight here. It needs some good news to lift it up. How are the founders reacting? so uh for crypto founders this is a hard time uh especially since last year a lot of them just raised money so now uh the few people that i spoke to they're saying that crypto founders are trying to survive and they're not going to just sit and wait for the market to turn they're trying to figure out new ways to take advantage of the money that they raised and then and provide services that could be still of useful. One company that I talked to was pivoting just last month from offering Bitcoin holders a yield product to providing an AI treatment trading tool.

3:46So they've pivoted from crypto lending to offering an AI product? AI, yes, AI powered tool for crypto traders to trade on crypto assets as well as non-crypto assets. And so what about the venture capitalists? Are they getting excited about these AI-powered crypto trading tools? Are they investing any money at all? We, of course, we've seen a couple funds. Well, really one fund. I'm thinking of Dragonfly Capital. They raised$650 million. We had Tom Schmidt on the show. But, I mean, that's like one fund. I don't see anybody making any big deals anymore. It is a hard time for crypto venture funds to be raising money.

4:29And then for the money that they are deploying into the crypto startups, that has slowed down dramatically as well. We have data showing that in the first quarter, private fundraising for crypto startups has fell to about$2.9 billion, and that's a 69 % down decline from the prior quarter. within the crypto fundraising ecosystem, there are a few bright spots. I will say that companies that are pivoting to AI-related businesses, such as Argentic Payments, they are raising money. Stablecoin payments companies are still raising money. And then there are also this whole genre of companies that are trying to bring real-world assets, which are traditional assets that are not crypto, onto blockchain and try to sell it to crypto investors.

5:25And we see fundraising deals in that sector as well. How do you contrast what is happening in the crypto markets with all of the excitement around prediction markets? Because this is my hunch, and I have no evidence for this, but it feels like it could be a similar crowd of traders that was really attracted to crypto, now pivoting to prediction markets. We've seen CalShit and Polymarket do some big things in the way of fundraising themselves. Do you see prediction markets and crypto markets being competitive in any way at all? Yeah, they're definitely competing for retail investors' attention and capital.

6:08I think I would say both markets are still very much retail driven. And retail investors now are moving on to prediction markets. And even prediction markets, some of the most popular bets are related to crypto prices. For example, you can bet on whether Bitcoin price will go up or down in the next five minutes. And so there's crypto elements to it. And they are indeed competing for a similar group of traders. Now, you spoke with an executive at Coinbase involved in their institutional business. Tell me a little bit about the questions that you had for him. who this executive is and what they oversee and what you ultimately gleaned from that conversation.

6:51Yeah, so I talked to Coinbase executive who leads their institutional strategy, and I asked him what is the biggest risk that would concern him right now in the crypto market. And then he gave me this answer that he thinks the biggest risk, in his opinion, is the massive regulatory swingback. And that could potentially happen if the crypto industry fails to get a crypto market structure bill passed this year. And if there's no legislation that will provide a legal framework for how crypto tokens are classified, how crypto businesses will be regulated, then when there's a change of administration, it is possible that you will see a new administration taking a harsher stance on crypto, similar to how the Joe Biden administration treated crypto.

7:44So he thinks that if that happens, that will be a potentially systematic risk for the broader crypto market. And just remind us, this crypto market structure bill, you referenced it a little bit in your story. This is the bill that has caught up in Congress, and it's one that the banks and the crypto industry have found themselves on separate sides of. Walk us through what the disagreement is here and also what the stakes are for the crypto companies? Sure. So this is the second piece of crypto legislation that President Trump is trying to pass after he passed the stablecoin legislation, which was the first piece last year.

8:22And then this legislation is about providing regulatory framework for the rest of the crypto market that are not stablecoins. However, there is one element that is still very much being debated And that's whether stablecoin can pay yields to the platforms that are distributing them, such as crypto exchanges. Now, that's an issue that's highly contentious right now, especially between the banking side and the crypto side. The banking side thinks that if crypto holders can earn interest from stablecoins, from crypto exchanges, then that will potentially cause deposit flight from the banking system.

9:07And then on the other side, crypto industry is largely speaking, hoping that they will be able to pay interest for stablecoin holders in the secondary market because that will just make stablecoins more attractive. You will be able to earn yield by using stablecoin. That's a great incentive. That does matter for people who are debating whether they want to use stablecoins or not. So that's an element of the bill that's holding up the whole discussion right now. We're seeing negotiations happening behind the scenes, and we're watching whether the bill will end up passing this year. Well, and so let me ask you this.

9:49So in this environment where crypto markets are continuously coming down, does that or could that impact the way that this bill goes in terms of which side has a leg up in the argument? because I'm just trying to think about this, you know, logically here. I mean, if people are saying, hey, we want to be able to earn yield from our stable coins or from our cryptocurrency, and the banks are saying, we don't want that because, you know, this is directly competitive with what we're offering via our own deposits. I mean, I'm just trying to think if crypto prices are coming down, I mean, the banks kind of have even more ground to stand on, right?

10:31saying like you know uh the yield i guess could conceivably depend on the crypto prices or i help me connect the dots here a little bit yeah i think uh for the banks they do wish to be able to pass a crypto market structure bill as well but ideally with a provision that will ban crypto exchanges from passing yield to stable coin holders and that's because without passing the bill, the current status quo is crypto exchanges like Coinbase are able to pay yield to stable coin holders. And so if they want to change the status quo, ideally you should try to push for the law to be passed. But then on the other hand, the market conditions do matter.

11:16It matters more for the crypto side because without a Crypto Clarity Act being passed, big institutions and traditional firms are unlikely to fully embrace crypto because there's still huge legal uncertainty with most of the tokens that are circulating in the market. And there's also legal risk for anyone who's running a crypto platform, who's operating a crypto business here in the US. So those uncertainty would deter big money investors from embracing the market. So the crypto side, I would say, is highly motivated to get a deal passed, ideally before the midterm happens. Right, because like you're saying, I mean, this is sort of what I'm getting at, is it feels like all of these bills getting passed, I'm wondering to what extent they are tied to crypto prices continuing to go up.

12:13Because if the answer is they got to keep going, then, well, then that's not a very, you know, we're getting further and further away from the underlying technology being a better alternative to the financial system. It's just, hey, it's token prices going up and that's the logic for embracing it. I don't know. It's all a very dynamic situation. Yuechi, I want to thank you for coming on. That is Yuechi Yang, our crypto reporter here at The Information. As SpaceX marches towards its IPO, other public companies that play in its ecosystem are also getting a boost in the stock market. My colleague Theo Waite wrote at length about those companies this past weekend.

12:53I want to bring him on to talk all about it. Theo, welcome back to the show. It's great to have you here. Good to be back. So I want to look at some of the companies that you identified in your story that are part of the SpaceX or the space ecosystem. system and we know that SpaceX has the launch business and they have a satellite business and so eventually what I'm going to get you to do is talk about some of the big players outside of SpaceX in those categories but I just want to set the expectation here for people listening I mean are any of these companies actually like significant competitors to SpaceX or are they

13:28Theo Wayt:far and away the leaders right now I mean you know there's some nuance but generally SpaceX is is the winner and the reason this this whole rally is happening is because spacex has kind of changed expectations for what's possible in space and kind of given everything that has space in the name or is associated with space a pretty significant boost okay but we should say that these companies that we're about to talk about i mean they're not like one two billion dollar companies i mean these are companies that are 30 billion dollars in market type at some point and there's i mean their stock prices have doubled and tripled to some extent over the past year yeah that's right rocket lab is worth like 33 billion right now ast space mobile is worth like 30 billion um yeah there's some pretty gigantic names okay and and the last category i just want to get to before we talk about those companies what about the lockheed martin and the boeings of the group because we're not going to talk about them.

14:31Their stock prices haven't jumped as much.

14:33Theo Wayt:I mean, they're definitely, you know, huge incumbents in aerospace, and they're worth a lot of money. I mean, they're both worth well above$100 billion. But, you know, they have gigantic businesses on Earth that SpaceX isn't really trying to compete with them on. Like, SpaceX doesn't want to build, as far as I know, you know, an alternative to a commercial, like a Boeing commercial jetliner or like a Lockheed fighter jet or something. Like they have gigantic businesses, you know, on Earth that are responsible for a lot of their valuations. And, you know, investors aren't really thinking about them in the same way.

15:08Right. Okay. So let's talk about SpaceX's rocket launch business then. In that category, tell me a little bit about Rocket Lab, the company you identified. I mean, how fast is that company growing? How much of a threat do you think that poses to SpaceX?

15:22Theo Wayt:So Rocket Lab is, you know, kind of an interesting company because it's a publicly traded company that does actually launch rockets. And that's kind of rare. They have a small rocket called the Electron that is, you know, does a lot of satellite launches. They're developing a somewhat bigger rocket called the Neutron that's potentially going to eventually be an alternative to the Falcon 9, which is the current SpaceX workhorse, basically. But, you know, they have a long way to go to get anywhere close to where SpaceX is. And they, as far as I know, don't have any kind of alternative to Starship, which is the gigantic SpaceX launch vehicle that's still under development.

16:09Theo Wayt:And in that category, kind of Jeff Bezos' Blue Origin is the only real competitor, like, you know, company that's out there publicly talking about plans to compete with Starship. But obviously that's not a publicly traded company and, you know, seemingly will never be. Okay. And what about in the broadband and in the mobile business for Starlink? Who are the companies that you're watching in those two categories? So in broadband, you know, the idea of getting internet from a satellite has existed for a long time. Like, you know, we've had internet on planes for a while now. It's just kind of been bad.

16:49Theo Wayt:And you've been able to get internet in a cabin for a long time now. It's just kind of been bad. One of the interesting names there is this company, EchoStar, which owns a service called HughesNet, which was kind of a original version of Starlink, basically. Like if you lived in, a remote area, you could get HughesNet and they'd have a technician come out to your house and put a big dish on your roof and then you'd have to pay them and the service would be quite slow. So Starlink has totally destroyed HughesNet. I mean, not totally, but it's in the process of destroying HughesNet. Their users are down more than 50 % over the past few years.

17:30Theo Wayt:um but echo star has found a way to benefit from spacex because the company recently sold a ton of spectrum to spacex and um frequencies that they that they need to operate basically like the right to to operate in a certain way and and they're that means they're going to have billions of dollars worth of spacex equity um when the deal closes which will happen next year but you know It's already trading as if they have this huge chunk of SpaceX equity. And it's interesting because it's become one of the ways that retail shareholders and other investors will try to buy SpaceX stock now is to just buy EchoStar stock.

18:12Okay, so that's EchoStar. What about GlobalStar?

18:16Theo Wayt:GlobalStar, similar name, totally different company. um they they until this until your story i will admit i thought they were the same company i i thought it's it's understandable um they both have kind of boring logos too i'll say um global star is the company that currently powers the satellite features on on your iphone so if you go hiking or something and you need to send an SOS, that's Globalstar that is powering that feature. And Globalstar also owns a bunch of spectrum. And the spectrum that Globalstar owns is kind of the reason it's worth so much money, because there's been a lot of speculation, or more than speculation, reporting.

19:04Theo Wayt:Our colleague Aaron Tilly reported in the fall that Globalstar was interested in selling itself. There have been other people in the industry talking about apple could buy them amazon um spacex like there are lots of big names that could potentially be interested in purchasing global star um and so the the valuation that the company has is like totally tied to that and not to anything that they actually uh currently operate because the satellite service is not very good compared to starlink um and there are other companies that also in spectrum that have also seen gigantic uh you know investor interest recently because uh all of a sudden people are like thinking of it as a um as a big asset so are the financials of any of these companies i mean you talked about acquisition interests as being one vector upon which these stocks have been surging certainly the interest in the spacex ipo So as you said, having sort of some component of space as equity baked into there, but have the financial performance of any of these companies, have they really taken off over the past year at all, or is it pretty much the same?

20:15Theo Wayt:I mean, some of them are growing, but not a lot of them are profitable or anything like that. You know, I think a good example of a company that's, you know, been able to get a ton of hype and a ton of interest is like AST, SpaceMobile, which I mentioned earlier. That's a company that's positioned itself as a Starlink competitor. It's gotten a ton of interest. They're at all these conferences talking about space and doing big press events. And they are developing these satellites. They're launching satellites, but they don't have commercial service yet. You can't, as a consumer do anything with their technology at all.

20:57Theo Wayt:And they're worth$30 billion. And they say they have, you know, over a billion dollars in revenue commitments, but their actual revenue is quite low. So, you know, there's a ton of, there's a ton of interest in the area. There's a ton of, you know, press releases and hype going around. But it's, it's yet to be seen, you know, whether it's possible to build a sustainable business in space. And I mean, And SpaceX is the ultimate example, you know, or the ultimate question mark there. But there's a ton of other companies to look at, too. The industry experts that you talk to, and I realize that I say industry expert, I mean, the industry itself is kind of being built as it's flying, as they would say.

21:40Do they think it's going to be a winner takes all in terms of SpaceX? Do they think there's room for multiple leaders to emerge?

21:49Theo Wayt:I think it depends on how much you believe what SpaceX is saying. I mean, if SpaceX does everything they say they're gonna do, then it's gonna squeeze out a lot of people. But you know, I think there are a lot of people in the industry that really want there to be alternatives to SpaceX for, you know, like partially for political reasons. Like the idea of a single company that's worth a bajillion dollars and is closely tied to the U.S. government, like controlling a lot of communication around the whole world is, you know, potentially, you know, troubling to a lot of countries and a lot of, you know, a lot of companies too.

22:31Theo Wayt:But that doesn't mean that building an alternative to SpaceX like makes financial sense, even if it makes political sense. Right. All right. Well, Theo, I want to thank you for coming on. That is Theo Waite, our Elon Musk reporter here at The Information. This week on the show, we are going to be featuring the investors that we included on our flagship NextGPs list last week. These are the folks who are most likely to become the next general partners running some of the top venture firms in Silicon Valley. Today's guest is Ashe Songvi, a partner at Haystack, which focuses on early stage software investments.

23:09Some of his notable deals include AI search engine Exa Labs, autonomous ship company Ceronic, and data search startup Lilac, which was acquired by Databricks in 2024. Ashe, welcome to TITV. It's great to have you here. Hey, Kosh. How are you? It's great to be here. Doing great. Well, congrats on making the information's next GP's list. It is a fun project that we put together. I want to talk about a blog post that you published nearly a year ago now. It was called Big Token and the Startup Ecosystem. And I was rereading it earlier over the weekend, and it was kind of an interesting thesis. Walk me a little bit through what the message was you were trying to send with that piece.

23:51Yeah, I think that a lot of what I was thinking around was like as value was accruing to the kind of large foundation model labs like Anthropic, OpenAI, XAI, Meta, Google, one sort of viable exit path for startups would be actually acquisition into these labs. And the reason that the labs would buy some of these startups would be for unique tokens, the data sets. And we're seeing some of these acquisitions happen already. And so it feels like a lot of that thesis is starting to play out. But the point that you made in the piece, which I thought was interesting, was the way that big pharma companies, for example, will buy smaller companies that have done a lot of the legwork in the way of R &D that maybe they just didn't have the, I don't know if capacity is the right word.

24:42They just couldn't organize themselves to do it fast enough. I mean, the point that I heard you making was you're basically just buying R &D and that's the way these big token labs will end up operating. Yeah. And I think that like a lot of the function that these startups serve is that they're out in the world collecting unique data unique tokens and that sort of discovery function in the way that pharma is paying for the discovery function of unique drug assets of the big token big labs will pay for with these startups actually um so that's kind of a lot of the way i was thinking about that thesis in particular when you say unique tokens what do you mean by that um i mean like things like robotics data things like um collecting video game data to help train robotics.

25:28So those are a couple of the examples in terms of really unique data sets that Foundation Model Labs are essentially paying for to get in front of. And have you seen that trend play out this year? I'm trying to think of the deals that we've seen in the past couple months. I mean, gosh, OpenAI has been making a lot of acquisitions, I guess, on the applications. I mean, they've been trying to bolster their Codex business, right? Yeah, they bought a developer tools company called Astral, which I would sort of place maybe slightly adjacent to this thesis. It feels like there's probably a couple other companies that they've been kind of circling that maybe we'll see some deals announced soon.

26:13But it feels as if it's early days still, but I think we'll be hearing more about this very soon. So how has this thesis then informed the companies that you are trying to invest in? I would say that it's something that we've been thinking about in terms of like, hey, are these companies that eventually would be sold into the labs if that's one path? But I think the other tricky thing is like you don't want to over index on it, right? You want to back companies that are going to be around for a long time. and maybe there is some strategic value in one case to the labs, but that's not the end goal in many cases.

26:55So it's something that I would say that we're considering, but not maybe the crux of the investment thesis at times. Tell me about what you are investing in, because your focus seems to be split across a number of different sectors. Yeah, I would say right now, I've been looking a little bit more at things in the kind of like AI infrachain. So data centers, energy, chips, materials, things like that, that kind of inform the sort of AI build out. So I've been looking a little bit more in that space. Now, before that, before venture, you built a furniture company. Yeah, so a bit of a whining path.

Read the full transcript

27:36Which is a big, I mean, furniture to AI infra is, I don't know, I guess building blocks, you could say that much. Sure. Yeah, of course. Yeah, I mean, when I was in college with a couple of folks who I'd been working with at different internships, honestly, just noticed the market opportunity in the mid-market between, you know, Herman Miller on the high end and IKEA on the low end. and we decided to sort of form Branch Furniture back in 2018 and the business has scaled a ton since then. I ended up joining Haystack in 2019 right out of school and I've kind of seen the Branch story from afar. But yeah, it's been a kind of wild ride.

28:16And how has building a furniture company informed the AI infrastructure companies that you are looking to invest in? I would say that, you know, that through line is a bit trickier. And candidly, Akash. Well, thank you. Thank you for being candid because a lot of people come on the show and say, you know what? I mean, the best founders, they really stand out. Totally. I mean, I think the thing that I learned in that experience was more around working with founders. So, I mean, the founders of Branch, Greg, Siv, Verdi are like amazing. And I think that, you know, building a sort of like lower margin business takes a lot of work and a lot of guts and a lot of courage.

28:56And I think I learned a lot more about working with founders. And I think that's the generalist kind of mindset I've taken overall at Haystack. Great. Well, Asha, I want to thank you for coming on. That is Asha Songvi from Haystack here on TI-TV. As AI moves from the demo to real-world deployment, one of the biggest hurdles isn't capability. It is compliance. In industries like healthcare and finance, getting AI into production means meeting strict standards and high levels of trust. So what does it take to sell AI into some of the most tightly regulated industries? Joining me now to answer that question is Cheryl Cheng, Managing Partner at M12, Microsoft's venture fund.

29:36Cheryl, welcome to the show. It's great to have you here. Good morning. Thank you for having me. So we talk a lot about AI adoption on this show, and we talk a lot about how it's going a little bit slower than I think some people would have imagined maybe two or three years ago. And then you have AI adoption in highly regulated industries, the ones that you have spent a lot of time in, industries like healthcare, financial services, insurance even. My hunch is that AI adoption is even slower in those sectors. But I want to ask you, I mean, is that the case? Well, it is and it isn't. So M12, Microsoft's venture fund, we invest in mainly early stage companies.

30:19So a lot of the businesses that we are working with are still in their early days of commercialization. So what we're seeing is in the regulated industries, especially things like healthcare, financial services, there is a huge need for bringing AI in because they just don't have enough workers to do all the things that need to be done. And there are so many complex workflows that they absolutely need AI to help with that. I think one of the challenges, and Gartner talks about this in a recent report, is that a lot of federally regulated industries don't have their data AI ready. And so even though they really want to push forward, they don't have the data in a format or in a secure fashion to really unleash to AI.

31:04So that is one of the hurdles. On the startup side, you see a different hurdle, which is that AI needs to adapt to the workflows of these regulated industries. And so being very domain specific becomes something that's very important. Okay. So, I mean, maybe let's break down the industries. Healthcare is where you spend a lot of time. You just mentioned there's a big need for AI there. I mean, financial services, insurance, and government, what's the story in those three sectors? Well, so it's all around developing the platform and technologies to have trust in handing over this data to AI, right?

31:47So in financial services and insurance, you're actually gathering a lot of very personal data about your customers. And you can't just have that open for agentic systems to do the work on that data. In traditional SaaS, there was software there that was holding all of this data, but humans were the ones doing the work. They were the ones making the decisions on things that were going to happen with that data. When you move into AI, you now have the potential for an agentic system to be the one doing the work. And so you need the controls and the context windows and the regulations to be able to control that.

32:22In the federal government, the standard is even higher, right? And so most people don't know, but there is a program called FedRAMP, which is if you want to sell software to the federal government, you must be FedRAMP. And that means hitting a level of security that's higher than what you would normally need to hit, even if you were selling to a Fortune 500 company. So actually, one of the investments that I'm actually very excited about is one that one of my colleagues, Alan Dew, did recently in a company called Knox Systems. And they are one of only two AI native companies that can actually enable software businesses to become FedRAMPed.

33:03And unless you get FedRAMPed, you can't sell to the federal government. So they're helping companies get the compliance. That compliance and level of security. And that is very important because the federal government is the largest buyer of software in the United States. more than any other Fortune 500 company. So it's a great customer, but you have to be compliant with their standards in order to be able to sell to them. How do you get compliant? So you go through this program called FedRAMP and there are a sequence of things that you need to be able to check off to show that you are secure, that you're handling data in the correct way, et cetera.

33:45Most companies, and there are only about 500 to 600 companies today that are FedRAMPed and therefore qualified to sell their software to the federal government. What you need to be able to do to go through that process is usually a multi-year process and usually costs you somewhere between$2 to$5 million today. With Knox and their very large AI-managed federal cloud system, they can do this FedRAMP process for you in less than 90 days at about 10 % of the cost. So here you're seeing AI help get companies ready to be able to sell to the government, including those who have AI systems and solutions that they want to sell.

34:30So are you saying that the, so, I mean, it's cost around$5 million, you said, to get the FedRank compliance. This company that you've invested in, it lowers the cost overall to get the compliance conceivably because you spend less time, you know, preparing all the documents and stuff like that. And so is the idea here that startups and companies, I mean,$5 million, it's a lot of money, but it's also, you know, for some of these bigger companies, I imagine it's sort of just the cost of doing business. Are you suggesting that companies can't afford that$5 million to get access to the government as a customer?

35:08No, I don't think it's, it's not necessarily about affording it, right? If you're a large company like Microsoft, like Amazon, like Oracle,$5 million is not the issue. If you're a startup,$5 million is a lot of money to be able to unlock federal customers. So depending on the scale of your business, the$2 to$5 million could be a lot, but it's also the time. So the FedRAMP process is very complex and usually takes two to three years. Some of the customers that we've talked to, you know, spend years getting themselves FedRAMPed. But the federal government and state governments are very, very good customers.

35:49What we're seeing right now is the standard for FedRAMP was always so high that it actually would demotivate companies from going after it. But now as we move into AI systems and you move into these multi-cloud solutions, the demand even for enterprises for companies to show that they're FedRAMP because it shows a quality of security and trust becomes very important. And so when we go back to the original topic that we were talking about, how do you go from proof of concepts and pilots into production? A big part of that is being able to demonstrate trust in handling data going through these AI systems.

36:32Go ahead. Sorry. You know, I just wanted to, before we leave the FedRAMP topic, I just want to talk about the compliance process here as a whole. Because, look, here at The Information, we've reported on even inside big companies like Meta, agents going rogue, for example. And there being a lot of security threats. I mean, Meta is as big a company as you could get, right? And even their agents are going rogue. And so on the flip side here, I hear you saying, hey, there's only about 500 companies that have compliance to sell software into the government. And that was probably before agents came along.

37:05Now agents are coming along. I mean, my question is, is the government actually going to adopt any of these agents if there is a risk that these, you know, bots are going to go rogue? I mean, is it going to happen? Okay. So I think you're talking about two very different things. So Knox and the FedRAM system is really about securing the data. And this is more from external threats. What you're talking about right now around agents going rogue is actually a little bit of a different issue. That is more around have you set up your agents within the enterprise, any enterprise, to have the right permissions, controls, observability, all of that.

37:49So the agents that might go rogue, as you say, they're not necessarily doing anything illegal. They just weren't given the permissions and the guardrails that they needed to have. That is an independent issue from are you FedRAMP or are you HIPAA compliant, et cetera. This is we are still in the very early days of understanding, programming, commissioning, and checking how our agentic systems are working. So those are very different things. You could have a company that is FedRAMP, which means their cloud services are completely secure. Right. But if you don't write the right agent, that agent could still go rogue, as you put it, within the permissions that you had given it.

38:38So those are – Which of these seem like the bigger threat to you right now, to the adoption of AI in the long run here? In the long run, I think it's probably – Well, short run. I mean, we're trying to get – maybe short run. We're just trying to get to zero to one, right? Yeah, we're trying to get there right now. So maybe short run is the better way to phrase it. Yeah. So I think most enterprises do not need to be FedRAMP. So I think that level of data security, you know, Knox Systems is named after Fort Knox. Like you don't need that level of security. So that's probably not as much the short-term problem.

39:15I think it's then more of the, how do we deploy agentic systems into an enterprise workflow that follows the workflow in permissions that that enterprise has deemed as appropriate. And so that, I think, is a more short-term block than this issue of FedRAMP. FedRAMP is pretty specific to selling to regulated industries, particularly the federal government. And I think looking at things like financial services, if you are FedRAMP, you are more than enough for financial services. So it's kind of, it would be more and then therefore make it easier to sell into those industries. And maybe my question was a little more targeted to those regulated industries like government, like you're saying, because the Fed ramp is one barrier and you have a startup here that is helping with that.

40:06But what about the other risk of the security threats and stuff like that? I mean, that's a risk not just to regulated industries, but to any industry. And like you said, it's a different risk. It's coming from within. It's coming maybe from teams not having built things as prudently as they could have. But that seems like an even higher risk for regulated industries and one that I don't – don't they have to have some kind of a certification process to prevent that from happening? I don't know that you need certification because each one of these agentic systems, think of them as like living organisms, right?

40:50And they're evolving and changing and learning as they're doing the work that you ask them to do. I think the bigger thing is traceability. So can we actually go back and look at the decisions that an agent has made and then course correct along the way? And so these are the types of fail-safes, if you will, in protections that we're seeing companies now start to build in to become enterprise ready. Before, when you were just focused on consumer, these were still issues, but they're not as prominent as a procurement agent at an enterprise making the decision whether they're going to bring this in and open it up to all of their data.

41:36So there's the intelligence layer, which we already see advancing very quickly with all of these new models. But then the piece that needs to catch up is all of the controls that need to be there and in place and tested in order for enterprise to be able to adopt. So it's really more of that observability, control, traceability of decisions that is sort of the next layer of AI development. And so the last question for you is then, I mean, that seems to be not an insignificant amount of work that needs to get done in order for not just AI adoption, but AI adoption and regulated industries in any industry is to happen.

42:17But so when is that going to happen, is my question. The observability that you're talking about, you know, it seems like, and I'm asking here, you know, you're in the industry, you're seeing the startups. It feels like that is a gap right now in the tech stack that exists for all these AI tools. Is that true in your opinion? It is true in that it is less developed than the intelligence layer, which is where we've seen all the high-flying startups. At M12, we're really excited to work with these companies because our strength as Microsoft is the enterprise. So our ability to help these companies as they're building applications, agentic applications for the enterprise, is helping them guide through the things that enterprise customers are going to want.

43:09So these are absolutely developable, solvable problems. Helping the young startups focus on these issues in addition to getting the distribution to the right enterprise customers is where we really hope to add value. And I think as more investors look to move their companies from proof of concept, pilot, into enterprise, it is both everything that we talked about today as well as a good channel of distribution to the right enterprise customers that will make AI stick in regulated industries and in enterprise. Great. Well, Cheryl, I want to thank you for coming on. That is Cheryl Cheng, a managing partner at M12 here on TIT.

43:54Thanks for having me. Well, that does it for today's show. A reminder, we are on this stream Monday through Friday at 10 a.m. Pacific, 1 p.m. Eastern. I want to thank you all for tuning in. We really do appreciate your viewership. Make sure to subscribe to the information on YouTube and follow us on X, Instagram, TikTok, and check us out wherever you get your podcasts. I'm already excited for our next show tomorrow. Have a great rest of your Monday. Bye-bye for now.

From the publisher

The Information’s Yueqi Yang talks with TITV Host Akash Pasricha about crypto’s "crisis of faith" and why Bitcoin’s value has halved since October. We also talk with Elon Musk Reporter Theo Wayt about how the SpaceX IPO hype is boosting—and threatening—the space ecosystem, and Haystack Partner Aashay Sanghvi about the "Big Token" thesis for AI startup exits and making it onto The Information’s Next General Partners list. Lastly, we get into the complex world of FedRAMP and regulated AI with M12 Partner Cheryl Cheng.


Articles discussed on this episode: 

https://www.theinformation.com/articles/cryptos-nasty-downturn-getting-worse

https://www.theinformation.com/articles/spacex-hype-boosts-stocks-crosshairs


Subscribe: 

Sign up for the AI Agenda newsletter: https://www.theinformation.com/features/ai-agenda


TITV airs weekdays on YouTube, X and LinkedIn at 10AM PT / 1PM ET. Or check us out wherever you get your podcasts.


Follow us:

X: https://x.com/theinformation

IG: https://www.instagram.com/theinformation/

TikTok: https://www.tiktok.com/@titv.theinformation

LinkedIn: https://www.linkedin.com/company/theinformation/

More from The Information's TITV

All 304 episodes
Crypto’s Nasty Downturn Worsens, SpaceX IPO Hype Halo Effect, Selling AI in Regulated IndustriesThe Information's TITV · 44 min
Listen in VO