Can You Really Buy A House Without Selling Your Crypto? | Vishal Garg
- The episode’s macro framing: the US mortgage system is a $15T government-guaranteed asset funded by government-guaranteed deposits, and the ~400bps spread between the 6.5% asset and 2.5% funding is mostly consumed by process. Garg says 300 of those 400 basis points are costs; Jason then frames that as US households losing $450B a year to the “PDF-manufacturing process,” roughly $4,500 annually across each of 100 million US households that own a home. Jason calls the industry’s paper-heavy structure “paperocracy.” Better is attacking it with AI on the front end and tokenization on the back.
- Better’s structural edge is unit cost: it can make a mortgage for under $2,000 against an industry cost of about $12,000, of which roughly $9,500 is labor. The company is at an ~$8B annualized origination run rate, up almost 2x for two straight years, and the “mortgage in a box” platform business—Coinbase, Credit Karma, Finance of America, LendingClub—is already over 50% of revenue with no real B2B sales motion: “we swarm, we do, and then we demonstrate.”
- The Coinbase product produced a Fannie Mae-eligible Bitcoin-backed mortgage: pledge BTC or USDC via smart contract on Coinbase Custody in lieu of a down payment, with no margin calls. Garg’s logic: “we don’t margin call you on your house when it goes up and down in price”—pay principal and interest on time and, in his claim, Bitcoin can go to zero. A Michigan couple got a 6.5% 30-year fixed, 100%-financed mortgage; “if the Bitcoin appreciates more than 8% a year... the Bitcoin paid for the house.”
- Tokenizing the mortgage beats tokenizing the house—the agency mortgage market trades $180B a day, and stripping out intermediation layers is where the value is. The planned $500M Sky credit line, backed by tokenized mortgages and not yet live, is estimated by Garg to cut Better’s cost of capital by 25bps on short-duration capital and ~100bps through long-duration tokenization, potentially allowing mortgage rates to fall 50–100bps. Sky’s competitor in this frame: “BlackRock.” MBB illustrates the leakage: it owns roughly $40B of agency mortgages with underlying rates around 6% or more, while its net yield is 4.1%.
- AI’s role isn’t primarily underwriting—the rules are preset by Fannie Mae, Freddie Mac, and FHA/VA—it is “the uniform application of the rules” without $210,000-a-year underwriters who cannot memorize 45 investors’ 800-page guidelines. Better’s top loan officer now has an AI digital twin with “all of his colloquialisms,” while the ChatGPT partnership’s Tinman app lets brokers quote and pre-approve customers live at open houses. Garg expects “way fewer employees” in five years outside data labeling, which he calls a core moat because every loan produces both a financial asset and a context graph.
- The tradeable claim on the equity: the street prices Better as “a call option on rates coming down” at ~2–2.5x run-rate sales, missing the pivot from D2C refi brand to AI mortgage platform. Garg says he has been buying the stock and has publicly committed to adjusted-EVA breakeven by September, saying he is “pretty confident.” His dependency ranking is categorical: “I think we could do it without blockchain... but we can’t do it without AI.”
- His message to depressed crypto natives is the 2003-internet analogy: the zealots ran from 1995–2000, the apocalypse followed from 2000–2003, and now “the pipes are laid” and utility can be unlocked. The filter for what becomes valuable: “if it drives consumer utility and it has distribution, it will be very valuable.” Garg sees tokenization as “the realization of P2P finance” conceived more than 15 years ago.
1. A $15 trillion market losing $450B a year to “the PDF manufacturing process”
- Garg’s top-down math sets the table: $15T of US mortgages, 95% guaranteed by Fannie Mae, Freddie Mac, FHA, and VA, with the rest in a banking system “effectively government guaranteed post the financial crisis.” The asset yields ~6.5%, the deposits funding it pay ~2.5%, and 300 of those 400 basis points—$450B a year—are consumed by “people, processes, making 550-page PDFs.” Jason calls this paper-heavy structure “paperocracy.”
- Jason asks whether banks exited mortgages because they got caught on the wrong side of rates; Garg’s flat correction is that rates had nothing to do with it. Post-GFC compliance pushed bank cost per mortgage to ~$15,000, so Wells Fargo—the largest originator in 2014—Chase, and Bank of America ceded ground to Rocket, Better, loanDepot, and UWM. Even the big nonbanks run $12,000–$13,000 because of linear processing: loan officer to processor to underwriter to closer to funder to quality control, with “each one of these people rechecking the work that the last person did.”
- Why mortgage is “especially ripe for disruption by AI”: unlike black-box consumer credit, the process is driven by rules preset by the GSEs, FHA, and VA. AI therefore works primarily in “the reasoning, communication and orchestration layer,” which Garg says is referenceable. There are 550,000 licensed loan officers, and “the bulk of what they do every day is stare and compare.”
- Garg’s broader fintech indictment: credit card rates are 36%, “the same as they were when Citibank rolled out credit cards in the 1980s.” Financial services absorbed the internet’s savings for themselves—“they’ve created friction that’s maintained the margins”—and tokenization could break that pattern.
2. Unit economics: $2,000 loans, and a platform business that’s already half of revenue
- Better has made $110B of mortgages over roughly ten years and runs at an ~$8B annualized origination rate between itself and partners, up almost 2x for two consecutive years. Its cost to make a mortgage is under $2,000 versus the industry’s roughly $12,000. D2C economics: about $6,000 all-in, including roughly $4,000 of customer acquisition, then sold to investors for about $8,500; Garg also gives a $400,000-loan example sold for about $8,000, producing roughly $2,000 of margin.
- The platform business—“mortgage in a box” for banks, fintechs, and brokers that bring their own customers—earns $2,000–$4,000 depending on the services used and is already over 50% of revenue. There is no major ad spend or B2B sales machine. The motion is: “we swarm, we do, and then we demonstrate”—a partner sees its full customer journey within a week. “Speed wins” against incumbents “asking them to come to Detroit.”
- Garg has abandoned the idea that Better.com must own every customer: purchase buyers “want hand-holding,” so AI should give the last mile 10x leverage instead. The live ChatGPT-partnership Tinman app lets brokers quote at open houses in real time, and Garg says 6,000 banks want back into mortgage without rehiring the full loan-officer, processor, and underwriter stack. His framing of the real buyer: the person in Peoria isn’t buying a $300,000 house—“they’re buying a house for $30,000 down and $2,200 a month.”
- On Better.com, customers can already go “all the way through lock” by talking or chatting; lock-to-fund is in progress. Full origination inside ChatGPT is still being developed around security “wormholes” so Social Security numbers and pay stubs do not reach the LLM. Garg notes that his child gave Gemini access to everything, and “there are people out there just pressing the yes button.”
3. The Bitcoin-backed mortgage: pledge it, don’t sell it, and no margin calls
- The idea is 12 years old and personal: when Garg was buying his own place, he would have had to liquidate investments for a 20% down payment rather than pledge them. No US broker could handle a single-stock pledge. Jason tells a similar story: Wells treated him as having no assets because they were held in crypto, even after the Gemini team joined calls with Wells. A 2020–21 attempt with CartaX died; Coinbase Custody’s smart-contract “tri-party pledge” revived the concept, which Better then took to Fannie Mae.
- Mechanics: pledge Bitcoin or USDC as the down-payment equivalent, get a Better pre-approval, and “show up at the closing table, and the only person writing a check is Better, not you.” Garg contrasts it with other crypto mortgages that have margin-call features, require pledging roughly 100% of the home value, and charge 9%–11% rates. This product is described as a 30-year fixed. A Michigan couple got 6.5% financing on a 100%-financed home; “if the Bitcoin appreciates more than 8% a year... the Bitcoin paid for the house.”
- Jason presses on the no-margin-call claim; Garg’s answer is the product’s essence: “we don’t margin call you on your house when it goes up and down in price.” Garg says Bitcoin can go to zero—as long as the borrower pays principal and interest on time, “you’re good.”
- The roadmap is broader tokenized collateral: $5T in household checking and savings, $35T in stocks, bonds, and digital assets, and the down payment—not the monthly payment—as the largest impediment to ownership. Garg discusses conversations with large ETFs and imagines an Amazon, SpaceX, or other public-company employee pledging stock, keeping the upside, and benefiting from a mortgage payment that stays fixed for 30 years.
4. Tokenize the mortgage, not the house—and Sky’s $500M line versus BlackRock
- Jason recalls the 2018–19 tokenized-real-estate wave, including Harbor, and its lesson that tokenization does not make an illiquid asset liquid. Garg agrees that fractional home ownership—“why can’t I own 62% of my house and pay rent to the other 38%?”—is conceptually interesting but “a long way away.” The more immediate opportunity is a standardized Fannie Mae or Freddie Mac mortgage, a market Garg says trades $180B daily.
- The Sky facility is a $500M credit line backed by tokenized mortgages, but it is not yet live. Garg estimates that short-duration capital from Sky could lower Better’s cost of capital by 25bps, while long-duration tokenization could lower it by 100bps. His goal is to use that difference to reduce mortgage rates by 50–100bps and potentially improve affordability by “$1,000 a month.” Asked who Sky replaces, he answers: “BlackRock.”
- His evidence for intermediation leakage: MBB owns roughly $40B of agency FHA/VA mortgages. Garg contrasts the roughly 6%–6.5% underlying mortgage rates with the ETF’s 4.1% net yield. Multiple layers of financial intermediation take 10–20bps at a time; going consumer-to-capital could let individuals earn the kind of mortgage return banks receive on their deposits and effectively have “their own little bank.”
- Garg also argues that stablecoin issuers’ short-duration Treasury yields become less attractive as rates fall, while a Fannie Mae or Freddie Mac mortgage is “possibly the single best long-duration asset” because it pays a fixed rate for 30 years. Institutions care about prepayment risk, he says, but individuals may be comfortable receiving their money back early.
- Jason observes the inversion: crypto natives are pessimistic while fintech and institutional participants are increasingly excited. Garg’s analogy is the internet: zealotry from 1995–2000, an apocalypse from 2000–2003, then “the pipes are laid” and utility emerges. His test for the depressed holder is, “if it drives consumer utility and it has distribution, it will be very valuable.” Lending protocols that price collateral instantly without traditional foreclosure machinery qualify, and tokenization is “the realization of P2P finance” conceived more than 15 years ago.
5. AI as “the uniform application of the rules”—digital twins, forced adoption, and the context-graph moat
- What AI actually solves in mortgage is “the uniform application of the rules,” minus the biases and knowledge gaps of expert labor. The average underwriter earns $210,000 and cannot hold 45 investors’ 800-page guidelines in their head—“how many humans remember the Encyclopedia Britannica front to back?” AI can also advise: pay off certain loans, raise the borrower’s credit score, and lower the mortgage rate. Then it can close: “I can just do it for you. Just tell me yes.”
- Better’s best operators are leaning in. Ryan Grant, its top loan officer in Orange County, now has an AI digital twin with “all of his colloquialisms,” implicit lessons, the underwriting guidelines across all 45 investors, and instant math. The effect is both top-line—24/7 coverage and more customer surface area—and cost-related: great employees get 10x leverage.
- The management doctrine, relayed by a prominent Silicon Valley VC three years ago, is that the difference between a company and its peers “is going to be determined by how hard you can force AI down every level in your organization.” Garg says employees resist because AI disrupts their work, workforce, and “personal sense of meaning.”
- Headcount is about 150 in corporate and over 1,000 including brokers, processors, and India-based data entry. In five years, Garg expects “way fewer employees” outside data labeling, which he calls a core moat and says he “probably should have” spun off earlier. Every loan creates two assets: “one is a financial asset and the second is the context graph.” His ambition is for every US house, appraisal, and consumer to pass through the company’s data-labeling process, enabling instant financing for anything. The platform could ultimately become larger than the Better.com direct brand—the “Stripe for loans.”
6. The stock as a mispriced AI platform, breakeven in months, and public-company scar tissue
- Priority one is profitability after “four really tough years”: Garg has publicly said Better will reach adjusted-EVA breakeven by September, which Jason characterizes as two or three months away, and Garg says he is “pretty confident.” The dependency ranking is stark: “I think we could do it without blockchain... but we can’t do it without AI. All in.”
- Garg says the street sees an online refinance business and prices Better “almost as a call option on rates coming down,” at roughly 2–2.5x current run-rate sales, missing the transformation into an AI platform modernizing the mortgage and home-equity business. “If this was a private company it would trade at a very different valuation,” he says, adding that he has been buying the stock.
- Lessons for founders considering 2027 listings: public companies move more slowly; public directors are typically “loss minimizing” where VC boards maximize outcomes; and some superstars may take their public-stock liquidity and leave. Garg’s stated regret is: “if I had an A as a number two, I would have wanted an A+”—build depth two layers down before listing.
Full transcript
Nothing said on Empire is a recommendation to buy or sell any investments or products. This podcast is for informational purposes only, and the views expressed by anyone on the show are solely their opinions, not financial advice or necessarily the views of Blockworks. Our hosts, guests, and the Blockworks team may hold positions in the companies, funds, or projects discussed.
All right, everyone. I’m very excited about this. I’ve gotten to know Vishal, who’s the CEO and founder of Better. I’m really excited about this.
The dream for this episode is to start with the macro—to really talk about the mortgage industry as a whole—then get into Better. I think you guys are the company that I’ve talked to among all the mortgage providers that is the most innovative on 2 fronts: AI and crypto or blockchain, whatever you call it. Maybe we can frame the problem, or just the big societal view of the mortgage industry in general, and then let that guide us.
Okay, cool. I think if you start right from the top, there’s about $15 trillion of mortgages out there in the United States. They’re all funded, and 95% of them are government-guaranteed: Fannie Mae, Freddie Mac, FHA, and VA. The remaining 5% are in the banking system, which is effectively government-guaranteed post the financial crisis. So you have a government-guaranteed asset that’s permeated throughout the financial system.
It yields about 6.5%, and it’s funded by deposits that are also government-guaranteed at 2.5%. So you have about 400 basis points per year on $15 trillion—$600 billion a year of intermediation expense—government-guaranteed deposits funding government-guaranteed assets. Three hundred basis points of that 400 basis points is lost to costs in the banking system, and those costs are people and processes making 550-page PDFs.
So in America today, with blockchain, tokenization, and AI, we are living every day in a world where US households lose $450 billion a year to the PDF-manufacturing process.
And that’s the mortgage industry as we know it.
That is true. Yeah.
Those are the margins. That’s the margin of the mortgage.
It’s the paperocracy, right? Four hundred and fifty billion dollars across 100 million US households that own a home—that’s $4,500 per person per year.
I mean, it’s massive. And so the goal is: how can we use the latest technology to make the process of creating these mortgage documents cheaper, faster, and easier?
How can we use it to make it cleaner? How can we enable trust, when post the financial crisis there has been nearly none? In doing that, how can we make homeownership more affordable and accessible for all Americans?
1. The Better Business Model
What is—how should someone think about the Better business? I’ve been following you guys for a while and, in preparation for this podcast, I spoke to a couple of your investors. I’d be curious how you think about what it is today and what the evolution has looked like.
Okay, so Better’s vision is to make home finance cheaper, faster, and easier; to have the highest approval rate and the lowest interest rate across the full range of products that any American consumer can qualify for to buy a home, refinance a home, or get home equity out of their home.
We’re a long way into our journey here. We’ve been around for about 10 years. We’ve been making mortgages, and we’ve made $110 billion of them. In that process, we first started with machine learning to automate the process. We built a rules engine that matched investors—45 different investors in the mortgage market—with consumers directly.
Then, over the past 5 years, we used the learning data from creating these mortgages to put an AI loan officer, an AI loan processor, and an AI loan underwriter in place. That now automates the process of making and underwriting a mortgage in minutes, instead of something that used to take 21 days or more.
We’re almost reincarnated as the leading AI mortgage platform, built on the learning data, process mapping, and rich context graph that we created in our first iteration as a digital mortgage company.
And we think the AI mortgage business is going to be really meaningful for American consumers because it will help everyone afford to be part of the American dream of homeownership.
What’s the scale and scope of the business, however you look at it—loan volume, customers served, market position, originations?
Between ourselves and our partners, we’re at about an $8 billion annualized origination run rate. A decent-sized fintech. We’re hoping to continue to double that for the foreseeable future. That’s up almost 2 times from where we were the year before, and 2 times from there the year before that. So we’ve now really been making progress.
I think the other big thing that’s really important to remember about Better is our marginal cost to make a mortgage. It costs the mortgage industry $12,000 to make a mortgage, and almost $9,500 of that is labor cost. We’re able to do that in under $2,000 today.
So we’ve got a competitive cost advantage that has gotten better and better with AI. Now we’re not originating only for ourselves; we’ve opened up our platform to fintechs, mortgage brokers, and incumbent mortgage companies. That is dramatically fueling our growth because we’re no longer relying only on Better.com as a distribution channel.
You started the business in 2014, or—
2014.
2014. Okay. So 12 years in, how has the mortgage space changed? I’m assuming it used to be that you’d go to Chase, Bank of America, or Wells Fargo. Now I see ads at every sporting event for Rocket Mortgage and SoFi. I don’t know if you’re doing the TV ads as well, but maybe just tell us how the space has progressed.
Post the financial crisis, coming out of it in 2014, Wells Fargo was the largest originator of mortgages in the country. Chase was big, and Bank of America was big. Over the past 10 years, more and more of those banks got out of the mortgage business because of the ups and downs in the mortgage cycle, and also because the cost to manufacture mortgages for a bank was about $15,000 because of the compliance and paperocracy that got created after the global financial crisis.
And did so many of them get caught on the wrong side of low interest rates, or did that not have anything to do with it?
No, it didn’t have anything to do with it. It was just their cost of making a mortgage.
Interesting.
You know, if you were trying to get a mortgage even today, you’d call Bank of America or Chase, and it would take them about 2 months to get you that mortgage.
I know. I did it with Wells. They actually treated me like I had no assets because it was all crypto. This was in 2020 or 2021—I forget.
Yeah. I know. The bank origination process for a mortgage is like the internet didn’t exist.
Yeah.
Right. And so those people have lost ground to companies like Rocket, Better, loanDepot, and UWM—the nonbank mortgage lenders that have been able to use technology to speed up the process.
That being said, even at the very large nonbank lenders, it’s costing them $12,000 to $13,000 to make a mortgage because they still run a linear processing platform. You have a loan officer who passes the loan on to a processor, who passes the loan on to an underwriter, who passes the loan on to a closer, who then passes the loan on to a funder, who then passes the loan on to a quality-control person, and so on and so forth.
Each one of these people is rechecking the work that the last person did, then either finding issues or coming back. When you go through the process as a consumer, you’re like, “Wait, I just dealt with this other person. Now I’m being passed on to this person, and this person is asking me all the same questions over again. They’re asking me for this other minor nit in a document. I qualify either way. What is your problem?”
And that’s really bad. This is why the industry is especially ripe for disruption by AI. Unlike even other forms of consumer credit, where you have a black box and instant underwriting, here everything is driven by rules that are preset by the government-sponsored enterprises, Fannie Mae and Freddie Mac, or by FHA and VA.
What you’re doing with AI is not actually in the underwriting layer. It’s in the reasoning, communication, and orchestration layer, and that’s entirely referenceable. You can train the AI to effectively do all of it, which I think is going to be really interesting for the industry to deal with.
You have 550,000 licensed loan officers in this country. You have hundreds of thousands of processors and underwriters, and they will tell you the bulk of what they do every day is stare and compare.
The amount of consumer surplus that can be unearthed and the amount of value that can be given to a consumer when you stop selling and just collecting data and looking at it, and instead start to help people analyze their full finances and be the debt adviser for people—to do all that—there’s real value you can create for the American consumer.
Yeah. One or 2 more questions about your business model. What is it? How do you make money?
We make money in 2 ways. The first way is our direct-to-consumer business, which is our traditional business.
We make a mortgage. We acquire a consumer, process the loan, underwrite the loan, fund the loan, and all of that costs us about $6,000 to do. Then we sell that loan to an investor for $8,500.
Instead of the $15,000 that it cost us.
Exactly. Our customer acquisition costs are still quite high. They're about $4,000 of the $6,000, so it costs us $6,000 to do, and then we sell that loan to an investor for $8,500.
Got it.
If you've got a $400,000 mortgage, it'll cost us $6,000 to make it, and then we'll sell that loan to an investor for about a 2% premium. We'll sell it for $8,000 and make $2,000. That's our direct-to-consumer business. The same applies to home equity lines of credit, where the math is about the same.
The second thing we've got is a platform business, where we sell “mortgage in a box” to banks, fintechs, and mortgage brokers. They bring the customer, so we don't have that customer acquisition cost upfront. They bring the customer, and then they pay us for processing, underwriting, closing the mortgage, and getting access to our investor network. In those cases, we typically make anywhere between $2,000 and $4,000, depending on how many of our services they're using. We also try to make about $2,000 in contribution margin on the loan.
Is this what you did with Coinbase or Finance of America?
That's right. We've got Credit Karma, Coinbase, Finance of America, LendingClub, and a whole bunch of large fintechs and incumbent mortgage companies using our platform to originate mortgages.
Do you have a thought on, if you fast-forward a couple of years into the future, how important it is to own the customer?
Actually, I don't think it's important to own the customer. I think it's important to change the industry. I think it's important to make the world we want to live in.
There was a time when we thought Better.com could actually reach all consumers everywhere because all consumers were migrating to the internet. What we learned over the past couple of years, particularly with respect to purchase mortgages, is that people want hand-holding. What AI enables the hand-holder—the last mile, the person on the ground walking you from open house to open house—to do is become way more efficient. Now we want to empower the last mile to provide exceptional service to consumers.
Can you speak to the ChatGPT thing? Is this live? I saw a couple of headlines.
Yeah, it's live. We have mortgage brokers around the country who are showing up at open houses. In the old days, they'd take your information and then, after the open house, go to their office and say, “Hey, you can afford to buy this house. Here's your pre-approval.”
Now they're doing it right there, just talking to the Tinman app and saying, “Hey, my customer's got this and this and this attribute. What's the best rate we can quote them? How much is this house going to cost per month?”
Interesting.
People in America don't buy houses. On the coasts and in some of the fancy places where we are, like this island in Manhattan, people buy houses for X dollars and Y dollars. But the regular person in Peoria, Illinois, isn't buying a house for $300,000. They're buying a house for $30,000 down and $2,200 a month.
That's what they need to know. They need to translate the price they see on Zillow into: How much money do I need to come up with to stop being a renter? How much down? And then what's my monthly nut?
So the ChatGPT product—the partnership—is for brokers?
It's for brokers. It's for fintechs. We designed it so that any bank could get back into the mortgage business. With some of the regulatory-capital changes taking place, where they're rolling back some of the things that happened after the global financial crisis, mortgages are a really attractive asset again.
We've got 6,000 banks in this country that want to get back into the mortgage business. But do they want to hire loan officers, processors, and underwriters? No. Do they want to put a person in their branch? No. What they want to do is enable anybody in the bank to say, “Hi, sir. Yes, can I help you? Can you tell me a little bit about these things?” Tell it those things, have that typed into the ChatGPT Tinman app, and have it come back with an approval nearly instantly.
And now that's a delightful thing. It's completely compliant and foolproof.
2. The Opportunity For Tokenization & AI
This is the amazing thing with AI. You go from all of the software that's been created in the last 50 years, which is all specialist-labor-driven software where you, as a person, have to understand, learn, and get trained on the software. Now you go to no interface: You can just talk to it, talk about your problem, it'll ask the relevant questions, and it'll give you the answer. That's magical. It's magical for ordinary people.
I agree, I agree. What do you think the timeline is to just get a mortgage through ChatGPT or through Claude?
Oh, you can do it right now.
You can do it right now.
Right now on Better.com you can go all the way through lock on by just talking to it or by chatting with it, and we're working on lock to fund right now. It's still in a task-frame-based framework, but we're working on getting that so the entire thing can be done entirely.
Could you—but you're still on the Better website, right? Can you do it inside ChatGPT?
We're working with ChatGPT to get that live. We're working with them. There's a bunch of security stuff, like social security numbers, pay stubs, things like that. So we have to create ways to collect really rich personal financial information from consumers without it being given to the LLM models. And so we're trying to create these sort of wormholes where you can upload stuff and give it back.
What was really interesting is my kid—he's given Gemini permission to everything: his Gmail, his Google Drive, his YouTube history. And so there are people out there just pressing the yes button. If you have, say, your pay stubs and your tax return in your Google Drive, we should be able to get it for you.
And I mean, they rolled out finance, right? It took me two days, and I was like, "Ah, should I connect? Should I click, click, click?"
Oh, you did the whole thing?
I did it with three of my accounts to see how good it was, because I also use this tool called Monarch that tracks everything. Then I was like, I wonder if ChatGPT is better.
I think that's really clever. I mean, what they've done with Plaid—eventually, most of what gets disrupted by AI is money. The fact that now each and every one of us can have the same power as the most excellent private wealth-management guy at Goldman Sachs.
Yeah.
And also that guy works 24/7, right?
Right? And it can just do it for you every day. On Betterment and Wealthfront, they optimize your portfolio daily. It could say, “Hey, I see you have $7,000 in your bank account, but you're paying interest on $1,700 of credit card debt. Why don't we pay that down, and then you can draw it later when you want to get something else?”
People don't even think about that. People pay credit cards only when the payment comes due. But you can pay your credit card at any time of the day, any time of the month, and not pay interest when you have money sitting in your checking account. If you plan to pay at the end of the month, you might as well pay early.
I think most financial services today have taken the benefits of the internet and absorbed them for themselves in the form of cheaper distribution, easier access, and lower credit risk. But they've created friction that has maintained the margins.
Yeah.
Credit card interest rates today, at 36%, are the same as they were when Citibank rolled out credit cards in the 1980s. You've had all this technology and all these savings, but the rates are the same, right? Shouldn't the rates have come down?
The same thing applies to mortgages, auto loans, student loans—all of these things. The rates are the same. Costs have come down. Financial services companies are more profitable than ever, but the moat is really just coming down. Tokenization changes that even further.
How have you had to adjust how the company runs now that you have these 2 business lines—the direct business and then the platform and partnership business?
I think we've always run the company with a no-customer-left-behind ethos. When I first started, any customer could email me at vgb@better.com. They'd get a personalized email from me talking about their circumstances, and I'd respond. I have a team that responds on my behalf when I'm sleeping.
So that ethos of treating the customer really, really, really well expands to our partners. So now we treat our partners' customers really, really, really well. There's nothing that keeps and grows your partnerships more than treating your partners' customers really, really well and making them happy. That is what has been fueling the growth of our partnership business and our platform business.
We're not buying ads for the platform business. We're not out there—other than at conferences doing demos. So we're kind of like this super startup inside where we do demos, and our existing customers tell us about other customers. We don't really have a massive B2B sales motion, but the business is already now over 50% of revenue.
So when you're doing these deals—Coinbase, OpenAI—who's selling? Who's doing it? Is that you directly doing that deal?
I try to do it directly, and then I have a team of 3 people with me. And we just swarm.
Yeah.
We swarm. We do. We demonstrate. I think that's the difference: We swarm, we do, and then we demonstrate. We're up and running with the partner, showing them what their whole customer journey is going to look like within a week of meeting them.
Yeah.
And speed wins, right? I'm competing against other people who are, you know, literally asking them to come to Detroit.
3. peaq Ad
Yeah. Exactly.
Robots and machines will outnumber humans onchain, and PEAK is how they get there. PEAK OS enables any machine to do business on any chain. It turns machines into autonomous actors and liquid assets with ease, giving robots access to capital and the ability to compound their value. That means funds, allocators, and institutions can underwrite, finance, trade, and route capital to machines the same way they handle any other asset. Millions of new potential consumers for any Web3 service from compute to storage to anything a machine may need without a human in the loop. Head to peak.xyz or click the link in the show notes to find out more.
4. Launching Crypto Backed Mortgages
Let's talk crypto. There are a bunch of angles that we could go here, right? There's the consumer side—borrowing against crypto, which is what you did with Coinbase. There's also DeFi borrowing against crypto, which is what you're doing with Maker—or I guess Sky.
Maybe we could start with the former. Can you walk us through—you guys did this Fannie Mae-eligible Bitcoin-backed mortgage funded with your Coinbase account. Walk me through how that came to be and why that's actually important.
Yeah. So this is a concept I've wanted to do for about 12 years now. When I was going to buy my place, I had made some good investments during the credit crisis, and we were going to have to put 20% down on this jumbo loan from a major bank here in New York. I was like, I'm going to have to sell this stuff. I'm going to have these capital gains. Why can't I just pledge this? It's liquid, just like cash, instead of pledging this down payment. And they were like, "No, not having it." Not at all.
The broker wouldn't even show me places until I showed him cash in my bank account equivalent to 20% of whatever the price point I wanted to buy.
I told you I had this with Wells.
Yeah. I literally had the Gemini team, which is where my Bitcoin was held at the time, get on with the Wells manager—the manager of the manager—trying to show them that Bitcoin is a liquid market, and it didn't work.
No. No. They wanted you to liquidate it. And it's like, why? Because they're spending, and you realize, well, if stock prices go up and down, therefore your buying power changes, the real estate broker is not getting the same commission. So they want you to be fully certain.
So anyway, I said that's crazy. We first tried it again in 2020–2021. We did a deal with my friend Henry Ward at Carta, where he was doing something called CartaX, and it was a private exchange where companies could list their private shares.
I remember this.
When you think about it, as a CEO of a company that was at that time about to go public, most of my employees immediately wanted to sell shares so they could buy a house. And I said, well, what if I don't want them to sell their shares? Why can't they just pledge their shares to get a mortgage? So we tried it there, and that worked for a while. Then Carta decided not to really move forward with CartaX, so it kind of died there.
Then it came back because I started reading about the changes in the administration and started understanding that so many people hold Bitcoin and USDC. Coinbase had worked out, on Coinbase Custody, the first way you could do a tri-party pledge via a smart contract.
When I wanted to do this, even once Better had gone public, my guy at Goldman Sachs was like, "Well, it's going to cost 20 grand. You're going to pledge your stock. We're going to have to move all the stock that we're pledging into a different account." All that because not one single broker in America can actually do a single-stock pledge. They have to do it in an account. They have only account-pledge agreements, and the brokers don't like doing those because they'd rather lend you money on margin. It's very profitable to lend you money on margin, and so they didn't really want to do that.
So I reached out to Coinbase: Why don't we take whatever is tokenized and, in the beginning, start with Bitcoin and USDC? Those are very large tokenized assets among Coinbase users. We can enable them to enter into a smart contract on Coinbase Custody, where they can pledge their Bitcoin or USDC and use it in lieu of a mortgage down payment.
Then we went to Fannie Mae and said, well, look, we can open up home ownership to this very large constituency of consumers who own tokenized assets in Bitcoin and USDC. But eventually, all tokenized assets could be included, if you think about it. There's $5 trillion of household savings in checking and savings accounts. There's $35 trillion of household savings in stocks, bonds, and digital assets.
When you think about what the biggest impediment households have today to buying a home is, they're already paying their rent. They're basically paying their landlord's mortgage, right? The biggest impediment is actually being able to come up with the money for a down payment, which is hard. If you're 32 or 35, you've got your student loans you're paying off. You might have your grad school loans, you might have other stuff, your expenses. So it's really hard to build that savings buffer.
And if you do have that savings buffer, you haven't invested in the market. You don't want to take it out of the market or out of crypto and then sell it all, incur capital gains, to go buy a house. So what if we made that process super easy, which is what we did? Now you can simply go on Coinbase, pledge your crypto, and it basically shows up as if you've made a down payment. Get a pre-approval letter from Better, go find the house you want to buy, come back, transact it, and show up at the closing table. The only person writing a check is Better, not you.
Interesting. How many checks did you write when you were at the closing table with your mortgage?
Too many.
Yeah. What are—there are other crypto mortgage products on the market, right? So what separates this from those?
They have margin-call features. They require you to basically pledge 100% of the home value in crypto.
Yeah.
Right. Here, you're pledging a percentage of your down payment amount. You can't get margin-called, and there are no margin calls. There's nothing there.
The rate is really great because it's not a 9%, 10%, or 11% rate, as it is with other crypto loans of this size. The duration is 30 years, and it's fixed. The couple that bought a house in Michigan got a 6.5% rate. Their monthly cost for a 100%-financed home—if the Bitcoin they pledged appreciates more than 8% a year, the Bitcoin paid for the house.
How is there no margin call?
Because we don't margin-call you on your house when it goes up and down in price.
Yeah. Interesting.
As long as you pay your mortgage on time, you're good to go.
Yeah. Interesting. What happens if Bitcoin—but, you know, a house doesn't go to zero?
Okay. Bitcoin can go to zero. As long as you're paying your mortgage on time, you're paying your P&I, you're good.
Then you're fine. Interesting. Interesting. When Mike and I first started building Blockworks, there was this whole idea of moving real estate onto the blockchain. There was a company called Harbor, and there were all these securitization companies in 2018 and 2019 tokenizing real estate.
It turned out, actually, tokenization probably made more sense for highly liquid assets. Tokenizing something doesn't make something that's illiquid liquid, right? What do you think? Do you have an updated thesis on tokenizing real estate, or is it more the tokenization of a mortgage that makes sense?
I think the tokenization of a mortgage makes a lot more sense because this is a market that trades $180 billion a day.
Yeah. Yeah.
Right. And so taking friction costs out of transactions that happen recurrently is a lot more valuable than taking friction costs out of transactions involving however many people are buying and selling and tokenizing their house. I do think that there is something to be said: Why should I own 100% of my house? Why can't I own 62% of my house and pay rent to the other 38% of the people? I think that's the conceptual basis behind tokenizing houses. But I think we're a long way away from that.
Yeah.
If you think about it, people think just because you put something on an exchange, it will trade. But how many public companies are out there that don't trade much at all?
Yeah, they trade less than $1 million a day.
Yeah. Yeah.
And so you can have a public company that's a $200 million company that trades $100,000 or less a day. I think tokenization of real estate doesn't make sense, but tokenization of a standardized product like a Fannie Mae mortgage or Freddie Mac mortgage makes a ton of sense.
And then what has been really interesting and encouraging is that the CEO and the presidents of Fannie Mae and the new administration, are so encouraging about the idea of not only enabling more people to be able to afford a home using a token-backed mortgage, but also the tokenization of those mortgages themselves. Fannie Mae and Freddie Mac, to date, have mostly sold their bonds to institutional investors. You, as a consumer, couldn't go out and buy a share of a Fannie Mae mortgage backed by a house and guaranteed by Fannie Mae.
Right. Right. Right.
And now you can—and you will soon be able to.
5. Partnering With Sky
That's right. You guys partnered with Sky.
Yes.
And, you know, our friends at Framework. The way I understand it, you guys partnered with Sky to basically secure access to a $500 million credit line.
That's right. That is backed by tokenized mortgages.
That's correct. Can you walk us through it?
If we think about it—and again, maybe I'm going to ruffle some feathers here—stablecoin issuers today are buying short-duration Treasuries. That was a really good thing for stablecoin buyers to invest in when short-duration rates were 4.5%. As rates come down, the yields become less and less attractive relative to keeping your money in the bank.
The surplus that they have, they have to invest in things that are great. If you think about duration, a Fannie Mae or Freddie Mac mortgage is possibly the single best long-duration asset out there because it's a fixed-rate mortgage. Today, you can buy a 6.5% mortgage guaranteed by the U.S. government or a U.S. GSE, and you earn that for a 30-year period of time.
Now, yes, you're exposed to prepayment risk. But if you're a regular person—if I borrow money from you and I pay it back early, what do you say?
Oh, no, no, no. Don't give me money.
You're happy to get your money back early. Institutions care about prepayment in the context of investment risk, but individuals are actually totally comfortable with prepayment risk as a natural thing. So I think it's an amazing product to put in tokenized format and then distribute to consumers. Basically, it will enable consumers to effectively have their own little bank.
Yeah. Yeah. Yeah.
Instead of being on the other side, instead of getting paid 2.5%—
I can get paid what Chase gets paid, 6.5%. Right. Right.
Chase just takes my deposit and puts it in Fannie Mae mortgages. Now I can do the same.
Yeah. Did you guys do any math on how much the cost of capital comes down for you by accessing the capital through Sky instead of the myriad of whatever that process looks like?
I think short-duration capital from Sky brings our cost of capital down by 25 basis points. Long-duration capital from Sky, and from tokenizing long-term, brings the cost of capital down by 100 basis points.
That means—really? Wow.
And that means we can take mortgage rates down by 50 to 100 basis points. It's huge. We can improve affordability by $1,000 a month.
Interesting. Are you doing that yet? I mean, look, if you can take it down 100 basis points, then—
Yeah, that's our goal.
Ah, that's interesting. Is it live?
No, the Sky facility is not live.
The Sky facility is not live. Interesting. So, wait, this 100 basis points—that's just because you get to remove all these intermediaries?
Yes.
Yeah. Yeah. Yeah.
No, there's an ETF called MBB. It owns about $40 billion of agency mortgages—agency FHA/VA mortgages, government-guaranteed mortgages—and the net yield on that, I saw, was 4.1%.
And the underlying assets, right? Today's mortgages are 6%.
Yeah. Yeah. Yeah. Right. And it's like the multiple layers: Fannie Mae is selling bonds, there's a securitization, then those bonds are getting bought by some fund, and so on and so forth. The multiple layers of financial intermediation that you relieve, each one 10 basis points or 20 basis points at a time, by going from consumer to capital and cutting all that out—that is huge.
Honestly, I've been doing fintech for 26 years. The original promise of peer-to-peer finance that tokenization brings is now out there.
Who does Sky replace in this sense, or who is Sky competing against?
BlackRock.
Sky is competing against BlackRock in that sense?
BlackRock.
You know, it's so funny. How long have you been doing fintech? You said—
26 years.
Yeah. You know, this is year 9 of Blockworks, right? I've been in crypto and, for a while, it was like the crypto zealots against the world. Now, for the first time, the crypto folks are the most pessimistic on crypto because everyone's been saying the tokens are all down. The fintech and institutional people are more excited about crypto and blockchain than ever. It's just interesting hearing your excitement for this stuff.
Oh, my God. You know why? Because I think it's like the first iteration of the internet. When people built all these companies, the zealots in Silicon Valley, from 1995 to 2000, were so high on it. Then, from 2000 to 2003, it was an apocalypse. Around 2003, people were like, “Wow, this stuff actually works now. You get a little Google IPO, and stuff's actually working. It's delivering user utility. It's not a dream. The pipes are laid.”
Obviously, the people who lay the pipes sometimes don't last.
Yeah.
But the work is done, and now that the pipes are laid, there's amazing utility to get unlocked.
Yeah. Can we go one level deeper there? What would you say to the crypto person who's been in the industry for 10 years? They're a little depressed because their tokens are down 90%, and they're trying to find a little bit of optimism for what the industry is going to look like for the next 10 years. What's your message to them?
I think the biggest thing is, if it drives consumer utility and it has distribution, it will be very valuable.
Yeah.
The biggest challenge so far has been: Which of these coins actually does something for the user? Which of these protocols actually does something for the user? Now, some of these lending protocols, if they make something where, effectively, you show up with collateral and they give you a loan that's instantly priced and run on a smart contract, you don't have the traditional framework of having to foreclose on the asset and so on and so forth.
That's super valuable.
You've suddenly replaced this entire layer of people who do loans in a physical environment. You can have one guy doing a pawn shop and another guy doing something else. Anything that can be tokenized can therefore be lent against without the physicality of all of it. That's really cool. Some of these protocols are going to be worth a lot of money. Those tokens are going to be worth a lot of money.
Yeah, 100%. What is your vision for, as all stocks, bonds, currencies and commodities—all capital markets—come on-chain and you have these tokenized assets, enabling all of these tokenized assets to be pledged for mortgages?
That's our goal.
That's the goal.
Yeah, that's the goal. We're starting with Bitcoin and USDC. The next step is conversations that are taking place with all the large ETFs—the S&P 500 guys, all that—
Top 100 public companies in America.
Yeah. Imagine you're an employee at a public company. You're an employee at Amazon and you get paid in RSUs, and you have to sell them. What's the point? You're paying people in stock, and the first thing they want to buy is a house. Once they've got enough stock, they're selling it.
So imagine you’re an employee at an XYZ company—at SpaceX, at Amazon—and you can just pledge your stock—
To go buy a house.
—and keep the upside of the stock and benefit from having the rent be the same for the next 30 years. That’s really cool.
Yeah, 100%. And the fact that the USGS sees that Fannie and Freddie are on board with this, right? There’s validation that this is something that’s good.
Yeah.
Flipping to anything else on crypto and blockchain?
I’m so excited. This is—
Yeah.
Tokenization is the realization of peer-to-peer finance in the original way that it was conceived 15-plus years ago and has not yet been able to be put in place.
Yeah, amazing. I love that. What is fundamentally broken in the mortgage process? We walked through what blockchain solves. What does AI solve here—more on the underwriting side?
What AI solves is the uniform application of the rules.
Yeah.
Minus the expert labor, the biases, and the lack of knowledge of the expert labor. The expert labor layer in mortgage is extremely expensive. The average mortgage underwriter gets paid $210,000. There’s no way that all of them remember all 45 investors’ guidelines, each 800 pages. Do you? That would be like remembering the Encyclopedia Britannica. How many humans remember the Encyclopedia Britannica front to back? No way. So the AI is able to do that, and therefore it improves the approval rate.
The AI is able to recommend things and do math in its head instantly. So it tells you, “Hey, if you pay off this loan, this loan, and this loan, it’ll increase your credit score, which will lower your mortgage rate by the same amount as the loan you just paid off. So that should be really great. You should do that.”
The last thing is, unlike most humans, the AI will say, “I can just do it for you. Just tell me yes.” I think the value of implementing AI is really enabling consumers to have access to the top 1% of salespeople, processors, underwriters, and their skills—but for the other 99% of Americans who don’t have that access.
Yeah. It democratizes access to financial advice and financial engineering.
Do you have salespeople at your company? Do you have sales brokers?
We do.
And will they all be—you know, the question they’re thinking—
The best ones—
—are embracing it?
The best ones are embracing it. They’re making AI twins of themselves. I think if you watch our earnings call from November, we literally have Ryan Grant, who’s the top loan officer in Orange County, California, and he now has an AI digital twin.
Interesting.
That AI digital twin has all of his colloquialisms, all of his implicit lessons, everything that he knows, and knows all the underwriting guidelines across all 45 investors. It can do math instantly.
Yeah. Do you think about this increasing your top line or decreasing it? Basically, are you going to grow AI, crypto, and all these things? Is it making the business more efficient, or is it actually increasing the revenue?
It’s both. It’s increasing the revenue because we’re capturing more surface area. We’re talking to more customers. We’re able to be open 24/7 when other people are not.
On the flip side, it’s lowering our expenses because it’s allowing us to have the people who are great at working with consumers have 10x leverage.
Interesting. Yeah. What’s the way you’re spending your time right now?
My biggest thing is the path to profitability.
Yeah. How close?
We’ve had 4 really tough years.
Yeah, yeah. You guys have had a journey.
Yeah, coming off the refi boom.
Yeah.
Reconfiguring the company, continuing to invest in the tech, continuing to invest in the AI. We’ve publicly been out there saying that we’re going to be breaking even on an adjusted EVA basis by September, and so I’m pounding away at that. I’m pretty confident that we’re going to be able to make that happen.
2 or 3 months?
Yeah, yeah.
Could you have done that without blockchain and without AI?
I think we could do it without blockchain.
Yeah.
But we can’t do it without AI.
Can’t do it without AI.
Can’t do it without AI. We have to be all in on AI.
Yeah.
All in. One of the very famous Silicon Valley VCs told me 3 years ago, “Vishal, the difference between any company and its peers, and any CEO and its peers, is going to be determined by how hard you can force AI down every level in your organization.”
The truth is, most of the people in your company don’t want it because it’s hugely disruptive to their day. It’s hugely disruptive to their workforce. It’s hugely disruptive to their personal sense of meaning and what they do because it’s able to do it better than they are. So you have to adapt, and you have to adapt really quickly.
We really pushed on that. I think the deep knowledge base, the knowledge graph, and all of the stuff that we collected from 2016 to 2021 helped us implement AI in a way that no one else in the mortgage industry has. That’s why I think we’re going to be able to get to our goal.
That’s the number 1 goal. Number 2 is finding more partners, doing more interesting things with tokenization and AI, and meeting more customers who have millions of customers—more fintechs who want to be in the mortgage and home equity business.
Yeah. What do you think the street misunderstands about your business?
Truthfully, I think there are a couple of things. One, I think the street views us as an online version of a traditional mortgage business and a traditional refinance mortgage business. Our stock price goes up or down based on whether interest rates go up or down.
Since we’ve lost money for 4 years, I think the street has a very high discount on whether we’re going to be able to make money unless rates come down. It views us almost as a call option on rates coming down.
I think the street misses the fundamental transformation that’s taken place in the business—from being a direct-to-consumer brand that did really well during the pandemic to now being an AI platform that’s modernizing the entire mortgage and home equity business.
6. Running a Public Company
If this were a private company, it would trade at a very different valuation than what it does today. It trades at something like 2 to 2.5 times current run-rate sales versus what AI companies trade at. But, again, that’s an opportunity for people. I’ve been buying the stock, as you know and have probably seen, and I think there’s an appreciation that will come. It will take a while.
Yeah, yeah. What if there are some founders listening to this who are thinking about taking their company public? There are a lot of crypto founders who listen to this. They’re at Series D stage, the last round was at a couple billion, and they’re looking at the public markets and getting a little excited.
Maybe there’s a lull right now because all the money’s been sucked into SpaceX and OpenAI and stuff, but maybe in 2027 they’re looking at the markets. What have you learned about running a public company?
I think you should be mentally prepared for things to go slower.
Yeah.
Running a public company, you should be really careful about the new board members that you put on because public-company board directors are very different from private-company board directors.
Private-company board directors are typically your venture capitalists, who want to maximize outcomes.
Supportive.
Supportive and maximizing outcomes. Public-company directors are typically loss-minimizing. You have to be very careful about who you put on. I learned those lessons.
The employee base is going to change. So many of your superstars—the layer below you—are going to take the money that they’ve now got as public stock and probably leave. So, what do you have in the layer below them and the layer below them?
If I had to do it over again, what I would have done is really make sure the layer below them and the layer below them was amazing. If I had an A as a number 2, I would have wanted an A+ as a number 2.1.
Right.
I think that’s probably the single biggest lesson I would take.
Interesting. How many employees do you guys have today?
Not counting the brokers and all those things, in corporate it’s about 150. Altogether, with the brokers, the processors, and the people in India doing the data entry, it’s about 1,000.
Okay, but it’s like 150 core?
Over 1,000.
5 years from now, do you have more or fewer employees?
I think it depends on the surface area that requires data labeling. Outside of data labeling and data learning, I think we’ll have way fewer employees.
On the data-labeling side, that’s a huge business in and of itself that can be monetized. You don’t monetize it because you monetize it for your own product.
That's right. But there are companies like Mercor and all. I mean, there are these huge data-labeling companies. I had dinner with a guy three nights ago, on Saturday night—
A $50 million deal—just that deal—for data labeling for one of the big companies.
Have you ever thought about spinning that off?
[gasps] Probably should have done it a long time ago. [laughter]
Yeah.
But no, I mean, because it's so bespoke to this particular process. I think it's actually one of the core moats.
Yeah.
We sat around labeling everything when other people just let it all die in some 550-page PDF that they ship off to a custodian. Now we have this moat, and it just keeps expanding. What's really interesting is, I tell people, there are 2 assets that are created every single time we make a mortgage. One is a financial asset, and the second is the context graph that we've just created, along with all of the data that we've just cleansed across the person, the asset, and all of those things.
Eventually, we'll have all houses in America having gone through our clean data-labeling exercise, all appraisals in America having gone through that, and all consumers in America having gone through that. That will then allow people to be able to finance instantly for anything.
Yeah. If you guys are wildly successful and you achieve all this, you are the largest mortgage provider in the US, in the world. You are the—
What is the Stripe for loans?
Stripe for loans. That's great. Cool. So that means the platform business becomes actually bigger than the direct business. Is that where you see things going?
Interesting. I think Better.com is a great brand, but it's not going to be better than all the other brands combined together, right?
Yeah. Interesting. That's cool. I like that. Is there anything that we missed, that we didn't cover?
No, thanks. This has been really great.
Yeah, good. Vishal, congrats on everything, man. Before the episode, I called a bunch of your investors and said, “What do I need to know about it? What do I need?” You have an excited group backing you.
Thank you so much.
Yeah, congrats on everything.
Just getting started.
Good. All right, man. Be well.
Thanks.