Whether I am building a business from scratch or buying a stock, I look for the same four things:

1. A large and growing market, ideally more than $10B

2. A 10x better product

3. Compelling unit economics

4. A moat that will allow the company to capture the benefits of their unit economics for a very long time

In public markets, there is an additional fifth condition that must be met: the market should not have already priced in all of the above.

Personalis ($PSNL) satisfies all five conditions for me.

The company has developed NeXT Personal, an ultra-sensitive blood test that can detect cancer recurrence far earlier than traditional methods. The test works by finding circulating tumor DNA (ctDNA), the tiny fragments of DNA that cancer cells release into the bloodstream. By picking up these traces, NeXT Personal can identify recurrence up to 12–18 months before a CT or MRI scan would show anything. In effect, Personalis has turned a simple blood draw into an early-warning system for patients already diagnosed with cancer.

At $7.30 a share, $650M in market cap, $174M in cash and zero debt, the market thinks Personalis is a sub-scale, no-moat, cash-incinerating diagnostics company whose best in class blood test for cancer recurrence (NeXT Personal) has no chance against Natera ($23B) and Guardant ($8B) because both of them have larger sales forces pushing “good enough” alternatives.

I believe the market is wrong and this disconnect presents an opportunity to 4x (and potentially a whole lot more) your money in the next 18 months with an imminent catalyst in Q4 ‘25.

The catch, and the source of this opportunity, is one of perspective. I am not a traditional healthcare investor. That said, my background as a hyper curious startup founder who spends a lot of time thinking about software and executing Go to Market (GTM) motions is precisely what allows me to see something the market is missing.

TL;DR: The NeXT Personal test is widely acknowledged to be best in class. The bear case is distribution; the market doesn’t think the test will be widely adopted. My counter is that the market hasn’t fully appreciated the value of PSNL’s exclusive distribution partnership with Tempus ($TEM, $15B), which is connected to more than half of U.S. oncologists and integrated across over 2,000 hospitals. Tempus owns 14.5% of PSNL and 2 weeks ago filed to have the option to increase their stake to 19.99%. Tumor-informed testing begins with a tissue biopsy, and Tempus controls that workflow with EMR integrations and turnkey billing. For clinicians already using a Tempus lab, adding NeXT Personal will soon effectively become a one-click decision, similar to an “add to cart” on Amazon. This isn’t something you could figure out staring at PSNL’s backwards looking financial statements, but its pretty obvious if you’ve spent time building software products. If NeXT Personal can capture 6% market share by 2029, there is a clear path to $100M of EBIT from NeXT Personal alone which currently accounts for <1% of PSNL’s revenues. At a 20x EBIT multiple, that implies roughly $2B versus a current EV of ~$500M. The rest of the business generates about $85M of revenue today and has significant embedded upside optionality.

Investing can be as easy or complicated as you want it to be. I’ll lay out a simple narrative style investment case for PSNL in Part 1, and then in Part 2 I am open sourcing a full blown hedge fund style write up with PSNL’s TAM, unit economics, a valuation model, and risks. Pushback is welcome and encouraged!

Disclaimer: I own a chunk of Personalis in a personal capacity and on behalf of funds I manage for UHNW families. I am extremely biased and this write up is not financial advice. Do your own research, don’t borrow conviction and no crying in the casino.

Part 1: The (Simple) Investment Case for PSNL

I define variant perception as holding a well-founded view that’s meaningfully different from the market consensus. - Michael Steinhardt

PSNL’s product NeXT Personal is a blood test that can detect whether cancer is coming back (i.e. recurrence) 12-18 months before CT scans pick it up.

Being able to detect the recurrence of cancer early is very important because:

1. it allows doctors to start/change treatment sooner which saves lives and

2. it enables biopharma companies to reduce the length of their enormously expensive cancer drug trials by proving or dispoving their drugs work faster

NeXT Personal currently accounts for less than 1% of PSNL’s $85M in revenues. So you’d be right to wonder why the NeXT Personal tests are not contributing to revenues yet.. is it a moonshot product still in development? Is the product an overhyped scam? No, NeXT Personal is being used by over 600 oncologists, has 98% retention(!!) and adoption has been growing exponentially every month for the last 24 months:

Woah, that’s pretty impressive. If NeXT Personal were a startup, VC’s would be tripping over themselves to throw money at them. So why isn’t NeXT Personal showing up in PSNL’s revenues?

Great question. The reason is because the NeXT Personal test isn’t currently reimbursed by Medicare or private insurance yet. As a result, Personalis has been eating the cost of each test as they drive adoption in advance of anticipated Medicare reimbursement in Q4 2025. This has set up a bizarre dynamic where the faster NeXT Personal grows, the larger ’s losses appear.

So when the NeXT Personal test receives insurance reimbursement, a giant revenue spigot that isn’t being captured in their financials today turns on? Yes. That is pretty interesting, but ’s management must be shouting this from the rooftops? The market must have priced this in, surely?

Until 30 days ago, was a $400M market cap company (today it’s $650M and still cheap IMO) so I don’t think many investors are paying attention. But among those who do follow this sector, the general consensus is that tiny Personalis has no hope of scaling their NeXT Personal product against the giant scaled players like Natera ($NTRA, $23B) who currently dominates this market with their Signatera test.

Signatera uses whole-exome sequencing (WES) to create a personalized blood test that tracks 16 of a tumor’s mutations. Signatera is on track to be the most profitable diagnostic test of all time (great blog post by Edward Larkin about Signatera here). 10/7 Update: A reader informed me that Natera now has a new test with 64 variants. That’s still fewer than PSNL’s 1800 and directionally proves that testing for more variants is a good thing.

Personalis is the challenger. Their NeXT Personal test was designed from the ground up on a fundamentally different (and imo better) architecture. It starts with whole-genome sequencing (WGS), giving it a far more comprehensive map of the patient’s cancer. From this superior map, it creates a monitoring panel that tracks up to 1,800 unique tumor variants.

Natera (who until this year was a customer of PSNL, until PSNL launched a competing product), recognized the threat and has recently attempted to counter PSNL by introducing an updated Signatera that also uses a WGS front-end. This test only monitors 16 variants (vs 1800 for PSNL) but allows Natera’s salesforce to create the perception of parity.

For now, the market is assuming that the tweaked Signatera is “good enough”, i.e. doctors will not switch despite PSNL being vastly superior because Personalis doesn’t have the marketing resources to dislodge Signatera. That is the consensus. I believe it is wrong.

Firstly, I think NeXT Personal isn’t just slightly better, it is dramatically better. If a random retail investor with no science background and a ChatGPT subscription can figure this out, I’m pretty sure oncologists can too. Gold standard studies like TRACERx have demonstrated that ultra sensitive tests like the one being pioneered by PSNL have nearly 4x better detection rate in Stage I lung cancer (57% vs. 14% found by a more limited, Signatera-like test) or this paper in GenomeWeb where NeXT Personal demonstrated higher baseline sensitivity compared to another gene-mutation based test (71% vs 30%). Pardon my French, but this is a big F-ing deal, especially for cancer detection. Read those papers and ask yourself which test you’d want your doctor ordering if a loved one unfortunately got cancer?

What the market has further failed to appreciate is Personalis’ Go-to-Market strategy. PSNL has embedded their test directly into the workflow of its distribution partner, Tempus ($TEM, $15B market cap), which already has touch points into 50% of oncology practices, runs the biopsy labs for more than 30% of U.S. hospitals and has over 200 full time sales reps. Ordering the test is going to soon be as easy as clicking “add to cart” on Amazon. Billing is already set up, so saying yes takes one click, while saying no means extra work to go outside the system and order a test they already know the patient needs.

Again, I’m not a professional healthcare investor, but I do spend a lot of time mapping workflows and building software products in my day role as a startup founder. I’ve mapped out below what ordering the NeXT Personal test would look like for an oncologist/nurse practitioner once Tempus deploys their one click integration vs ordering Signatera. The key takeaway here is that ordering Signatera will actually require more work than NeXT Personal. People tend to default to the option that has the least amount of friction.

I am making the bet that Tempus will implement this one-click upsell for NeXT Personal. It’s what any half decent product manager would do, and if you spend 5 minutes on Tempus’ website you’ll see that they are pretty damn good at building software. Now might also be a good time to mention that Tempus owns ~14.5% of Personalis and, on Sept 11, 2025, amended their standstill to allow open-market purchases up to 19.99% 👀.

You’d be right to worry that Tempus has all the leverage in this relationship and will make the bulk of the economics. Turns out however that Personalis negotiated a fixed cap of $9.6M on the distribution fee paid to Tempus, laid out in a publicly filed contract that apparently none of the sell side analysts on Wall Street has read, making the NeXT Personal test unit economics incredibly attractive. Here is the contract: https://contracts.justia.com/companies/personalis-inc-8225/contract/1260239/

Why would Tempus give Personalis such a sweetheart deal? Well, Tempus makes gobs of money by anonymizing and selling the genomic sequencing data generated from these tests to pharma companies who use that hard-to-get data to develop drugs. These data licensing deals are extremely lucrative, so Tempus has a strong incentive to help PSNL as much as they can. More tests performed by PSNL = more data for Tempus to sell.

Bonus: If you read the contract I linked above carefully, you will also see that Personalis is entitled to 10-20% of Tempus’ data licensing revenues generated from their tests. It’s early days and its hard to say how big this could get, but this is pure margin revenue for PSNL when it turns on and not at all priced into $PSNL today.

Sidebar: I am in San Francisco at the moment watching a friend’s startup go from $0 to $10M in revenue within a week (true story) after he figured out a way to capture data not easily available elsewhere. No comment on the durability of his business model, but it is undeniable that the big AI labs are all thirsty for proprietary data right now. PSNL’s data licensing agreement with Tempus isn’t core to my thesis, but I do think both Tempus and PSNL are sitting on dramatically under-appreciated data assets.

Valuation

The NeXT Personal test is the core of my thesis. It contributes under 1% of revenue today, yet with 6% market share of MRD recurrence testing I think there is a realistic path to $100M of EBIT by 2029. At a 20x EV/EBIT multiple, this one product alone is worth about $2B against a ~$500M enterprise value today. Could they get more than 6% marketshare? That is upside. Beyond NeXT Personal, Personalis currently makes roughly $85M from three lines of business: biopharma services (with partners like Merck and Moderna), population sequencing supported by government contracts and the aforementioned nascent data licensing deal with Tempus that the market has completely missed. Each of these businesses could be meaningful in their own right but I don’t include them in my valuation.

Imminent Catalyst

PSNL’s breast cancer, lung cancer, and immunotherapy response monitoring tests are expected to gain coverage in Q4 2025 to Q1 2026 (all have been submitted and awaiting MolDx approval for reimbursement). Personalis should then receive roughly six months of retroactive payments, and every additional test going forward converts into recognized revenue. Volumes are already compounding exponentially ahead of reimbursement and management has already signaled that they will step on the gas post reimbursement, so a revenue inflection and dramatic re-rate is around the corner.

Could I be wrong? Sure, and the failure point will be either a) PSNL does not get insurance reimbursement or b) there is an execution stumble - scaling a lab business is orders of magnitude harder than scaling a software business. The former seems unlikely given how strong PSNL’s data is, but you never know. The latter almost feels like an advantage rather than a disadvantage in the age of AI where software increasingly feels commoditized.

Regardless, the downside is limited with 2 years of cash on the balance sheet, no debt, and likely takeout interest from Tempus (who own 14% and just filed to increase that to 19.99%) or Merck (who also own 15.8% of PSNL and have the right of first refusal to mitigate against a take-under from Tempus). If I am right, this is a grand slam. If I am wrong, there’s probably 40% downside from here. Given the probabilities I assign to the two paths, it feels highly asymmetric to me. When I find situations like this, I take my shot ¯\_(ツ)_/¯

Subscribe now

Part 2: The Numbers Supporting the Investment Case for Personalis

Value investing is at its core the marriage of a contrarian streak and a calculator. - Seth Klarman, Baupost

Here’s a hedge fund style deeper dive into Personalis covering:

• What do they do and how do they make money

• How big is the TAM

• Unit Economics of the NeXT Personal test

• Valuation

• Competition

• Risks

What Personalis Does and How They Make Money

At its core, Personalis is a genomics company transitioning from a research-focused services business into a high-growth clinical diagnostics company. Its revenue comes from two distinct areas: a legacy biopharma business and a new, rapidly scaling clinical platform.

Biopharma and Research Services

• In 2024, this part of the business generated approximately $85 million in revenue and is on track for similar revenues in 2025 despite losing a large contract with Natera who released a competing product.

• Personalis provides comprehensive genomic sequencing and data analysis to pharmaceutical companies to support their clinical trials.

• Their main product in this segment, ImmunoID NeXT, helps drug developers understand tumor biology and optimize patient selection.

• This segment also includes large-scale population sequencing projects for customers like the U.S. Department of Veterans Affairs.

Clinical Diagnostics (The crux of my thesis)

This segment currently generates almost no revenues today. It is centered on two products:

• NeXT Dx: A test that analyzes the genetic signature of a patient’s solid tumor to help guide therapy. It has Medicare reimbursement, proving the company can navigate the regulatory and payment landscape.

• NeXT Personal: The flagship minimal residual disease test for detecting cancer recurrence. While it is not yet generating significant clinical revenue pending broad reimbursement, adoption is strong, retention is at 98% (Q4 2024 earnings call) with test volumes up 59 percent in the most recent quarter and set to accelerate massively post reimbursement. It is already used and paid for by biopharma customers in drug trials, with that revenue stream expected to grow 300 to 400 percent in 2025.

In short, a growing services business is currently funding the launch of a transformative clinical platform. The value will unlock when reimbursement decisions convert NeXT Personal’s test volume growth into a high margin, recurring revenue stream.

The TAM

I estimate about 5 million surveillance tests per year in the U.S. across breast, lung, and colorectal cancers (all cancer types PSNL has tests for).

My calculation is based on the number of U.S. patients in major solid tumors who are in the surveillance setting, meaning they have received initial treatment and are now being monitored for recurrence.

First, I do not look at the number of newly diagnosed patients each year (incidence). The true market is the total number of patients who are alive and being monitored at any given time (prevalence). A patient diagnosed with Stage II/III cancer may be monitored for 5 years or more.

I focused on the three largest initial markets for MRD testing (all of which PSNL has a test for):

• Breast Cancer: ~300,000 new cases/yr

• Lung Cancer: ~240,000 new cases/yr

• Colorectal Cancer (CRC): ~150,000 new cases/yr

Together, these three account for nearly 700,000 new diagnoses annually in the U.S.

I estimate that roughly 50% of these patients are diagnosed at a stage (II or III) where they receive treatment with curative intent and then enter long-term monitoring.

700,000 annual cases x 50% eligibility = 350,000 new patients entering the monitoring pool each year.

Assuming an average monitoring period of 4 years, the total pool of patients under surveillance at any given time is: 350,000 patients/year x 4 years = 1.4 million eligible patients.

Testing Frequency: The standard clinical protocol for MRD surveillance is quarterly testing to detect recurrence as early as possible. I assume 4 tests per patient, per year (and I’m being conservative. Medicare allows for up to 1 test per month).

TAM: Multiply the number of eligible patients by the annual testing frequency.
1.4 million eligible patients x 4 tests per year = 5.6 million annual tests.

I round this down to a conservative 5 million annual tests to account for imperfect patient compliance and variable physician ordering patterns.

Note: This number represents only the breast, lung, and colorectal cancer markets. When I include other major solid tumors like bladder cancer, melanoma, and head and neck cancer, the true long-term TAM is substantially larger.

Unit Economics of the NeXT Personal Test

To triangulate the unit economics, we need to know:

a) the average selling price per test (i.e. what the reimbursement rate from Medicare and private insurance companies will be) and

b) how much does it costs PSNL to manufacture and distribute their tests.

Average Selling Price Per Test

On the Q2-25 call, Personalis’ CFO framed this as “about $7,000 per patient over roughly four tests”, which implies ~$1,750 per draw in a full-reimbursement scenario. While I’d be over the moon if they got this pricing eventually, I think a more conservative scenario is $1145 per test in 2026 (roughly in line with Signatera) which scales up to $1500 in 2029 which feels reasonable given that coverage is likely to broaden and Medicare Advantage/commercial compliance tightens.

After coverage is approved, management plans to apply for ADLT status. Under CMS rules, new ADLTs are paid at list charge for the first three quarters, then reprice to the weighted median of private-payer rates. If granted, this can lift realized pricing above my assumed $1500 base in 2029, but that is upside.

How much does it cost to make a test i.e. COGS

I triangulated today’s lab cost from what the company discloses. Each quarter they provide total revenue, reported gross margin, and the gross-margin headwind from unreimbursed clinical tests. Multiply the headwind percentage by total revenue to get the dollar headwind. Divide by clinical tests delivered to get the implied cost per test. This isolates the cash cost of tests that hit COGS without offsetting revenue.

This shows PSNL’s scale at work: ~$1,088 → $933 → $755 → $593 per test from Q3’24 to Q2’25.

My 2026 base assumes lab COGS of ~$650 per draw plus a Tempus operations fee of ~$200 per draw, for all-in COGS of ~$850. By 2029, lower sequencing and kit costs, tighter pipelines, higher utilization, and a lower ops fee take all-in COGS to ~$525.

One of the under-appreciated aspects of PSNL’s business is that approximately 50% of PSNL’s COGS (sequencing costs and compute) are falling dramatically year over year, driving gross margins higher.

Gross Margin Per Draw

• Today: $1,145 ASP minus $850 COGS = 25% unit gross margin.
  

• 2029 base case: $1,500 ASP minus $525 COGS = 65% unit gross margin.

How much money can PSNL make from the NeXT Personal Test?

Given PSNL’s superior product and distribution partnership with Tempus, I think there is a reasonable chance they get to 6% market share in MRD cancer recurrence testing in the next 4 years. The ride will probably be lumpy instead of the smooth curve I’ve modeled below, but you get the idea..

As demonstrated above, I think there is a COGS path to about 65 percent gross margin. This is what that could look like:

Operating Expense Discipline: As volumes scale and their platform matures, I think SG&A as percent of revenue can go from 35 percent in 2026, to 17 percent in 2029. R&D as percent of revenue can go from 60 percent in 2026, to 25 percent in 2029. Again, these are broad directional assumptions but it intuitively makes sense. Note: The SG&A and R&D costs should be distributed over their entire revenue base not just NeXT Personal but you get the idea..

Putting it all together, this is how I think NeXT Personal generates $100M in EBIT by 2029.

So the NeXT Personal product alone can do $100M in EBIT in 4 years with some fairly reasonable assumptions. We can debate what the right multiple to stick on this is, but this business is worth a lot more than $500M in enterprise value today. A 20x multiple gets us to $2B.

But wait, there’s more!

PSNL currently does $85M in revenue excluding NeXT Personal. Their biopharma business is actually growing rapidly (after adjusting the loss of Natera as a customer, who launched a rival product and a few delays from pharma customers). I’m doing them a huge disservice by sticking this business at the end and lumping it into “upside” because this biopharma business might just be a whole franchise in itself. Intuitively you can probably grok that any business that can shave off a year from a billion dollar drug trials is probably going to be able to capture a lot of value. That is what this segment has the potential to do.

I’m going to very briefly go through each of these business segments and show some very lazy, conservative, hand wavy math for illustrative purposes. The key take away is that there is significant, explosive, upside inside Personalis.

A) Biopharma testing & services

What it is: Trial support for pharma (tumor profiling/WGS, trial MRD endpoints, regulatory reporting) under MSAs; analytics and dashboards reuse across programs.
2029E economics: EBIT $36M (shared lab/compute leverage from clinical MRD).
Valuation: 15× EBIT ⇒ $0.54B EV.

B) Cancer-vaccine sequencing (option)

What it is: End-to-end sequencing and neoantigen calling for personalized cancer vaccines. This cancer vaccine is almost a bet the company moment for Moderna and they are set to do a read out in Q1 2026. MRD monitoring tail per patient (package ≈ $1.1k per new patient).

I think Moderna has a low probability of succeeding given the difficulty of what they are trying to do and the political environment, so I am assigning $0 to this, but these numbers will look bonkers if they do pull it off.

C) Population sequencing (government)

What it is: Large-scale, grant/award-driven sequencing for public programs; capacity smoothing and reference datasets, with pass-through costs capping margins.
2029E economics: EBIT $9M.
Valuation: 10× EBIT ⇒ $0.09B EV.

D) Tempus data royalty participation

What it is: PSNL share of Tempus third-party data licensing; minimal incremental cost. (Assume $200M 2029 third-party royalties; PSNL at 15% ⇒ $30M revenue.)
2029E economics: EBIT $24M (80% margin).
Valuation: 15× EBIT ⇒ $0.38B EV (strategic, dependency risk).

Valuation: SOTP (midpoints, 2029E)

• Core MRD (NeXT Personal): $2.00B

• Biopharma testing: $0.54B

• Cancer-vaccine option: ?

• Population sequencing: $0.09B

• Tempus royalties: $0.38B

  
  Total SOTP EV: $3.0B

Discount 4 years at 12% ⇒ PV of $2.0B before net cash and any dilution.

Competition

I’ve framed this as a battle between Natera (Signatera) and Personalis, but Guardant and Quest have competing products too. The reason I don’t spend too much time on competition is because a) I think PSNL has the best in class product and 2) my whole thesis hinges on them getting 6% market share which should be quite do-able given the Tempus partnership.

Risks

Believe, but buckle up - bumper sticker I saw on a Tesla in SF

1. The single biggest risk is that PSNL gets denied insurance reimbursement and/or there is a major delay. If this happens, be prepared for a 40% chop.
   
   Mitigant: The data is exceptionally strong, the product is already being used by over 600 oncologists in practice and the company is ramping up their own sales team and hiring billing specialists. Every indicator points towards imminent reimbursement approval, but you never know crazy things do happen.

2. Execution risk: What looks like a simple blood draw actually relies on insanely cool and difficult technology. Scaling this up will come with its own challenges. Can they go from 0 to 6% market share in 4 years? There may be some execution stumbles along the way.
   
   Mitigant: They’ve spent the last 2 years eating the cost of the NeXT Personal test precisely to plan for this reimbursement moment when they can step on the gas. That’s sufficient time to iron out most wrinkles.
   
   Conclusion

   
   If you’ve made it here, congrats you are in the top 1% of curious people in the world. I think the risk / reward on PSNL is pretty damn compelling. We will have an answer on the reimbursement question pretty soon and I expect the stock to react violently (in either direction) when that happens. Let’s find out!
   
   Stay curious,

   Rush

I started buying Blend ($BLND) in May 2023 (avg $0.80/share) and still own it. The price today is $3.18/share. In 2 years I think there is a reasonable chance Blend is worth $6/share and potentially a whole lot more.

This is a long post, so I have split it into two parts:

Part 1: I share how I discovered Blend and a few lessons learned.

Part 2: A deep dive investment write up where I explain why Blend’s best days are still in front of them and what Wall Street isn’t appreciating, yet.

Disclaimer: This is not financial advice, I am very clearly biased, no crying in the casino etc.

Part 1: Finding a Diamond in the Mud

July 16, 2021: After 9 years and $665 million in venture funding, Blend Labs ($BLND) went public at $21 a share, good for a $4.4 billion market cap.

March 19, 2023: An anonymous Twitter account, @Dalibali2, unveiled his now-infamous “ShitCo Index”; a basket of 38 VC-backed SPACs and IPOs all taken to the woodshed and down 95-99% from their post Covid ‘21 euphoria highs.

March 20, 2023: I randomly came across the post and immediately drafted a tweet telling Dali exactly where he could stick his index. At that point, I was three years into banging my head against a wall working 80-100 hour weeks trying to get my own venture backed startup to the promised land of product market fit. Regardless of the performance of those stocks, I knew on a visceral level that every single ticker on the “Shitco” index represented the life’s work of a founder who had taken a very real risk to create something out of nothing.

Thankfully, before rage Tweeting at Dali, I came to my senses and realized that he had identified something extraordinarily compelling. The post covid crash of 2022-23 had effectively turned these battered listings into a public-market proxy for late-stage VC. All the companies in his list had some sort of product-market, none of the early stage startup blowup risk, and now anyone could buy these companies for literally a fraction of the total VC funding that had gone into them.

Dali took an equal weighted basket approach to owning all of the companies in his index. So I figured what if I, a startup founder and product builder by day, could go through the same list in my free time and identify individual companies that had legitimate products?

So over the next several weekends I systematically went through the 38 companies in Dali’s index, kicked the tires on the products that these companies were selling and identified three companies (Blend, Carvana and Cardlytics) that I thought had leptokurtic upside.

May 5th, 2023: I realized that while Wall Street had crushed Blend’s share price by 97%, the company had built a best in class platform, expanded its market share from 12% to 20%, grown their unit economics by ~50%, and taken out $100 million in operating costs all during a toughest mortgage industry operating environment of the last 50 years (including the the Great Financial Crisis).

When life presents you with opportunities like this, you have to swing hard and swing big. So I made Blend a 40% position at an average price of $0.80/share ¯\_(ツ)_/¯

June 20th, 2025: Blend is at $3.18 per share, I still own it and I think on a risk-adjusted basis, Blend is more compelling today than it was back then.

Lesson #1: One Man’s Shitco is Another Man’s Treasure

Eighteen months later, the ShitCo Index beat the NASDAQ by more than 80% (see Dali’s follow-up for the numbers). My three picks from the basket; Blend, Carvana, and Cardlytics did meaningfully better.

The point isn’t to take a victory lap, I’m trying to illustrate a core lesson: returns largely flow from the pond you choose to fish in.

For investors: Bombed-out sectors and geographies often have the best asymmetric payoffs. Sidebar: U.S. biotech, the single worst-performing sector of the past three years, looks really interesting to me here. AI has made it easy for anyone with agency, not just domain “experts,” to get up to speed on any topic in the world. You can do the work on individual names if so inclined, or just grab the sector ETF ($XBI) as 2025’s ShitCo proxy.

For founders: Paradoxically, you want to launch in markets where sentiment is max bullish. Raising capital is easier, TAMs are expanding, and customers are open to new solutions. Right now, that market is obviously AI. We are in the first innings of a once in 5 generation opportunity and ignoring AI as a founder (assuming you’ve found a real problem to solve) is just straight up reckless. This is not the time to be a contrarian, run towards the heat!

Lesson #2: One Fish Can Feed the Village

If you can get the pond (market) right, often that single decision alone will literally pull great returns or companies out of you. Don’t believe me? Imagine investing $100 into each of the 38 companies ($3800 total) in the Shitco index on the day the list was published. Now, let’s assume that 37 of the 38 companies in the index went to zero. The remaining 1 company was Carvana at $7 (true story). Today, you’d have $4700 dollars or a 25% return in 2 years, not too shabby.

Lesson #3: Keep A Radically Open Mind

I subsequently exchanged messages with Dali and it turns out that he’s thoughtful and routinely shares great ideas. Go follow him on X if you don’t already!

I’ve been fortunate to interact with some of the world’s sharpest investors in both public and private markets and I’ve noticed a pattern: the best investors always seem to respond to cold DMs/emails and somehow make the time to talk to yahoos like me. I think it is because they don’t judge people on the credentials they come wrapped in; they are constantly underwriting the world for themselves. It is an incredibly exhausting way to approach life, yet the market consistently rewards independent thinkers, so act accordingly.

Part 2: The Investment Case for Blend ($BLND)

Price: $3.18

Fully Diluted Shares: 325M

Market Cap: $1.0B

Cash & Equivalents: $109M

Debt: $0

Enterprise Value: $926M

Target Price & Expected Return: $6 in <2 years, ~100%

What does Blend do and how do they make money?

Banks and lenders use Blend’s software to let their customers easily apply for financial products like mortgages, personal loans and bank accounts online. Each successfully completed transaction triggers a small fee, making Blend a digital tollbooth on the flow of money between institutions and their customers.

• Mortgage Suite (~58% of 2025E revenue): Blend earns fees when a loan is funded using its software. They earn about $78 per loan plus ~$15 for related services (income verification etc) which are delivered via 3rd parties operating on Blend’s platform.
  

• Consumer Banking Suite (~34%): Blend earns fees when consumers open or use deposit accounts, personal loans, or HELOCs via the bank’s Blend-powered interface.
  

• Professional Services (~8%): One-time fees for setup, integration, and custom work.

Elevator Pitch In Startup Founder Speak: Blend’s product is awesome, competitors are weak, their founder is a killer and their profits are going to grow a lot when either home prices fall or interest rates come down. Some of the world’s best private equity and hedge funds are buying Blend at these prices. Buy it, forget it and thank me in 5 years.

Elevator Pitch in Hedge Fund Speak:

• Platform Transition & Margin Inflection
  Blend is converting its mortgage-only franchise into a multi-product banking platform, driving stickier, less-cyclical revenue. As this mix shift accelerates, gross margin, unit economics, and share penetration should all inflect well above current consensus.
  

• Setup & Upside
  At $3.18 the stock already generates FCF even at trough mortgage volume and carries substantial operating leverage. The trade offers rare “recovery torque” in mortgages plus secular growth from commercial banking; a >2x re-rate inside 24 months is realistic as rates normalize and cross-sell ramps.
  

• Forecast Delta vs. Consensus
  Street: 75% GM / 14% top-line growth FY-26.
  My view: 80% GM / 25–30% growth.
  Valuing the consumer-banking segment (45% CAGR, non-cyclical) at a 12x sales (in line with recent acquisition comp) implies you are effectively buying the mortgage franchise at ~3x mid-cycle operating income.
  

• Embedded AI Option
  Market has yet to price Blend’s data-advantaged position. Owning the top-of-funnel for banks gives first-party data access, enabling a suite of AI products that can materially lift economic value per funded loan (evPFL). Any AI-driven revenue should command a significantly higher multiple on discovery.

Bottom Line: Consensus underestimates both cyclical recovery and structural expansion. The risk-reward skew is attractive; doubling the share price in <24 months is well within base-case parameters.

Ok, I’m intrigued, can you give me more details Rush?

Sure, I was hoping you’d ask. Here’s a deep dive on Blend covering the following topics:

1. Generational Founder in the Making

2. Blend’s Distribution Advantage & Platform Moat

3. Unit Economics: Compelling & Inflecting

4. Competition

5. Valuation

6. Who Else is Buying Blend?

7. Risks

Generational Founder in the Making

“Never underestimate a man who overestimates himself” - Franklin D Roosevelt

Great companies are forged by exceptional people, and Blend has one in founder-CEO Nima Ghamsari. A Stanford CS grad and early Palantir engineer, Nima owns 5.9% of the company and controls 59% of its voting power.

When I first encountered Blend in early 2023, I almost passed after seeing Nima selling large chunks of his stock near the lows. My knee-jerk reaction was that I didn’t want to own shares of a company where insiders were abandoning ship. A closer look, however, told a very different story. Here’s what was actually going on:

After the IPO, Nima borrowed against his Blend shares to.. Wait for it.. buy even more Blend shares 🤯 The subsequent 70% collapse mortgage volumes drove the share price down, triggered margin covenants forcing him to sell, which in turn drove the price even lower. How did I know this? Because he published an open letter on Blend’s website explaining it to anyone that was paying attention:

“…I increased my personal stake in Blend by borrowing against my existing shares. The sharp decline in fintech valuations triggered loan-to-value covenants that forced me to sell a portion of that collateral… I remain a long-term holder and continue to control the company.”

Most founders use an IPO to (understandably) take some chips off the table. Nima, who literally moonlighted as a semi-professional poker player in college, used his IPO to borrow money to double down. Just pause for a minute here to recognize how insane and rare this is.

Many investors will look at this and think of Nima as reckless. To be fair, Nima did drink too much of the 2021 koolaid. He even signed up for an Elon Musk style options package that incentivized him to 10x Blend’s share price.

So while he did get high on his own supply, dismissing Nima as reckless completely ignores the prior 9 years of great decision making that got Blend to an IPO in the first place. More importantly, this level of self conviction (often erroneously mistaken for arrogance) is exactly what I actively look for in a founder/leader. All courage and no skills just gets you a kamikaze pilot (fun fact: despite all the lore surrounding them, no dive bombing kamikaze pilot ever sunk an Allied fleet carrier, battleship, or cruiser). But when you pair next level confidence with smarts and product building chops which Nima clearly has, you get a relentless founder in it for the long haul worth betting on.

As a side note, I also had a brief interaction with Blend’s CFO, Amir Jafari, formerly of ServiceNow, and found both him and the materials produced by his IR team to be impressive.

Blend’s Platform Distribution Advantage

“He who controls the spice controls the universe.” — Baron Vladimir Harkonnen, Dune

A company’s long-term destiny is determined first by its people and second by the quality of its product(s). The financial statements that Wall Street loves to obsess over are lagging indicators of these two things. The good news is that Blend has a best-in-class product and, even more importantly, is now transforming that strength into a durable platform moat.

Why banks need Blend

To understand why Blend has an advantage, it's important to understand why Blend’s customers need solutions like Blend in the first place.

The reason is that most banks and lenders are simply really bad at building software. To fill the gap, they started buying point solutions from various SaaS vendors like Blend. This however led to a spaghetti of systems that no one could fully untangle:

Blend’s misstep and 2023 course-correction

Early fintechs including Blend initially ironically repeated their customers mistake, selling siloed modules. In 2023 Blend rebuilt everything on Blend Builder, a single data and workflow layer that fixes the issues in the table above.

Importantly, Blend’s competitors have NOT truly unified their various offerings into a platform i.e. they are still a hodgepodge of different products that they have acquired at different points in time being held together by duct tape and web hooks underneath the surface, and this gives Blend’s platform a competitive structural advantage.

Tellingly, competitors still rely on acquisitions to introduce new products. To illustrate my point, here is one (of many) examples: when rising rates pushed banks to prioritize capturing deposits, Alkami paid $400m for MANTL. Blend didn’t acquire anyone, they just built an in-house deposits module and pushed it live through Blend Builder making it instantly accessible to all their customers if they wanted it.

Result: Blend has been able to grow their economic value Per Funded Loan (evPFL) from $69 in Q1 2022 to $99 in Q3 2024, a 43% increase in the face of a 70% decline in mortgage volumes, a heroic achievement. They were able to do this because they are leveraging their entrenched distribution into financial institutions and selling new products via Blend Builder.

Per Nima’s comments on the Q1 2025 conference call, Blend’s largest customers who are using all their products are already at an evPFL of $170. Intriguingly, the evPFL today is a tiny fraction of the total cost of originating a mortgage ($10,697). Translation: Blend is undermonetizing their product today and their customers aren’t going to mind too much if Blend raises prices in the future, an excellent position to be in as a business.

Lastly, Blend put out a roadmap (below) for their product expansion plans which showed a path to an incremental $325 in evPFL. Most excitingly, they have designs on expanding into the Sales and Fulfillment piece of the origination process which is much juicier. I think they are being conservative with the $325 estimate because if they are able to deliver on even half of what is being promised on this roadmap, they will capture a lot more given the nature of the products they are moving into. This does however put them on an eventual collision course with the 800 pound gorilla ICE (they own Encompass) who are currently mostly downstream of Blend, so it’s something to keep an eye on.

Their consumer banking business? Same story. Most investors didn’t even take this segment seriously 2 years ago, and now it is a $45M ARR run rate business growing at a 45% CAGR. Here are their RPO numbers, which tell the same story: Blend’s platform is gathering steam.

Notably, their RPO target for 2025 was $150M and they ended up exceeding it in Q1 2025.

Unit Economics

“If the unit economics don’t work, the business can’t work” - Fred Wilson, USV

Blend doesn’t give us the split in unit economics of their two businesses: mortgage and consumer banking but we can try to estimate it.

Mortgage suite: We know that in Q1 2025 Blend earned an evPFL of $93. This number has been declining for the past 2 quarters because of a mix shift towards smaller customers (independent mortgage banks) and new customers onboarded into the platform who are still ramping up.

Historically Blend provided add-on services like KYC, verifications, insurance, etc. in-house, eating data-provider fees and support costs that suppressed contribution margins. They are now flipping that model: they are plugging trusted third-party vendors into the platform, letting those partners run the heavy ops, and keeping a software-like take-rate on every transaction. As the mix shifts from owned services to partner-delivered services, each incremental dollar flows through at near-pure gross profit.

My variant perception on mortgage gross margins: By keeping core-software margins at ~77% and lifting third-party services to 90% while they represent roughly 16% of mix, the Mortgage Suite’s blended gross margin rises to ~80% (0.84 × 77% + 0.16 × 90% ≈ 79%).

The gross margins on the Consumer Banking suite are more straightforward. The combined software platform runs ~79% gross margin, and the legacy Mortgage Suite (still 66% of software revenue) is at ~75–77%. Plugging in the numbers therefore puts the Consumer Banking Suite at roughly 84–87% gross margin today.

Putting the two together, I model 80-82% gross margins vs the street who are still modeling 75% gross margins even into 2026.

Operating expenses Blend gives us (below) of ~65% in 2025 which I reckon will go to 60% by 2026. This leaves us with a 20-25% Operating Profit Margin.

The most important takeaway: Blend is transitioning away from their service heavy business model and essentially now operating an App Store where third party providers operate on Blend’s rails to provide banks and lenders services like income verification, title insurance etc. This business model shift will result in Blend’s gross margins inflecting higher and the company operationally more focused on building great software. The App Store model is incredibly powerful and starts forming a real compounding moat for Blend; just as the App Store has served as a moat for Apple, Google, Shopify and Salesforce. TL;DR Blend is in excellent company and on the right path.

Competition

Mortgage Suite: Blend’s only credible rival IMO is SimpleNexus, which is owned by nCino. My tire kicking led me to the conclusion that SimpleNexus lags Blend in functionality, and Nima Ghamsari has noted for two consecutive quarters that Blend is the sole vendor appearing on bank RFP short-lists. It’s revealing that, even though nCino and Blend share several major customers (with nCino serving their consumer needs and Blend mortgage), as far as I can tell by cross referencing logos using the Wayback machine, SimpleNexus was unable to coax any of those clients onto its platform, even when Blend was on the brink of bankruptcy. Roostify has languished inside CoreLogic, while ICE’s Encompass sits farther downstream. The bigger threat IMO comes from Rocket and United Wholesale Mortgage, which compete directly with Blend’s bank customers for origination share. But with only ~40% of U.S. mortgages completed online, the addressable runway remains massive for everyone.

Consumer Banking Suite: This market is more crowded: nCino, Alkami, Q2, and others are active and have competitive products, but Blend’s suite is outpacing its own targets. Lenders appear drawn to Builder’s one-platform delivery of both mortgage and consumer-banking products, giving Blend a differentiated edge even in a competitive field.

Valuation

Blend’s Consumer Suite is expected to make $62M ARR in 2026 and be growing at ~35%. MANTL was just acquired by Alkami for $400M for 13x 2025E revenues which are growing at 30%. Using this as a comp, we can use a 12x multiple for Blend’s consumer business and estimate it will be worth $744M in 2026.

Estimating the value of the Mortgage Suite is trickier because revenues are highly dependent on mortgage volumes (which are inversely tied to rates). Given that we appear to be at the peak of the rates cycle and there is considerable political pressure to reduce interest rates, and there is a ton of pent up demand for re-finances, it’s not crazy to think that mortgage volumes will go back to ~7.2M by 2026 which is the 20 year historical median mortgage volume. Blend currently has a ~20% market share, I think it is fair to assume they’ll be at 25% by 2026. Blend currently makes $93 per funded loan. They’ve introduced a whole bunch of new products and existing current cohorts of customers will mature, so it’s not crazy to think evPFL will be $130 by 2026. Putting it all together:

Essentially at $3.18/share today you’re buying Blend’s best in class mortgage platform for 3x next year’s estimated operating income. For context, autos which are also cyclical trade and less profitable trade at 16x operating income on average.

If we assume the mortgage business is worth 5x EV/Sales or 20x Operating Income, the stock doubles.

Bonus: Blend invested $2.5M in Bilt’s seed round at a $60M valuation. It looks like Bilt is on their way to being a decacorn (last valuation of $3.25B), so we could see a return on this hidden asset on their balance sheet in the future if Bilt goes public or gets acquired.

7/11/2025 Update: Turns out Blend invested in Bilt at a $350M valuation not $60M. I erroneously got the $60M number from a press article that misconstrued the round size ($60M) for the valuation ($350M). Bilt also announced a private round of funding on 7/10/2025 at a $10.75B valuation.

Who Else is Buying Blend?

“Oh my God, it’s like we’re sharing a brain!” — Alexis Rose, Schitt’s Creek S02 E02

While it is important to do independent work and reach your own conclusions in investing, it’s good to know other (much smarter) investors have also converged on Blend seeing the opportunity.

Brian Sheth’s (formerly President of Vista Equity Partners) new fund Haveli Investments invested $150M in Blend via preferred that convert at $3.25/share and Brian sits on Blend’s Board. Extraordinarily thoughtful and high conviction public markets investor Dennis Hong of Shawspring has been adding to his BLND position per their latest 13-F filing. Joe Lonsdale, cofounder of Palantir and seed investor in Blend through 8VC took the unusual step of publishing a public letter defending Blend. Lightspeed, another early investor, remains fully invested in the company. Peter Thiel led their Series C. Last but not the least, the company was buying back (small amounts) of their own shares in the open market at $3.30/share this past quarter.

Risks

“Trust Allah, but tie the camel” - Arab proverb

1. Rates stay flat or go higher: Blnd’s destiny for better or worse is tied to interest rates that are outside of their control. In a world in which rates go higher, the Blend re-rating story will get delayed.

   Mitigant: I think we’re at the point where higher rates will cause a meaningful home price correction, making homes more affordable offsetting some of the demand destruction. If you want to get fancy (I never get fancy), you could hedge out your rates exposure via derivatives.

2. Capital allocation: Blend made a disastrous acquisition of Title365 from Mr. Cooper in 2021. They paid $450M for it and just recently sold it off to Covius for an undisclosed sum (wouldn’t surprise me if they gave it away for free). Past shareholders have paid the price for this acquisition and the goodwill’s already been written off, but another bad acquisition will sink the company.

   Mitigant: With Dennis Hong and Brian Sheth at the table and a new CFO (ex ServiceNow) in place, I don’t think there is a high probability of this happening.

3. As alluded to in the competition section, both Rocket and UWMC are making massive investments in technology. They compete directly with Blend’s customers so there is a chance that they start taking share from banks and lenders on an accelerated basis.

   Mitigant: Only 40% of mortgages are currently done digitally. There is a strong structural tailwind of digitization that is greater than any share loss. Further, Blend will continue to win new customers offsetting any share loss.

4. Feels a bit silly to send out an investment idea the same weekend stealth bombers are dropping bunker buster bombs on nuclear research facilities, so I’ll just say it: if oil spikes up leading to sticky inflation or a nuke is launched, we’re in trouble.

   Mitigant: In the long run this should be noise but if it isn’t we’ll have bigger problems to worry about.

If you’ve made it to the end, I’d welcome your feedback and/or pushback on this write up.

Stay curious,

Rush

Subscribe now

Coming Out of Stealth

Dear friends,

I’ve decided to start writing publicly about two parts of my life that I have up until now kept firewalled: building startups and investing in public companies.

If you are receiving this email, I value your opinion and either want to keep in touch with you or get back in touch with you. So I’ve taken the liberty of subscribing you into this newsletter. Below I’ll explain what I’ve been up to and what you can expect from my writing, but if you’d rather not receive this please feel free to unsubscribe - my feelings won’t be hurt, pinky promise.

1. What will this newsletter cover?

To give you a flavor for what you can expect, here are a few titles of future blog posts:

1. Investing in the Age of Infinite Intelligence

2. What Torching $5M Dollars and 6 Years of Your Life Feels Like

3. Meta $5000

4. What Consoling 9000 Heartbroken Women Taught Me About Value Investing

5. Why I Invested 40% of My Net Worth In a “Shitco” Run By a Pro Gambler

And more 🙂

2. Why should you give one half of two schnitzels about what I have to say?

This is an excellent question, so let’s look at the data:

My record as a startup founder: I’m 0 for 2. My first startup was bootstrapped and my second was venture backed with millions of dollars by some of the best early stage investors on earth. Both were zeroes. I’ll have a lot more to say about this later, but for now all I’ll say is ouch.

My record as a public markets investor: I've been fortunate to be a concentrated investor in a number of companies including Shopify (58x), Sea Ltd (30x), Carvana (25x), Meta (6x) and Blend Labs (5x). On the basis of this track record, I've started managing a small amount of external capital for a few families.

I have no f*cking clue if I’ll be able to keep this up and I’ll be the first to acknowledge the role of luck in investing. Compounding with small starting amounts of capital is a lot easier than compounding hundreds of millions of dollars and I’m also aware of the importance of humility in investing, so this will be the first and last time I share these numbers publicly. Still, I thought it important to establish my bonafides.

3. Why start writing now / what took you so long?

I finally managed to resolve an existential knot that has been vexing me for my entire professional life: am I a tech bro or a compounder bro?

I’d like to say it was the countless hours of therapy or months of travel that helped me resolve this identity crisis, but actually it was anonymous Buddhist monk #2 in Episode 1, Season 3 of White Lotus (Same Spirits, Different Forms) who dropped this pearl of wisdom:

“Identity is a prison. We build the prison, lock ourselves inside, then throw away the key.”

Hot damn.

As those words pierced my skull, I brushed the Dorito dust off my pajamas and realized I didn’t need to pick between these two identities.

There was a third option that felt deeply true to me and allowed me to escape the identity prison: I’m just a deeply curious person.

IMO (good) startup founders and hedge fund investors tend to have diametrically opposite personalities and hate each other's guts, but really they are both just abnormally curious people who look for opportunities and act on them with force of conviction.

4. What is in this for you Rush?

Another great question, understanding incentives is important! If you are receiving this email from me, I want to stay in touch with / reconnect with you because I value your opinion. Also, a big motivation to write publicly for me is to get disconfirming evidence and get pushback on my (frequently heretical) ideas, so if anything I write tickles your brain, please hit reply!

5. What’s next?

Before I let you go, I have a quick favor to ask you:

1. Can you reply "received" to this email?

2. If this email ended up in a spam, promotion, or social folder, please move it to your primary folder! On Gmail, you can do this by clicking the three dots in the top right corner of this email, selecting “move”, and clicking on “primary”.

"Dude, you’re sending me emails on a Sunday… I just want to read your blog in peace. Why the extra steps?" Another great question! Let me explain:

Replying "received" tells your inbox to trust SecurityAnalysis.ai emails, so they won't end up in your spam folder.

The whole process will take less time than it took you to read this sentence so far: one "received" reply to this email, and move the email to your primary inbox if needed.

6. Bonus

We’ve entered an incredible, unprecedented period of history with everything around us accelerating all at once. As a startup founder I’m an unapologetic optimist. As an investor, I’m a bit of a skeptic. Navigating this period is going to require equal doses of both and it will be fun to do it together.

So until the next post I’ll leave you with my favorite artwork; Maxell’s “Blown Away Guy” (pictured below). The man is shown desperately clinging to the armrests but defiantly looking ahead at the source of the music through his sunglasses while his hair, necktie, the lampshade and his martini glass are being blown back by the tremendous sound from speakers in front of him supposedly due to the audio accuracy of Maxell’s product. Can’t think of a better metaphor for the paradigm we are in.

Stay curious,

Rush

www.securityanalysis.ai

https://x.com/updating_priors