To Our Investors,

Three months ago I wrote two specific things: that we had begun managing this business as if our consumer asset class was entering a recession, and that the private credit market faced a structural duration mismatch that would force itself into the open. In the ninety days since, both calls kept proving out, and a third thread arrived that I did not spend much time on last quarter: how fast a company can convince itself that a person is replaceable by a model, and how fast it discovers it was wrong.

The headline reading is still nominally constructive, though softer than last quarter. The Bureau of Economic Analysis's advance estimate puts Q2 2026 real GDP growth at 1.5% annualized, decelerating from Q1's (revised) 2.1%. The Atlanta Fed's final GDPNow nowcast for the quarter had actually run hotter, settling at 2.6%, one of the wider nowcast misses I have seen in the two years I have been citing this series to you. The U.S. dollar sits near 98 against major currencies, essentially unchanged from last quarter. Unemployment is 4.2%.

But the standard reading is still the incomplete reading. The University of Michigan Consumer Sentiment Index registered a preliminary 54.0 in July, a modest recovery from the record lows earlier this year but still depressed by any historical comparison. Wage growth and inflation crossed and re-crossed each other this quarter as gas prices spiked and eased with the renewed conflict in the Middle East, a reminder that the underlying picture moves in weeks, not quarters. Investors have continued to press for redemptions across the largest private credit platforms, and at least one of them has now closed the door on redemptions altogether rather than continue to sell assets to fund them.

The aggregate still looks fine because the aggregate is still being held up by a narrow band of asset owners and high earners while the wage earner, our customer, continues to absorb pressure they cannot offset. This is the K-shaped economy I have been describing for several quarters now. It is also, increasingly, the environment in which the labor market itself is bifurcating, not just the consumer's balance sheet. That is where I want to start this quarter.

I had the chance this quarter to spend an afternoon with one of my heroes, Howard Marks, whose firm now manages on the order of $224 billion. His whole career has been a demonstration that a good, boring business, underwritten conservatively and held over a long enough time horizon, beats almost everything else the market tries to sell you as more sophisticated. That is the model for how I think about risk in an environment that changes this fast. I do not believe you can predict the future. What you can do is bound it: run the worst case, run the best case, and build a business that survives the distance between them. We do that with four tools layered on top of each other. Decades of direct lending experience inform the priors. Patented and patent-pending machine learning processes catch repayment and fraud patterns that conventional scorecards miss. Large language models adjust our risk strategy in near real time as portfolio performance shifts. And stock-and-flow modeling, a discipline I have taught for close to twenty years, lets us simulate the portfolio through historical worst-case scenarios, the way we did through COVID and the way we do on a standing quarterly cadence today.

The largest variable that changed since my last letter is renewed energy volatility tied to the continuing conflict in the Middle East. Four-week gasoline prices swung as much as 30% higher earlier this year before round-tripping most of the way back, and were up 6.9% on that same trailing measure as of early August.

That volatility briefly pushed headline inflation above wage growth for the first time in a while: the Atlanta Fed's wage growth tracker read 3.6% overall in June, and the Bureau of Labor Statistics' Consumer Price Index rose 3.5% over the same twelve months, with the June print itself coming in cooler than forecast as energy prices eased. Wage growth is, on the most current reading, again running fractionally ahead of inflation, but the two lines sit close enough together that a second energy spike would flip them. My read from last quarter still holds: a short-term shock is absorbable and even stimulative for short-duration loan demand; a year of sustained $110 to $150 Brent would be a different conversation. We have not seen that.

The Fed and Bond Market Disconnect

The most underappreciated signal remains the distance between the Federal Reserve and the long end of the Treasury curve, and if anything that distance widened this quarter. Kevin Warsh was confirmed as Federal Reserve Chair in May, generally understood to favor lower rates. Yet at its most recent meeting the Federal Open Market Committee held its target range at 3.50% to 3.75%, with three officials dissenting in favor of a hike, not a cut. The market's response was to send long rates higher, not lower: as of this letter, the 10-year Treasury yield has climbed to 4.72%, its highest level since January 2025, and the 30-year sits at 5.25%, both up meaningfully from where they stood when our investor deck was finalized earlier this quarter, and both at levels last common well before this rate-cutting cycle began. For the consumer, mortgages, auto loans, and credit cards remain expensive regardless of what the Fed committee intends. A representative pattern in our own applicant pool: a borrower with a stable paycheck and a clean repayment history gets declined twice by a conventional lender, not because the underlying risk is bad, but because the traditional scorecard requires a cleaner box-check than a real life in this economy tends to produce. That pattern is a meaningful part of why our applicant mix keeps skewing toward higher quality even as volume grows.

Labor: The AI Layoff Trap

The labor market itself sent a genuinely mixed signal this quarter. Job openings retook the lead over the number of unemployed people for the first time in this cycle: 7.4 million openings against a 4.4% vacancy rate, versus a 4.2% unemployment rate, both as of June. That is a constructive rotation on its face

But it sits alongside a second, faster-moving story: AI-attributed layoffs have run to roughly 87,714 year to date in 2026, more than the prior three full years combined (4,247 in 2023, 12,742 in 2024, 54,836 in 2025), per Challenger, Gray & Christmas and the Forrester AIQ Index.

What is almost comedic, and instructive, is the following slide: 55% of companies that laid people off citing AI have already come to regret it, and in a meaningful share of cases have quietly rehired. Forrester's own data on why is worth sitting with: AI success rates run 58% on single-step tasks but only 35% on multistep tasks, and a quarter of leaders admit they do not actually know which roles benefit from the technology they are cutting headcount around.

We have not participated in that trade. Across this company's history, we have continued hiring, particularly in technology and analytics, on the theory that if an individual's productivity is compounding, you want more of those individuals, not fewer. The gap I hear about consistently from other CEOs and community leaders is an upskilling gap, not a headcount gap: the people and companies who treat AI adoption as a continuous-learning problem are pulling away from the ones who treat it as a cost-cutting problem.

Our platform processed 7.1 million loan applications in a single representative week this quarter, of which our lender partners' models selected roughly 4,195 for approval, an acceptance rate under 0.2% that reflects how much of the funnel we are able to see and price before a dollar is ever put at risk. We hold on the order of 21,000 individual data points and attributes on every applicant who reaches us, drawn from a partner network that has processed applications from tens of millions of Americans since inception. That scale is what allows the next two paragraphs to be specific instead of anecdotal.

Median loan size increased quarter-over-quarter to $650, up meaningfully from roughly $600 in recent prior quarters. Median bank balance across the applicant pool increased to $190. Overall consumer expenses increased 3% quarter-over-quarter, a real deceleration from the 7% pace we reported last quarter. Food and living expenses remain the single largest reason borrowers give for a loan, consistent with the pattern in prior quarters. Read together: the applicant is carrying a slightly larger loan, holding a slightly larger cash buffer, and seeing expense growth cool. None of that is a boom. All of it points the same direction, and that direction is stabilization rather than deterioration.

The composition behind those averages is what matters most. Forty-four percent of mid- and upper income consumers now say it is harder to obtain credit than it was a year ago, per the Federal Reserve Bank of New York and the Experian Economic Strategy Group, versus roughly a third as recently as two years ago. That is the traditional credit channel tightening exactly as I described last quarter, and it is precisely the mechanism that keeps pulling higher-quality applicants into our funnel who would not have considered installment lending when conventional credit was easier to obtain. Lower-quality applicants remain in the pool too, because their financial position continues to erode. This bimodal expansion, both tails growing at once, is the same pattern we watched at the onset of COVID, and it remains a reliable leading indicator of where the credit cycle is heading next.

One important counterweight: U.S. household debt to GDP sits at 68.8%, well below the pre-2008 peak above 100% and essentially flat with where it stood a year ago. Despite genuine and specific pressure on the lower and middle of the income distribution, the household sector in aggregate remains structurally less levered than it was heading into the last major credit event. Whatever this cycle produces, the shape of it will look different from 2008.

Blue Owl's OBDC II, a $1.6 billion vehicle, moved this quarter from selling assets to fund redemptions to permanently closing its redemption window altogether, eliminating quarterly liquidity for its investors rather than continuing to sell into a market that does not believe the marks. That is how the private credit story I described last quarter has gone: not resolved, but evolved, and not gently. BlackRock's HPS managed lending fund, a $26 billion vehicle, received redemption requests equal to 9.3% of assets and honored roughly half of that, capped at the standard 5% quarterly limit. Morgan Stanley's North Haven Private Income fund saw requests equal to 10.9% of assets and paid out $169 million, also capped at 5%. Cliffwater's flagship fund received requests equal to roughly 7% of assets and is expected to cap at the same 5% ceiling. Blackstone's BCRED, the largest vehicle in the category at $82.5 billion, is the exception that proves the rule: it received $3.8 billion in requests, equal to 7.9% of assets, and paid out every dollar of it in full, backstopped by a $400 million capital injection from Blackstone's own balance sheet.

You can see the same story in net asset value. Hive has held at par, one hundred cents on the dollar, in every quarter we have measured it to date, validated independently by our third-party administrator and auditor. BDEBT, BlackRock's non-traded vehicle, sits at 98.5 cents on the dollar. BCRED, which just demonstrated it can honor redemptions in full, still trades at 95.4 cents on the dollar against par, continuing to slide from the high 90s where it sat as recently as two years ago. When you lend Hive a dollar and we lend that dollar out, it remains worth a dollar. When you give a dollar to a vehicle marked at 95 cents, you may still collect a yield, but you have already lost part of your principal on paper, whether or not the fund ever gates you.

I will repeat what I wrote last quarter, because the market keeps validating it: the underlying credit assets across this industry are not, in aggregate, bad. The problem is duration. A platform that promises daily, weekly, or monthly liquidity against five- to ten-year illiquid loans has a structural mismatch that becomes a crisis the moment capital flow reverses, regardless of how sound the underlying loans are.

Hive's Structural Position

Hive's structure was built to make this category of failure highly improbable, not to react to it after the fact. The fix is duration matching: investor capital is committed for three years while the underlying loan assets mature in twelve months or less, and our newer three-, six-, and nine-month products shorten that average further as they scale. As a matter of policy and structure, we do not cross-collateralize our paper, we do not securitize the portfolio for short-dated liquidity, and we do not give any single capital partner the right to pull credit lines on short notice. I have described before a Florida-based lender that built a half-billion-dollar book over a decade and lost the company in a single week because a major investment bank in their capital stack, with cross-collateralized covenants across multiple lines, withdrew all at once. That is still the lesson on the negative space: what you deliberately choose not to do structurally matters as much as what you choose to do.

Our lender partners' Q2 FY2026 net unit margin is 27.0%, holding at the prior quarter's level. To restate the definition for newer investors: that is the post-default, post-cost margin on every dollar deployed, after marketing, data, technology, operations, payment processing, collections, and call-center costs are all subtracted. Across six consecutive quarters that figure has progressed 22.0%, 23.1%, 23.0%, 25.0%, 27.0%, 27.0%. The plateau is the point. Margin improvement stalling at a healthy level is exactly what should happen when underwriting tightens into a stress environment rather than loosens to chase growth.

May was the single largest lending month in the company's history, ahead of what is typically our seasonal peak in the back half of the year, and 20% of our volume now comes from repeat customers. That second number compounds in a way the first one does not. Nothing tells you more about a customer's ability to repay, assuming continued employment, than how that customer performed on their last loan with us. A repeat customer arrives with functionally zero incremental marketing cost and a demonstrated repayment history, which is why we can respond with more capital, at better terms, and still improve portfolio quality at the same time.

YTD return through Q2 FY2026 is 6.84%, bringing cumulative cash-on-cash return since inception in mid 2017 to 128.23% (manager-reported, gross of investor-level fees; methodology disclosed in our Q2 FY2026 Performance Report). Over the same window, the ICE BofA U.S. High Yield Index has returned 59.3% and the S&P 500 has returned 124.5% on the comparison basis we have used in this chart every quarter. We are not a hedge fund, and the comparison to public equity indices is imperfect by design given the very different risk and volatility profile involved, but the data is the data, and Hive has now compounded ahead of the S&P 500 since inception on this basis in addition to remaining meaningfully ahead of high yield credit.

Hedge Fund Comparison

We sit above four of the platforms in this comparison and behind only two, Bridgewater and Millennium. The composition behind that number matters more than the ranking: our return is produced from a portfolio of duration-matched, contractually defined cash flows, not from leverage on directional positioning. That is a meaningfully different risk profile than most of the funds we are listed alongside, even where the headline return is similar.

Credit Fund Comparison

The gap between Hive and the next-closest name has continued to widen since we began running this comparison. Duration matching, conservative underwriting, and consistent deployment compound. The par chart above is the mechanism; this table is the result.

Capital Deployment and the Capacity Constraint

Through Q2 we have deployed $27 million year to date against an FY2026 forecast of $60 million, our largest deployment year on record, with $10 million currently committed for the balance of the year. AUM stood at $168 million at quarter end and has continued climbing since; as of this letter it sits at approximately $177 million. The constraint on this business remains capital, not opportunity. We could take on another $50 million of committed capital tomorrow and still hold that same 27% unit margin, because the demand-side ratio of qualified application volume to deployable capital across our partner network remains comfortably above 10 to 1.

I want to address the question I get asked most often: with margins this strong, why aren't they going higher? It is the right instinct and the wrong conclusion. Our 27.0% unit margin is in part a sign of capital scarcity, not an achievement to maximize further. With meaningfully more capital, we could bring margins down toward the high teens by serving more of the qualified demand we already see and turn away today, generate substantially more total dollar profit in the process, and extend better terms to better customers. A well-run lender at ten times today's scale earning a high-teens unit margin generates more compounded cash over time than the same business at today's scale earning 27%, and the larger version is also less exposed to any single vintage going bad. Scale, in this business, is a risk-management tool before it is a profitability tool.

Our three-, six-, and nine-month products continue to scale alongside the standard twelve-month term, and the strategic value is compounding faster than the headline growth number suggests. Shorter duration turns capital faster, but the more important effect is on customer selection: a shorter term expands the addressable market to higher-quality customers who would not accept a twelve-month commitment, while compressing the window over which any individual loan can go bad. The Multi-Arm Bandit and Thompson Sampling framework that governs our underwriting allocation was built to handle exactly this kind of dimensional expansion, and each additional quarter of seasoned data on the newer products tells us how aggressively to lean into the mix.