The most dangerous thing in financial markets is a number that is technically correct but strategically misleading. GDP growth of 3% sounds robust. An unemployment rate of 4.4% sounds near-optimal. But these are aggregate statistics, and in an economy growing increasingly K-shaped, aggregates can obscure more than they reveal.

One year into President Trump's second term, the tariff picture has been notably less inflationary than many analysts feared. Aggregate CPI has remained relatively stable — moving sideways with modest downward drift on an overall basis. The U.S. dollar, which some predicted would deteriorate sharply under trade confrontation, has held firm near the 97 level, well within historical ranges and far from crisis territory.

But the tariff story is still being written. China's nominal GDP growth has effectively flatlined in recent quarters even while Chinese exports hit record levels — a paradox that creates complex dynamics for global supply chains and domestic price pressures. The trade reconfiguration will take years to fully manifest in consumer price data, and we are watching it closely.

The most underappreciated macro risk is the one that rarely makes frontpage news: real wages for lower-income workers have been negative on an inflation-adjusted basis for an extended period.⁶ This is not a technical recession indicator. It does not show up in the Sahm Rule. But its effects are entirely visible in the spending behavior, savings behavior, and loan application patterns we observe every quarter.

Consumer sentiment, as measured by the University of Michigan, registered 56.6 in the final February 2026 reading, with 46% of respondents citing high prices as a persistent drag on their personal financial situation.⁷ These are not the numbers of a population feeling prosperous. They are the numbers of a population managing.

Perhaps the single clearest recession signal in the current data is the relationship between job openings and unemployment. Through 2022 and into 2023, job openings dramatically outnumbered unemployed workers — a historically unprecedented imbalance that drove wage inflation. As of December 2025, that dynamic has inverted. The U.S. Job Vacancy Rate stood at approximately 4.0% while the unemployment rate reached 4.4%,⁸⁹ meaning unemployed workers now exceed available job openings. This is not catastrophic. But it is a structural shift. The labor market has moved from employee surplus to employer parity — and trending toward employer advantage.

The tech sector is the most vivid illustration. Google, which grew headcount by nearly 70,000 employees in the pandemic hiring wave, has been aggressively reducing that footprint.¹⁰ In 2025, over 127,000 tech workers were laid off across the industry. The intellectual debate — whether this represents a permanent AI-driven restructuring or a mean reversion to 2019 staffing levels — is genuinely unresolved. I believe it is both. The normalization thesis explains some of the numbers. But the AI displacement thesis explains the structural direction of travel.

The one sector defying the gravity of the K-shaped economy is AI infrastructure. Top hyperscalers — Amazon, Google, Meta, and Microsoft — are entering 2026 with a combined annual CapEx trajectory approaching $600 billion, with data center investments alone projected to exceed $400 billion this year.¹¹ Dell'Oro Group projects that AI-driven data center CapEx will reach $1.7 trillion annually by 2030. As I noted in my presentation, approximately 1% of U.S. GDP is now flowing into data centers and AI infrastructure — an investment wave that has made cloud computing and infrastructure investors fabulously wealthy while reshaping the productivity landscape for every other sector.

For Hive, this wave matters for two reasons. First, it confirms the direction of the technology curves we have been riding for years — the AI tools that are now compressing risk modeling from a week-long simulation to five minutes, the agent-based systems that are scanning news feeds for employer layoff signals, the natural language interfaces that are making our entire data stack queryable in real time. Second, it sets the context for why AI productivity gains are so unevenly distributed: the companies positioned to capture them are thriving; the companies that cannot adapt their workforces are shedding jobs. Our consumer lives at the intersection of these forces.

I want to be precise about something that gets lost in macro analysis. When we say "the consumer is under pressure," we are not describing a statistical abstraction. We are describing specific people facing specific crises — and we have the granular data to see it with unusual clarity.

Our platform now holds data on approximately 50-60 million Americans. We have originated almost 300,000 loans across six lender partners. Every quarter, we run what we call Hive Insights — a systematic analysis of the applicant pool, spending behavior, and financial health indicators of the population that turns to installment lending. What we see this quarter is stark.

As of late 2025, 67% of American adults live paycheck to paycheck according to PYMTS's widely cited research,¹² with some surveys reporting even higher figures depending on income bracket and methodology. Approximately 13% of adults report being completely unable to cover an unexpected $400 expense by any means,¹³ while estimates for those unable to cover a $1,000 emergency from savings alone exceed 50% of the population. These are not edge-case statistics. They describe the core financial reality of the majority of American households.

The data we see in our applicant pool is consistent with — and in some respects more acute than — the national picture. Median bank balances among our applicants declined significantly quarter-over-quarter to $175.¹⁴ Think about what that means in practical terms: a single car repair, a medical co-pay, or an unexpected utility bill creates an immediate financial crisis for the median person applying to our lenders. The most frequently cited reason for loan applications this quarter was food and living expenses — cited by 32% of applicants — followed by emergency and unexpected expenses. Overall applicant expenses increased 6% quarter-over-quarter.

With 82% of American adults carrying at least one credit card, and revolving balances at cycle highs the structural cost of credit for lower-income consumers is compounding in troubling ways. Unlike our installment loans, which use simple interest and are structured to allow early payoff, credit card debt uses compounding interest. A consumer carrying a balance at 20-29% APR and making minimum payments will pay three to five times the original balance over the life of the debt. This dynamic is not peripheral to our business thesis — it is central to it. The consumer seeking our lenders' products is often not making an irrational decision. They are making a rational comparison: a fixed-rate, fixed-term installment loan with a clear payoff date versus a revolving credit card balance with an indefinite repayment horizon and compounding interest.

The most alarming single data point in the consumer stress picture is auto loan delinquencies. Fitch Ratings' auto ABS data shows the subprime 60+ day delinquency index reached 6.15% in December 2025 — a record for a December reading.¹⁵¹⁶ Meanwhile, vehicle repossessions reached 1.73 million in 2024, the highest since 2009. Average new vehicle prices are now approximately $50,000, and the combination of elevated purchase prices, 7-8% financing rates, and rising insurance costs has created an affordability wall for lower-income consumers.

We want to be explicit about what this data means for our portfolio. While we do not make auto loans, the subprime auto delinquency trend is historically one of the earliest and most reliable leading indicators of broader consumer financial stress. The pattern we saw in 2006-2007 — subprime auto and mortgage delinquencies rising before mainstream credit metrics deteriorated — is visible again in the subprime auto data today. We have been watching this signal closely, and it informed our decision in October 2025 to formally begin modeling our business for a recessionary scenario in our asset class.

Here is the counterintuitive dimension of the story: consumer financial stress is simultaneously our most significant risk factor and our most powerful demand driver. In Q4 2025 alone, almost 20 million unique Americans applied for installment loans across the lenders we work with.¹⁷ 20 million. In a single quarter. That is roughly 6% of the U.S. adult population seeking access to credit products — and it represents expansion on both ends of the quality spectrum.

Higher-quality consumers who would not previously have considered installment lending are now applying as their savings cushion depletes and traditional credit products become less accessible. Simultaneously, lower quality consumers remain in the pool as their financial position continues to deteriorate. This bimodal expansion is exactly what we observed at the onset of COVID-19, and it is a reliable leading indicator of a market slowdown. From a business model perspective, it means we have more high-quality customers to select from — which is precisely why we have been able to maintain and improve our underwriting metrics even as the broader consumer environment weakens.

I have been saying since the early days of Hive that there is a structural mismatch in the private credit industry that would eventually come to a head.

The model is straightforward to describe and deeply problematic in practice: large platforms have been offering daily, weekly, or monthly liquidity to retail investors while deploying capital into five- to ten-year illiquid underlying assets. This works beautifully when capital is flowing in. It breaks catastrophically when capital starts flowing out.

That moment has arrived.

On February 18, 2026, Blue Owl Capital announced the sale of $1.4 billion in direct lending investments — assets from three of its BDC vehicles — at 99.7 cents on the dollar, with proceeds earmarked to return approximately 30% of net asset value per share to investors.¹⁸ This transaction was not an isolated event. It was the front page of the Wall Street Journal for a reason: it is the visible edge of a much larger phenomenon.

Robert A. Stanger & Co. data shows Q4 2025 redemption requests for nontraded BDCs accelerated dramatically quarter-over-quarter, with the trajectory suggesting $30+ billion in annual outflows across the non-traded BDC market if the pattern continues.¹⁹ Apollo's private credit vehicles have seen material portfolio markdowns.²⁰ Multiple fund managers are implementing gates — restrictions that prevent investors from withdrawing their capital — precisely because the underlying assets cannot be liquidated quickly enough to meet redemption demand.²¹

This is the Jimmy Stewart scenario. In "It's a Wonderful Life," the run on the Bailey Building & Loan is terrifying not because the loans are bad, but because the duration mismatch makes the fund illiquid at exactly the moment when depositors want their money back. Today's private credit equivalents made the same structural error at scale.

From the very beginning, we designed Hive to avoid this trap. Our investor capital is committed on three-year terms. Our underlying loan assets mature in twelve months or less — and with our new three-, six-, and nine-month product offerings, the duration shortens further. This match between liability term and asset duration is not an accident. It is the architectural choice that allows us to say, with complete confidence, that we will never gate this fund.

When the market is stable and capital is cheap, this structure looks conservative. When the private credit markets are facing liquidity crises, it looks prescient. We prefer not to need vindication, but we are gratified that the model holds.

In another context, I might lead with the performance numbers. But I've always believed that performance without context is marketing, not communication. Now that you understand the environment we're operating in, the performance data should read differently.

Our lender partners' Q4 FY2025 end-consumer net unit margin is 27% — meaning that for every dollar lent, after all defaults and loan-level unit expenses including data and technology costs, IT infrastructure, payment processing, call center operations, collections, and customer acquisition costs, the lenders return 27 cents of net margin on each loan originated.²² This is not a trailing twelve-month figure. It is our current operating performance, validated against a conservative modeling approach that assumes the worst of each vintage until the loans have fully matured over a 12-24 month cycle.

To put this in perspective: our unit margin has improved consistently from 22.0% to 23.1% to 23.0% to 25.0% to 27.0% over consecutive periods. This improvement occurred during a period of macroeconomic stress and consumer financial deterioration — which is the expected behavior of a well-designed anti-fragile system. More consumer stress means more demand. More demand means better customer selection. Better customer selection means lower defaults.²³

From inception through Q4 FY2025, our cumulative cash-on-cash return to investors has been 121%+.²⁴ We have consistently outperformed the ICE BofA High Yield Index and the S&P 500 on a risk-adjusted basis. I say this not to be immodest — we are a direct lending fund, not a hedge fund, and the benchmark comparison is not the right frame for evaluating our strategy. I say it because the data is real, and because investors deserve to know how their capital has performed against the alternatives available to them.

The most meaningful comparison, now that the private credit market is going through its stress test, is against the large direct lending platforms. BlackRock's BDEBT, Blackstone's BCRED, and Apollo's ADCF have all experienced meaningful performance headwinds in 2026. We continue to outperform. We remain confident that our focus on balance-sheet lending with strong unit economics and proper duration matching will demonstrate its full advantage as the private credit repricing cycle plays out.

People sometimes ask what gives us our edge. The honest answer is that it's a combination of duration, data, and mathematics — none of which is glamorous, and all of which compound over time.

Our risk management framework rests on four pillars: decades of direct lending experience, patented AI and machine learning processes for repayment and fraud detection, large language model-based dynamic risk strategy adjustment, and stock-and-flow modeling for cash flow optimization. At the core of our underwriting is an adapted Kelly Criterion model²⁵ — the same mathematical framework that has been used in professional gambling and portfolio management for decades — combined with a Multi-Arm Bandit / Thompson Sampling approach that allows us to simultaneously optimize across hundreds of different combinations of product type, loan term, APR, and customer credit quality.

The practical effect: we can determine not just what product a customer should receive, but what exact product configuration — term, APR, loan size — maximizes both affordability for the customer and risk-adjusted return for the lender. With the addition of three-, six-, and nine-month term products this year, the dimensionality of this optimization problem has grown significantly. The result should be lower first-payment default rates, better portfolio velocity, and higher risk-adjusted returns as the product suite matures over the next 9-12 months of performance data.

In October 2025, I made a decision that many would consider counterintuitive at a time of strong performance: I formally declared, internally, that we should begin managing this business as if our consumer asset class is entering a recession.

My reasoning was simple. If I was wrong, the cost of extra conservatism is minimal — we just make slightly less money. If I was right and I hadn't prepared, the cost could be severe. This asymmetric risk calculation is how I think about risk management in all weather. One of my favorite books on this subject is Nassim Taleb's Antifragile, which describes systems that not only resist shocks but actively get stronger from them.²⁶ The design principle we've built into Hive — matching liability duration to asset duration, conservative underwriting even when demand is extraordinary, building shorter-duration products to accelerate capital velocity — is the anti-fragile pattern applied to credit.

The numbers support the approach. Our projections for 2026 show $41MM+ in deployment capacity, up from $33MM in the prior period, while the new three-, six-, and nine-month products will increase capital velocity and allow us to serve a broader range of customers with appropriate risk-adjusted pricing. If the market turns worse than we expect, we can simply tighten underwriting. If it turns better, we're positioned to accelerate.

Building a fund management firm is not just about performance. It is about building the institutional infrastructure that allows more capital to reach our strategy — responsibly, with appropriate oversight, and through channels that give investors confidence in what they own.

This past year, Hive was approved on the Bank of New York Mellon Pershing platform — one of the largest custodial and clearing platforms in the financial world, with extensive reach into the registered investment advisor community. We were previously approved by Schwab, whose platform manages approximately $12 trillion in client assets. We are currently in the process of completing the Fidelity approval process.

These approvals are not paperwork milestones. Each involves extensive due diligence, operational review, and regulatory examination. Each one expands the universe of qualified advisors who can allocate client capital to our strategy and opens access to institutional allocators who require these platform relationships as a prerequisite for investment. We are building the plumbing of a durable institution.