GUY: Good morning, Ava. It is Monday, July 27, 2026, and today’s Morning Signal has a pretty clean macro tension running through it: private-sector balance sheets may be handling the rate shock better than the headlines suggest, while the more dangerous leverage problem may be migrating onto government balance sheets.

AVA: Good morning. And the technology version of that tension is just as interesting. Capital is already committed to AI infrastructure, but nobody has fully settled the utilization, return-on-capital, or regulatory questions. So today is really about separating what is improving now from what could break later.

GUY: Let’s start with Monetary Matters. Jack Farley interviewed Nicholas Brooks, the head of economic and investment research at ICG, and Brooks brought data from roughly four hundred to five hundred private companies. His latest medians showed EBITDA growth around eight percent in Europe and six percent in the United States.

AVA: Which is a useful correction to the idea that the whole private-company universe is already rolling over. But Monetary Matters also made clear that medians are not the same thing as safety. The aggregate can improve while the weakest borrowers remain trapped below viable coverage, especially when leverage, refinancing timing, and sponsor behavior differ company by company.

GUY: Exactly. On Monetary Matters, Brooks said median interest coverage in Europe fell from roughly three times to two and a half times during the rate shock. In the United States, it fell from about two and a half times to two times. Europe has stabilized, and the U.S. has started to improve as policy rates ease.

AVA: So the mechanism is straightforward. For floating-rate borrowers, lower policy rates reduce the interest bill faster than a modest slowdown in EBITDA necessarily damages the numerator. But that does not mean every credit is healing. It means the broad debt-service arithmetic is becoming less hostile.

GUY: Right. And Monetary Matters gave us the proper trade construction. This is not a reason to declare private credit safe. It is a reason to prefer dispersion trades over a blanket systemic short. Look for weak cash coverage, aggressive EBITDA adjustments, recurring payment-in-kind interest, revolver draws, near-term refinancing needs, or a bad position in the capital stack.

AVA: Hold on though, because software is where the narrative could still outrun the reported numbers. On Monetary Matters, Brooks estimated software at about fourteen percent of ICG’s European sample and twelve percent of its U.S. sample. He said broad EBITDA pressure from AI disruption is not yet visible in the software borrowers he tracks.

GUY: That matters because people sometimes cite software exposures of twenty to thirty percent for the more software-heavy 2020 and 2021 vintages. Farley noted that classification differences also muddy the comparison. Software, IT services, and healthcare technology do not always land in the same bucket.

AVA: The signal to watch is a divergence. Monetary Matters showed why stable adjusted EBITDA is not enough. If cash interest coverage weakens anyway, or revolver draws rise, the borrower may be consuming liquidity while the headline earnings measure still looks fine. That would tell us the adjustment stack is masking stress.

GUY: Now here is the bigger shift. On Monetary Matters, Brooks argued that the more consequential medium-term leverage risk sits with sovereigns. Governments absorbed debt through the global financial crisis, the pandemic, and aging-related spending. He cited U.S. fiscal deficits in roughly a six-to-eight-percent-of-GDP range and saw no obvious consolidation path.

AVA: And the transmission channel is the long end, not simply the next central-bank decision. Monetary Matters laid out a reflexive chain: higher sovereign yields lift the benchmark cost of financing, higher interest expense damages fiscal credibility, and weakened credibility can push yields higher again until policy makers are forced to respond.

GUY: That is the scenario where the normal hedge can fail. If government bonds sell off during risk stress instead of rallying, duration stops cushioning the portfolio. The confidence trigger matters more than any magic debt threshold.

AVA: Monetary Matters used the United Kingdom’s 2022 mini-budget as the template. A policy announcement changed investors’ belief about fiscal control, long-end yields moved quickly, technical vulnerabilities amplified the move, and policy makers reversed course. The United States has deeper natural demand for government paper, but repeated bond selloffs during stress would still be a warning.

GUY: So Wednesday, July 29, matters. Farley referenced the Federal Reserve decision, and the curve response is more informative than the policy rate alone. If front-end yields fall while the long end rises, that supports the idea that easier monetary policy can repair private coverage without restoring confidence in sovereign duration.

AVA: I like that framing because it keeps two time horizons separate. Private borrowers can benefit now from lower floating-rate costs. Meanwhile, fiscal term premium can stay elevated or even rise. Easing at the front end is not automatically easing for every borrower if the long end and credit spreads refuse to cooperate.

GUY: Let’s bring in Capital Allocators. Ted Seides interviewed NYU chief investment officer Michelle Knudsen about rebuilding an endowment that grew from about six and a half billion dollars when she arrived in 2024 to roughly eight billion dollars today.

AVA: On Capital Allocators, Knudsen described a portfolio targeting about sixty-five percent in public and private equity, about twenty-five percent in absolute-return and opportunistic strategies, and smaller allocations to cash, fixed income, and real assets. NYU has turned over more than one-third of the portfolio and re-underwritten roughly another third in two years.

GUY: The important point is that diversification has to earn its place. Capital Allocators said NYU still wants equity-like long-run returns, but a non-equity strategy needs a clear reason to exist, a plausible source of persistence, and a defined portfolio role.

AVA: And risk is evaluated before commitment, not after the drawdown. On Capital Allocators, Knudsen emphasized correlated drawdowns, gap risk, liquidity needs, and multi-year shocks. That is especially relevant as NYU adds trading-oriented and higher-leverage hedge-fund strategies, macro and relative value, venture, emerging managers, and lower-middle-market buyouts.

GUY: What I liked is the governance design. Capital Allocators said the investment committee sets direction, risk appetite, and portfolio frameworks, while staff selects managers until an allocation approaches three percent of the endowment. At that point, the committee approves the concentration of risk.

AVA: Even better, the main internal challenge session happens around seventy percent through diligence. The team has enough information to form a view, but the recommendation is not yet frozen. That is when objections can improve the work instead of merely ratifying it.

GUY: There is also a warning for today’s private-market environment. On Capital Allocators, Knudsen flagged aggressive fundraising and fear of missing out. Access to a famous manager does not prove the vintage is attractive, and a benign aggregate credit backdrop does not justify rushing commitments.

AVA: Exactly. Preserve liquidity as an option. Capital Allocators showed that commitment pacing should be tied to distributions, portfolio stress, and genuine diversification. If new strategies merely add correlated leverage, then the portfolio looks broader while becoming more fragile.

GUY: Now connect Capital Allocators back to Monetary Matters. Brooks says broad private-company coverage is improving. Knudsen still sizes leveraged strategies against correlations going toward one. Those are not conflicting views. Improving medians help you avoid the wrong systemic short; they do not eliminate the need for liquidity buffers.

AVA: And Capital Allocators also offered a practical AI use case inside institutional investing. NYU is using AI tools to turn raw notes into shared formats, capture discussions, and organize recurring information from financial statements across funds.

GUY: Which is much more believable than an autonomous machine picking managers.

AVA: Exactly. On Capital Allocators, the promising project is consistent time-series extraction from audited statements. The system can flag changes in fees, leverage, cash conversion, exposures, or valuation policy. Historical data generates questions; the investment team still owns the forward-looking judgment.

GUY: That human ownership brings us to the a16z Podcast. Steven Sinofsky argued that broad AI regulation is arriving before the technology, market structure, and failure modes are stable enough to define well.

AVA: On the a16z Podcast, Sinofsky’s preferred sequence was to enforce existing laws against harmful conduct, amend sector-specific statutes where AI creates a demonstrated gap, and keep the licensed human or institution accountable for decisions. An AI system cannot hold a medical, legal, or optician’s license.

GUY: So if a clinician uses an AI tool to prepare notes and harms a patient, the clinician remains responsible. The regulator can examine whether existing law clearly covers AI-assisted conduct before creating an all-purpose permissioning regime.

AVA: But the a16z Podcast also gave the falsification condition for that position. The anti-regulatory thesis weakens if identifiable AI-specific harms repeatedly escape existing law, if accountability cannot be assigned to deployers, or if open-weight distribution creates security externalities that cannot be controlled at the use layer.

GUY: That is the right discipline. Do not argue “regulation good” or “regulation bad.” Ask what object is being regulated: model weights, chips, compute, deployment, professional use, procurement, or harmful output.

AVA: On the a16z Podcast, Sinofsky also treated support for sweeping regulation from large model companies skeptically. His concern was regulatory capture: incumbents can turn their preferred technical assumptions into compliance barriers that smaller entrants and open-model ecosystems cannot absorb.

GUY: And his open-model point was grounded in how research historically works. The a16z Podcast argued that academic and government-funded research has depended on published methods and open software. Restricting open models because established labs fear competition could protect market structure without necessarily improving public safety.

AVA: The investable implication is a wide distribution of outcomes. Licensed vertical applications may get clearer accountability while retaining room to innovate. Frontier labs may pursue compliance moats. Open-model ecosystems remain exposed to export controls, procurement restrictions, and security rules even if a comprehensive domestic licensing regime stalls.

GUY: Now, on the a16z Podcast, Sinofsky put the U.S.-China relationship in the frame of innovation leadership. Governments use export controls, procurement, subsidies, national champions, and restrictions on adjacent inputs because those are the tools they have.

AVA: But indirect constraints can change cost and location without restoring competitiveness. The a16z Podcast used the U.S. auto industry’s response to Japanese competition as the cautionary example. Trade pressure shifted Japanese manufacturing into the American South, but it did not restore Detroit’s lost product advantage.

GUY: Applied to AI, chip controls can slow training or raise costs without determining model quality. Restrictions on Chinese open models can also shelter U.S. incumbents. Industrial policy can redirect supply chains and capital while leaving the domestic capability gap unsolved.

AVA: Monetary Matters added the European side of that problem. Brooks argued that Europe has strong researchers, universities, and entrepreneurs but lacks a unified financing market at U.S. scale. Companies often reach a point where deeper later-stage capital pulls them toward U.S. listings, financing, or acquirers.

GUY: So Europe can create technical value while the value accrues elsewhere. Monetary Matters said European defense, energy-security, and infrastructure spending, especially in Germany, provides a fiscal impulse. But financing fragmentation and slower capital formation still constrain technology scale.

AVA: Let’s stay with AI capex. On Monetary Matters, Brooks said already committed AI-infrastructure spending should support U.S. growth over the next several quarters and probably beyond. Korea, Taiwan, semiconductor supply chains, power infrastructure, and hyperscaler-linked industries are the nearer-term beneficiaries.

GUY: That is a factual growth floor, not a perpetual guarantee. Monetary Matters also left the later utilization test unresolved. Suppliers can benefit while infrastructure is being built even if application revenue, pricing, and return on invested capital eventually fail to justify the installed base.

AVA: In other words, near-term macro optimism can coexist with a 2027 or 2028 capex-digestion risk. The exit multiple should reflect that timing gap. Watch project deferrals, power constraints, utilization, depreciation, and whether application revenue catches up before the committed pipeline is completed.

GUY: And for software credit, the burden of proof cuts both ways. Monetary Matters said ICG’s current data do not show broad AI-driven EBITDA damage. The market may be discounting future disruption before it appears in the median.

AVA: The next software reporting cycle is therefore critical. From the Monetary Matters discussion, we should track retention, seat contraction, pricing, cash conversion, and AI-driven displacement. If those weaken together, future disruption becomes operating evidence rather than a narrative.

GUY: Time for the wonderfully concrete deployment lesson. On The Indicator from Planet Money, Wailin Wong, Darian Woods, and Stephen M. looked at restaurant-payment frustrations, including handheld terminals, all-in menu pricing, and airport loyalty programs.

AVA: On The Indicator, handheld payment terminals reduced repeated server trips, kept the customer’s card in view, simplified split bills, and could improve table turnover. A Toast representative said more than one hundred seventy thousand U.S. locations were using its handhelds as of March.

GUY: But devices can cost hundreds of dollars, and workflow inertia is real. A product can be useful and still spread slowly when the buyer pays the cost while the benefit is split across labor productivity, customer convenience, and payment security.

AVA: The Indicator’s airport loyalty example was even better. Airport restaurants are often run by concession companies operating multiple brands through their own point-of-sale systems. Those systems may not connect to each brand’s rewards platform, and individual franchisees can control participation.

GUY: So the customer sees a missing loyalty credit, but the economic cause is fragmented ownership and incompatible infrastructure. The moat sits at the system-of-record and workflow layer, where data, accountability, and user behavior meet.

AVA: The Indicator also described all-in menu pricing as a collective-action problem. A restaurant that includes tax can appear more expensive than competitors displaying pre-tax prices, while state, county, and city tax fragmentation raises the cost of implementation.

GUY: That connects nicely to the a16z Podcast. The Indicator showed a case where a common rule might solve a coordination problem. Sinofsky warned that a common rule imposed too early can freeze the market around an incumbent AI architecture. The policy question is whether coordination fixes a documented problem or entrenches the wrong system.

AVA: Let’s pull the cross-currents together. First, Monetary Matters showed lower policy rates improving private coverage, but also warned that fiscal deficits can keep sovereign term premia elevated. Risk can migrate from floating-rate company debt to long-duration government paper.

GUY: Second, Monetary Matters showed committed AI capex providing a near-term growth floor. Capital Allocators showed NYU increasing innovation exposure while using AI for workflow memory. The a16z Podcast showed that the market structure and policy endpoint are still unsettled. Capital is committed before demand, regulation, and returns are fully proven.

AVA: Third, the a16z Podcast and Capital Allocators shared an incumbent problem. Large AI labs can use regulation to reduce open competition. Established private-market franchises can use brand and access to attract commitments despite weak vintage economics. In both cases, institutional comfort can be mistaken for edge.

GUY: Fourth, Capital Allocators and The Indicator both said integration beats invention. NYU’s useful AI work turns scattered notes and statements into a repeatable investment workflow. Toast’s product works best when the payment, loyalty, ownership, and operating systems actually connect.

AVA: And fifth, Monetary Matters and Capital Allocators reminded us that aggregate resilience can conceal left-tail fragility. Healthy median EBITDA and improving coverage can coexist with a leveraged fund, bad vintage, or weak borrower that fails abruptly. Do not confuse avoiding a systemic short with relaxing the stress test.

GUY: Let’s finish with the dashboard. On Wednesday, July 29, watch the Federal Reserve decision and the full yield curve. Falling front-end yields with a rising long end would strengthen the sovereign-term-premium thesis.

AVA: From Monetary Matters, watch the next ICG private-company update for EBITDA growth, cash interest coverage, leverage, non-accruals, payment-in-kind usage, and the share of borrowers below one-times coverage.

GUY: Also from Monetary Matters, watch private-credit fundraising and redemption data. Wealth-channel flows, gates, asset sales, and business-development-company discounts will tell us whether improving borrower medians are translating into real liquidity and realizations.

AVA: From the combined Monetary Matters and Capital Allocators discussions, watch hyperscaler capex guidance for timing, power availability, depreciation, utilization, and application revenue. The near-term floor holds only while committed spending becomes deployed capacity.

GUY: From the a16z Podcast, track U.S. and European proposals in the second half of 2026 on open-weight models, chip exports, model imports, procurement, and licensed professional use. Identify the exact regulated object and compare the burden on incumbents with the burden on entrants.

AVA: From Capital Allocators, watch NYU’s next pacing cycle: private commitments versus distributions, absolute-return leverage and liquidity, and whether new venture and buyout exposure truly diversify a concentrated public-equity book.

GUY: And from The Indicator, watch restaurant-technology disclosures: Toast location growth, hardware economics, gross payment volume, and evidence that handheld adoption actually improves labor productivity and retention enough to repay the installation cost.

AVA: The bottom line for Monday is disciplined separation. Private-company medians are improving, but the left tail still matters. AI infrastructure supports growth now, but utilization and policy remain open. Workflow integration is producing real value, but institutional ownership and accountability determine who captures it.

GUY: And the biggest macro warning is that leverage may not disappear when private coverage heals. It can move onto the sovereign balance sheet and return through long-end yields, currency confidence, and financing costs.

AVA: That is your Morning Signal for Monday, July 27. Keep the curve, the cash coverage, and the workflow layer on the same screen.

GUY: We’ll be back tomorrow. Have a good one.