Compute-backed lending through CompuX delivers a net portfolio yield of 9-12%. Default rates range between 2% and 5%, supported by blockable credit collateral that enables 70-85% recovery on defaulted positions. For capital partners evaluating this asset class, portfolio analytics provide the foundation for allocation decisions, ongoing monitoring, and performance benchmarking.
Key Takeaways:
- Net portfolio yield of 9-12% — After accounting for defaults, servicing costs, and recovery proceeds, lenders realize returns that exceed most fixed-income alternatives and compete favorably with venture debt.
- IRR of 10-13% over fund life — Internal rate of return depends on deployment speed, default rates, and recovery timing, but consistently outperforms treasury and revenue-based financing benchmarks.
- MOIC of 1.2x-1.4x over 3 years — Multiple on invested capital reflects the compounding effect of the credit multiplier and high recovery rates on the small percentage of defaults.
- 2-5% default rate with 70-85% recovery — Blockable credits enforce collateral seizure in under 500 milliseconds, compared to 3-12 months and 20-40% recovery in traditional unsecured lending.
- Real-time portfolio visibility — Dashboard analytics surface exposure concentration, utilization trends, and early warning signals before defaults occur.
How Returns Are Calculated
Portfolio returns in compute-backed lending are measured using four complementary metrics. Each captures a different dimension of performance, and institutional allocators typically evaluate all four before making commitment decisions.
Internal Rate of Return (IRR)
IRR measures the annualized return on deployed capital, accounting for the timing of cash flows. In a typical CompuX lending portfolio, capital is deployed over 3-6 months as loan originations close, with repayments flowing back over 12-24 month terms. The IRR calculation incorporates the credit multiplier effect: because $1M in capital produces $1.25-1.5M in compute credits, the effective yield on deployed capital exceeds the stated interest rate.
A portfolio with a 10% stated rate, 1.35x average multiplier, and 3% default rate with 80% recovery typically produces an IRR of 11-12%. The range of 10-13% reflects variation in deployment speed (faster deployment improves IRR), default timing (early defaults reduce IRR more than late defaults), and recovery speed.
Multiple on Invested Capital (MOIC)
MOIC is the total value returned divided by total capital invested. For a 3-year fund with $10M committed, a 1.3x MOIC means $13M returned to investors. The MOIC in compute-backed lending ranges from 1.2x to 1.4x over a 3-year fund life, depending on portfolio composition and recycling strategy.
The upper end of this range requires active capital recycling: as early loans are repaid (typically at 12-18 months), the capital is redeployed into new originations rather than held as cash. A fund that recycles capital 1.5x over its life achieves meaningfully higher MOIC than one that deploys capital once and waits.
Cash-on-Cash Yield
Cash-on-cash yield measures the annual income generated relative to invested capital, excluding unrealized gains or losses. This metric matters most for investors who need predictable income streams rather than total return. A typical CompuX lending portfolio generates 8-10% annual cash-on-cash yield from interest payments alone, before accounting for any recovery proceeds from defaults.
Net Portfolio Yield
Net portfolio yield is the headline metric: total return after all costs, defaults, and recoveries. It answers the question fund managers care about most: what did the portfolio actually earn? The 9-12% net yield range accounts for origination costs (1-2% of loan value), servicing costs (0.5-1% annually), defaults (2-5% of portfolio), and recovery proceeds (70-85% of defaulted amounts). This is the number that appears in quarterly LP reports and drives allocation decisions.
Use the interactive calculator to model net yield under your specific assumptions for default rate, recovery, and deployment pace.
Portfolio Composition
A well-constructed compute-backed lending portfolio balances diversification against operational complexity. Based on data from active CompuX capital partners, the typical portfolio has the following characteristics:
| Parameter | Typical Range | Rationale |
|---|---|---|
| Number of loans | 10-20 | Sufficient diversification without excessive monitoring burden |
| Average loan size | $200K-$500K | Large enough for meaningful yield, small enough for granular risk management |
| Facility size | $2M-$50M | Minimum viable for institutional deployment; upper bound reflects current market depth |
| Tenor | 12-24 months | Matches AI startup compute consumption cycles |
| Interest rate | 8-12% APR | Risk-adjusted for asset class; competitive with venture debt |
| Credit multiplier | 1.25x-1.50x | Driven by bulk purchasing discounts from GPU providers |
| Collateral coverage | 100-125% | Blockable credits provide real-time collateral enforcement |
Diversification Dimensions
Effective portfolio construction diversifies across three axes:
Borrower stage. A mix of seed-stage AI startups (higher yield, higher risk) and Series A/B companies (lower yield, lower default probability) produces more stable aggregate returns. Portfolios concentrated in a single stage exhibit higher volatility.
Compute provider. Loans backed by credits across multiple GPU providers (such as AWS, GCP, CoreWeave, Lambda) reduce concentration risk. If a single provider experiences outages or pricing changes, only a fraction of the portfolio is affected.
Use case. Borrowers focused on training-heavy workloads have different risk profiles than those running inference-heavy applications. Training workloads are lumpy and front-loaded; inference workloads generate steadier, more predictable compute consumption. A balanced portfolio includes both.
Default Rates and Recovery
Default rates in compute-backed lending through CompuX range from 2% to 5% of outstanding portfolio value. This is well below the 8-15% default rates typical of unsecured startup lending. The lower default rate reflects two structural advantages. First, borrowers are spending capital on a productive input (compute) rather than discretionary expenses. Second, the blockable credit mechanism creates a powerful incentive against default.
How Blockable Credit Recovery Works
When a borrower misses a payment or triggers a covenant breach, the CompuX platform can freeze the borrower's remaining compute credits in under 500 milliseconds. This is not a legal process or a collections effort. It is an automated enforcement mechanism built into the credit infrastructure. The frozen credits retain their full economic value. Compute demand consistently exceeds supply in the current market, so frozen credits find buyers quickly.
Recovery proceeds from collateral enforcement typically reach 70-85% of the defaulted loan value. The range depends on how much of the original credit allocation the borrower had already consumed at the time of default. Early defaults (within the first 6 months) tend to have higher recovery rates because more credits remain unspent. Late defaults (after 18+ months) have lower recovery because the borrower has consumed most of the allocated compute.
Comparison to Traditional Recovery
| Metric | CompuX Blockable Credits | Traditional Unsecured | Traditional Secured |
|---|---|---|---|
| Recovery rate | 70-85% | 20-40% | 50-70% |
| Time to recovery | Hours to days | 3-12 months | 6-18 months |
| Legal costs | Minimal (automated) | $50K-200K+ | $100K-500K+ |
| Collateral liquidity | High (compute always in demand) | Variable | Asset-dependent |
The speed of recovery is as important as the rate. In traditional lending, recovery is a months-long legal process that ties up capital and generates legal fees. Blockable credit recovery is effectively instantaneous, allowing the lender to redeploy recovered capital into new originations within days.
Risk Modeling
Institutional allocators need rigorous risk modeling before committing capital. The following framework covers the three primary risk areas in compute-backed lending.
Sensitivity Analysis: Default Rate vs. IRR
The table below shows how portfolio IRR changes as default rates increase, assuming a base case of 1.35x multiplier, 10% stated rate, 24-month tenor, and 80% recovery on defaults.
| Default Rate | Net Portfolio Yield | IRR | MOIC (3-Year) |
|---|---|---|---|
| 1% | 11.5% | 12.8% | 1.38x |
| 3% | 10.2% | 11.5% | 1.32x |
| 5% | 8.9% | 10.1% | 1.25x |
| 7% | 7.5% | 8.7% | 1.19x |
| 10% | 5.8% | 6.9% | 1.11x |
Even at a 10% default rate, which is double the upper end of the typical range and well above observed historical performance, the portfolio still generates positive returns that exceed treasury yields. This stress tolerance comes from the combination of the credit multiplier and high recovery rates.
Macro Risks
GPU price deflation. If GPU prices drop sharply, the value of programmable collateral could decline. However, compute demand has historically grown faster than supply, and the shift from training to inference workloads is expanding the addressable market. Partial mitigation comes from the short loan tenors (12-24 months), which limit exposure to multi-year price trends.
AI market contraction. A broad contraction in AI spending would increase default rates as startups lose revenue. The 2-5% default rate assumes continued market growth. In a contraction scenario, default rates could reach 7-10%, which the sensitivity table above shows still produces positive returns.
Provider concentration. If a single GPU provider represents more than 30% of a portfolio's credit collateral, provider-specific events (outages, bankruptcy, pricing changes) create outsized risk. Portfolio guidelines should cap single-provider exposure at 25-30%.
Regulatory Risk
Compute credit financing is a relatively new asset class without established regulatory precedent in most jurisdictions. Lenders should monitor developments in securities classification (whether compute credits could be classified as securities in certain structures), consumer lending regulations (if any borrowers qualify as consumers rather than businesses), and cross-border considerations for international deployments.
Yield Comparison: CompuX vs. Alternatives
Capital partners evaluating compute-backed lending typically compare it against other alternative lending categories and fixed-income allocations. The table below provides an apples-to-apples comparison.
| Asset Class | Net Yield | Default Rate | Recovery Rate | Liquidity | Collateral |
|---|---|---|---|---|---|
| CompuX compute lending | 9-12% | 2-5% | 70-85% | Medium (12-24 mo) | Blockable credits |
| Venture debt | 8-12% | 5-10% | 30-50% | Low (24-48 mo) | Warrants + IP |
| Revenue-based financing | 6-9% | 3-8% | 40-60% | Medium (12-18 mo) | Revenue share |
| Treasury / money market | 4-5% | ~0% | 100% | High (daily) | Government-backed |
| Unsecured startup lending | 12-18% | 10-20% | 20-40% | Low (12-36 mo) | None |
| Equipment financing | 6-8% | 2-4% | 60-80% | Low (24-60 mo) | Physical equipment |
CompuX compute lending occupies a distinct position: it offers venture-debt-level yields with equipment-financing-level default rates and recovery rates that exceed both categories. The primary trade-off is market maturity. This is a newer asset class with a shorter track record than venture debt or equipment financing, which means less historical data for modeling long-tail risk scenarios.
Dashboard and Reporting
CompuX provides capital partners with real-time portfolio analytics through a dedicated lender dashboard. The dashboard is designed for fund managers who need to report to LPs, manage risk exposure, and make rebalancing decisions.
Real-Time Metrics
The dashboard surfaces the following metrics, updated continuously:
- Portfolio utilization rate — Percentage of allocated credits currently consumed by borrowers. High utilization (above 70%) indicates healthy borrower activity; low utilization may signal borrower distress or over-allocation.
- Days past due (DPD) distribution — Breakdown of loans by payment status (current, 1-30 DPD, 31-60 DPD, 60+ DPD). This is the earliest warning signal for potential defaults.
- Provider concentration — Visual breakdown of credit collateral by GPU provider, flagging any single-provider exposure above the 25-30% threshold.
- Weighted average remaining term — The portfolio's average time to maturity, weighted by loan size. This informs liquidity planning and capital recycling decisions.
- Collateral coverage ratio — Real-time ratio of blockable credit value to outstanding loan balances. A ratio below 100% triggers automatic alerts.
Monthly Reports
Monthly LP reports include portfolio yield (net of defaults and costs), origination and repayment activity, default and recovery summary, and a portfolio health score. The portfolio health score is a composite metric (0-100) that weights utilization, DPD distribution, diversification, and collateral coverage. Scores above 80 indicate a healthy portfolio; scores below 60 warrant active intervention.
Integration
The dashboard exposes an API for integration with existing portfolio management systems. Fund managers who use tools like Carta, Juniper Square, or custom LP reporting platforms can pull CompuX portfolio data directly into their workflows.
Benchmarking
Performance benchmarking for compute-backed lending is still maturing as the asset class develops. However, several reference points provide useful context for evaluating portfolio performance.
Cambridge Associates alternative lending benchmarks. The Cambridge Associates Private Credit Index tracks returns across direct lending, mezzanine, and specialty finance. Compute-backed lending's 10-13% IRR places it in the top quartile of specialty finance strategies, comparable to top-performing direct lending funds without the 5-7 year lockup periods typical of those vehicles.
Venture debt industry data. According to industry reports, venture debt funds targeting early-stage companies typically produce net IRRs of 8-12% with default rates of 5-10%. CompuX compute lending achieves comparable or superior returns with lower default rates and faster recovery, driven by the structural advantages of blockable credit collateral.
Equipment financing comparables. Equipment-backed lending, the closest structural analog to compute-backed lending, produces net yields of 6-8% with default rates of 2-4%. Compute lending generates higher yields because the underlying asset (compute) is appreciating in demand value rather than depreciating like physical equipment.
The most important benchmarking distinction is recovery speed. In traditional alternative lending, recovery is measured in months or years. In compute-backed lending, recovery through blockable credits is measured in hours. This difference compounds over time: faster recovery means faster capital recycling, which means higher MOIC for the same amount of committed capital.
For a deeper understanding of how the collateral mechanism works, see the blockable credit collateral guide. For modeling how the multiplier affects lender economics, use the credit multiplier mechanics overview and the interactive calculator.
Frequently Asked Questions
What is the minimum facility size for a capital partner?
The minimum facility size is $2M, which allows construction of a portfolio with 8-10 loans averaging $200K-$250K each. This provides sufficient diversification to smooth out individual borrower outcomes. Facilities above $10M unlock additional benefits including priority deal flow, custom reporting, and dedicated portfolio management support. The maximum facility size is $50M, reflecting current market depth for compute-backed lending originations.
How does blockable credit collateral compare to traditional loan collateral?
Blockable credits are programmatically enforceable in under 500 milliseconds, compared to weeks or months for traditional collateral seizure. The recovery rate of 70-85% exceeds traditional unsecured recovery (20-40%) and is comparable to or better than secured lending recovery (50-70%). The key structural advantage is that blockable credits do not require legal proceedings, court orders, or physical asset seizure. The enforcement mechanism is built into the credit infrastructure itself, eliminating the legal costs and time delays that erode recovery value in traditional lending.
What happens to portfolio returns if AI market growth slows?
The sensitivity analysis above models this scenario. If default rates double from the typical 3% to 6-7%, net portfolio yield drops from approximately 10% to approximately 7.5%, and IRR falls from approximately 11.5% to approximately 8.7%. The portfolio still generates positive returns above treasury yields even under stress conditions, because the credit multiplier and blockable credit recovery provide structural downside protection. A severe contraction scenario (10%+ defaults) would reduce returns to the 5-7% range, which is still positive but below the target return for most alternative lending allocators.
How are lender returns affected by the credit multiplier?
The credit multiplier directly improves lender economics. When $1M in capital produces $1.35M in compute credits, the borrower receives more compute value than the loan principal. This means borrowers have stronger incentive to repay (they received outsized value) and lenders have more collateral coverage (the credits are worth more than the loan). The multiplier effect increases the collateral coverage ratio above 100%, providing a buffer that protects lender returns even when individual loans default.
Can I integrate CompuX portfolio data with my existing fund management tools?
Yes. The CompuX lender dashboard provides an API for integration with portfolio management platforms including Carta, Juniper Square, and custom LP reporting systems. Data exports are available in standard formats (CSV, JSON, PDF) for manual integration. Monthly LP reports follow industry-standard formatting conventions used by alternative lending funds, making them compatible with existing investor communication workflows. Contact the CompuX capital partnerships team to set up API access and discuss custom reporting requirements.