Vault report · Upshift · Ethereum → Solana

upSBRWA: levered carry on tokenised real-world assets

The SharpByte RWA Ecosystem Vault takes USDC on Ethereum, forwards it to an operator wallet, burns it to Circle's CCTP and reissues it on Solana, where the strategy actually runs. It buys tokenised real-world assets, borrows against them in isolated Kamino markets, and keeps the difference between what the collateral earns and what the borrowing costs.

The vault is denominated on a chain it holds almost nothing on. A dashboard reading only the Ethereum side sees an empty vault; one reading only Upshift's API inherits a full day of lag. This report reconstructs the position from chain data, sets it beside the reported figure, and shows the gap instead of reconciling it away.

Implied AUM
$1,518,070
Upshift's last daily stamp, carried forward by the net flow observed since. Reconciled in exhibit 1.
Composition
51.2% ONyc · 45.1% AUTO · 3.8% idle
By gross exposure, at this chain read.
Gross exposure
2.49× net equity
Idle cash counts on both sides of the ratio.
Blended LTV / min buffer
62.1% / 12.2%
Buffer is the tightest position, not the average. Every obligation currently carries a stale oracle price (exhibit 5).
Return
15.83% annualised on the last 30d · 11 weeks since inception
Net of the performance fee.
Fees
10% performance (weekly on gains) · 0.25% management
Redemption
Batch on request · 2% liquidity reserve · no fixed window
Inception
2026-07-31
Ethereum · SolanaVault 0x5776B5…9F3A26Chain read just nowWritten with AI assistanceWeekly reports →

Upshift last stamped the vault at $1,518,070, 13h ago. Since that stamp, this page has observed $0 of deposits and $0 of redemptions in the vault's Ethereum transfer logs, a net flow of $0 that the published figure does not yet include. Implied AUM is therefore $1,518,070.

Separately, this page identifies $1,102,375 of net equity by valuing positions directly on Solana. That figure is a floor, not a competing NAV: it counts only what a free indexer will decode, and the vault demonstrably uses venues that none will.

Exhibit 1:Vault valuation, reported versus reconstructed

MeasureUSD% of implied AUM
Today
Reported TVL$1,518,070100.0%
Deposits since stamp$00.0%
Redemptions since stamp$00.0%
Implied AUM$1,518,070100.0%
Identified net equity on chain$1,102,37572.6%
Not independently priced$415,69527.4%
Since inception
Lifetime subscriptions$1,496,076
Lifetime redemptions$1,925
Lifetime deployments$1,391,000
Reconstructed figures read from chain at 13:05 UTC, 13 Oct 2026. Reported TVL is Upshift's most recent daily stamp, 00:01 UTC, 13 Oct 2026. Flow figures are the vault's Ethereum transfer logs. Redemptions are shown negative because the group is a build-up to implied AUM. The gap between implied AUM and identified net equity is the part of the book this page cannot value, not a claimed loss. Source: Upshift API, Blockscout, Kamino API, Solana RPC, SharpByte Labs.

implied AUM = reported TVL + (deposits − redemptions) since the stamp

The obvious version (reported plus deposits) overstates AUM on any day somebody redeems, so net flow is used instead.

The reporting lag

Upshift recomputes vault metrics once a day, at roughly 00:01 UTC. Between snapshots its reported TVL (the number on the public vault page) can be arbitrarily wrong, and wrong in a specific direction: it lags every deposit and redemption made since the last stamp.

The magnitude is not theoretical. On 2026-08-19 a $436,161 deposit landed at 04:55 UTC; the reported $250,576 understated real AUM by roughly 63% for the rest of that day. Any figure derived from reported TVL (a share of portfolio, an APY denominator, a fee accrual) inherits the same error.

This page reads the chain every five minutes. That does not make it authoritative over Upshift, which is the accounting record; it makes it a check on the one window where the accounting record is known to be stale. The current snapshot is 13h old.

The strategy

The vault buys tokenised real-world assets, supplies them as collateral in isolated Solana lending markets, borrows stablecoins against them, and recycles the proceeds. The return is the spread between what the collateral yields intrinsically and what the borrowing costs, multiplied by the leverage applied to it. It currently runs at 2.49× gross exposure to net equity, at a blended loan-to-value of 62.1%.

Three conditions gate a deployment. They are stated as conditions, not features, because each one can stop being true.

Exhibit 2:Strategy gating conditions and current status

ConditionWhy it gatesToday
Collateral qualityOnly tokenised assets with a published redemption path, supplied into isolated markets, so a failure at one venue cannot reach the other.2 isolated markets
Borrow cost under collateral yieldLeverage amplifies a thin spread. The trade works only while the cost of borrowing stays below what the collateral earns in its own exchange rate.2.49× at 62.1% LTV
Liquidation bufferPositions are sized by how far collateral can fall before liquidation, not by the yield they quote.12.2%
Read from chain at 13:05 UTC, 13 Oct 2026. Conditions and thresholds are SharpByte policy, not venue or Upshift requirements. The figures beside them are read live. Source: Kamino API, Solana RPC, SharpByte Labs.

Size over time, and what a depositor earned

These are two different questions and they are answered separately here, on purpose. Total assets move when somebody else deposits; plotting size beside share price on one canvas invites reading a subscription as performance.

Realised return, as the 30-day compound figure Upshift shows on its vault page, is 15.83%. Maximum drawdown since inception is 0.04%. Every return figure quoted here is already net of the performance fee, which is deducted before the share price is published.

SharpByte also sends LPs a report each week, struck at the Monday close. Those weekly reports are published here as sent.

Exhibit 3:Total assets by day since inception

Latest daily snapshot, 13 Oct 2026. 124 snapshots across 75 days since inception; most land at roughly 00:01 UTC and the surplus are intraday recomputes. Steps are subscriptions and redemptions, not gains. Source: Upshift API, SharpByte Labs.

Where the capital sits

$2,742,230 of gross exposure is located across the venues below, against $1,102,375 of net equity and $1,639,855 borrowed. Of that, $103,211 has arrived but is not yet working: $103,151 still USDC in the vault contract on Ethereum and $60.11 at the Solana operator wallet FqYuNG…zGPk, of which 0.5236 SOL is held for transaction fees.

Exhibit 4:Venue allocation and net equity by position

Gross exposure by venue, before netting borrowings

  • Kamino · OnRe$1.40M51.2%
  • Kamino · AUTO$1.24M45.1%
  • Idle / unallocated$103.2K3.8%
Capital located$2.74M

Gross, net equity and share by venue

VenueGrossNetShareValuation
Kamino · OnRe$1,403,605$629,16751.2%valued
Kamino · AUTO$1,235,414$369,99845.1%valued
Idle / unallocated$103,211$103,2113.8%valued

Idle balances at the Solana operator wallet FqYuNG…zGPk

HoldingAmountValue
PYUSD2b1kV6…4GXo4.4723$4.47
PST59obFN…ZAGw0.0760$0.09
PT-ONyc2W5zZc…vhiJ0.0000$0.00
wONycG1qbuP…HosF25.9785not priced
ONycGVHsb3…uXnr7,418.1864not priced
SOL, held for fees0.5236$55.55
Read from chain at 13:05 UTC, 13 Oct 2026. Gross exposure includes borrowed capital. A venue marked not valued, and a holding marked not priced, is one this page can see but cannot put a number on; both are excluded from the totals rather than estimated. Idle cash is drawn neutral in the ring because it is the absence of a venue, not another one. Idle holdings carry their mint, because a symbol is not an identity on Solana. Source: Kamino API, Solana RPC, Jupiter, SharpByte Labs.

What each venue is

Curator commentary, not a chain read. The sizes in exhibit 4 are read live; what follows is a view, and is labelled as one.

  • Kamino · OnRe. ONyc, OnRe's yield-bearing token, held as collateral in its own isolated Kamino market. Levered against borrowed stablecoins to amplify ONyc's native yield. Isolated, so a problem here stays here.
  • Kamino · AUTO. Hastra's AUTO, held as collateral in its own isolated Kamino market. Same levered structure as OnRe, at a smaller size.
  • Idle / unallocated. USDC and other liquid balances in the operator wallets and the vault's reserve. Not deployed and not at risk. Includes the liquidity buffer held to serve withdrawals.

Leverage and liquidation

A levered loop carries one dominant risk: if collateral reprices far enough for loan-to-value to cross the liquidation threshold, the position is closed at market and the loss is realised. The drawdown buffer measures the distance to that point. That buffer, not the yield, is the number worth watching.

Exhibit 5:Position leverage, liquidation thresholds and health

PositionCollateralBorrowedLTVLiq. LTVBufferHealth
Kamino · AUTO #2 †$563,192$395,40870.2%80.0%12.2%1.14
Kamino · AUTO #1 †$672,223$470,00869.9%80.0%12.6%1.14
Kamino · OnRe #1 †$1,335,894$737,44255.2%75.0%26.4%1.36
Kamino · OnRe #2 †$67,710$36,99654.6%75.0%27.1%1.37
Read from chain at 13:05 UTC, 13 Oct 2026. Rows marked with a dagger carry stale oracle prices, flagged rather than corrected. Buffer is 1 − LTV ÷ liquidation LTV; health is unhealthyBorrowValue ÷ borrowedValue, liquidatable below 1. Source: Kamino API, SharpByte Labs.

buffer = 1 − LTV / liquidation LTV

health = unhealthyBorrowValue / borrowedValue

carry = (Σ supply·APY − Σ borrow·APY) / equity

Kamino publishes obligation values as fractions scaled by 260; the derivations above are computed from those directly, not from a dashboard summary.

Blended rate carry currently reads -11.45%. That figure is Kamino's supply rate less its borrow rate and excludes the yield the collateral earns inside its own exchange rate, so it measures the cost of the leverage, not the return of the trade. A working carry position reads negative here.

Exhibit 6:Net equity against gross exposure

  • Gross exposure
  • Net equity
Latest daily capture, 13 Oct 2026. 38 daily captures to date; the gap between the lines is borrowed capital. Upshift publishes vault-level snapshots only, so leverage and venue allocation are captured locally once a day. That capture reads Kamino's cached obligation totals rather than pricing each leg from its live amount, as the rest of this page does; the cached borrow runs below the live one, so this series carries more equity and less leverage than exhibits 4 and 5 show for the same day. The difference is named here rather than smoothed away. Source: Kamino API, SharpByte Labs.

Fees and terms of participation

What it costs, how to get in, and how to get out. Fee figures are read from Upshift's API, not copied from marketing material, so they reflect what the platform is charging at this read.

Exhibit 7:Fees and terms of participation

TermDetail
Performance feepublished10%, charged weekly on gains and deducted before the share price is published
Management feepublished0.25% annual on total assets
Deposit assetpublishedUSDC on Ethereum · 0xa0b86991c6218b36c1d19d4a2e9eb0ce3606eb48
Receipt tokenpublishedupSBRWA
Minimum depositpublishedNone published
RedemptionspublishedupSBRWA is redeemed for USDC by withdrawal request, settled in batches as the operator recalls capital from Solana. A 2% liquidity reserve is targeted so routine withdrawals do not force an unwind. No fixed settlement window is published.
EligibilitypublishedNon-US persons, 18 and over. Upshift's terms exclude persons located or resident in the United States, and sanctioned persons.
PointspublishedUpshift Points 5× · OnRe Points 1×
Vault contracton-chain0x5776B576474ad2513E550190C5855613859F3A26
EVM operatoron-chain0x3c6061efEF36B972887cF3bD9838045e9d967386
Solana operatoron-chainFqYuNGHZckycWFTN86fGrm3E6U3ohYPcXecrfppnzGPk
Inceptionpublished2026-07-31
Read from the Upshift API at 13:05 UTC, 13 Oct 2026. Terms not published by Upshift are shown as "none published". Operator addresses are published by Upshift under the vault's operators field; they are not inferred. Source: Upshift API, Upshift terms of use.

Curator

SharpByte Labs designs and runs the vault. It sets the allocation policy this report describes and takes the performance fee set out above.

No performance claims are made here on the curator's behalf. The argument this report makes everywhere else is verify it, not trust us, and a curator section that broke that rule would undercut the rest of the page.

What this report cannot tell you

Stated up front because it changes how the totals above should be read. A number that omits a position is worse than a number with a declared blind spot beside it. Last chain read just now, last operator transaction 2d ago.

  • Net equity is a floor. Kamino positions are priced leg by leg from live amounts, not from the obligation's cached USD fields, which run below the real debt. Positions on venues no free indexer decodes (Exponent and Loopscale, when the wallet uses them) are not valued here: Jupiter's portfolio interface does decode them, but the endpoint behind it refuses a keyless caller, so this page cannot fetch and reproduce the figure.
  • Reported TVL is a day old at worst. Upshift recomputes once a day at roughly 00:01 UTC. Anything that cannot be explained by that lag is shown in exhibit 1 as capital this page cannot independently price, instead of being absorbed into a total.
  • Stale oracle prices are flagged, not corrected. Kamino sometimes marks an obligation's prices as stale. Those rows carry a dagger in exhibit 5 and are left as reported, not substituted.
  • Backing and assessment are curator commentary. Sizes and rates in exhibit 4 are read live. The qualitative notes beside it are the curator's own view and are labelled as such wherever they appear.
  • Everything here is keyless and public. Upshift's API, Kamino's API, a public Solana RPC, Blockscout and Jupiter. Public endpoints only, which is what makes the figures reproducible, and what bounds what they can cover.
  • Programs seen in recent activity. Kamino Lend, Kamino Farms, Jupiter, Kamino Vaults, Circle CCTP, Loopscale, Orca Whirlpool, Raydium CLMM. Anything here that is missing from exhibit 4 is capital this page is not counting.

Appendix: collateral memos

SharpByte's own diligence on each collateral asset, reproduced in full. Each memo states the facts of the asset, then why the position is held, how it works, what would have to go wrong, and how it is sized. This is curator research, not chain data: where a figure below is a market estimate, not a contract term, it says so, and none of it is independently verified by this page.

ONyc, OnRe tokenised reinsurance

A NAV-accreting Solana token giving onchain exposure to real-world reinsurance yield, issued by a Bermuda-regulated reinsurer. Value accrues through NAV appreciation, not a $1 peg.

Issuer
On Re SAC Ltd, Bermuda segregated accounts company (reg. 55347)
Regulator
Bermuda Monetary Authority, Class IIGB (Insurance Act 1978) and Class F (Digital Asset Business Act 2018)
Backing
Active reinsurance contracts, liquidity reserves, and diversified reserve assets (T-Bills, sUSDe, syrupUSDC, USCC, USDC, USDG, sUSDS, USYC)
Yield
~9–15% blended market estimate, ~12% average, a mechanism, not a fixed APY
Attestation
Apex Group, monthly, NAV, treasury and reconciliation
Liquidity
Orca pools around $5M each against a Kamino ONYC market above $150M
Backers
Forward Industries (NASDAQ: FWDI, up to $25M) and RockawayX

Why we hold it

  • The yield comes from underwriting, not from crypto. Reinsurance premiums are priced against storms, claims and the underwriting cycle. None of that runs through token prices, funding rates or DeFi flows. That is unusual for an asset paying double digits onchain, and it is the reason ONyc is the core of the book rather than a satellite.
  • The mark is stable enough to lever. NAV has accreted steadily since inception with a maximum drawdown of 0.04%. Premiums earn continuously while losses arrive in clusters, so most of the time the price barely moves. Leverage on an asset like this multiplies the spread instead of the volatility, which is what makes the carry trade work.
  • The offchain claim is something we can underwrite. On Re SAC Ltd holds BMA Class IIGB and Class F licences. The collateral sits in a segregated account ring fenced under Bermuda's Segregated Accounts Companies Act, and Apex Group attests to NAV every month. Most offchain RWA gives you a promise. This gives you a licensed insurer, a legal firewall and an independent check on a fixed schedule.

How it works

  • Qualified investors deposit yield-bearing stablecoins into the Bermuda-regulated pool and mint ONyc.
  • OnRe's actuarial team underwrites short-duration reinsurance, property catastrophe plus specialty lines (D&O, cyber, tech E&O), sourced through global brokers.
  • Premiums net of claims, plus the yield on the reserve assets, accrue to ONyc's NAV. The Solana program controls mint and redemption; the collateral itself sits off-chain in the segregated account.

What the risk is

  • Catastrophe tail risk: a major insured event draws down collateral and cuts NAV. It is the return engine and the tail risk at once.
  • The backing is legally enforced, not code enforced. It sits in a Bermuda segregated account, ring-fenced under the SAC Act, with multi-sig custody, recourse in a failure is a Bermuda legal process, not an onchain claim.
  • NAV is independently verified only monthly, so the onchain price can lag a real claims event between attestations.
  • Secondary liquidity is thin relative to the lending markets built on the token; primary redemption is gated to qualified investors on a cycle. In stress, exiting size is hard.

How the position is sized

  • Isolated Kamino market, conservative LTV, and sizing that respects liquidatable depth instead of headline demand.
  • The monthly attestation cadence, claims events and live DEX depth are monitored outside the issuer's own reporting.

AUTO, Hastra auto-loan yield token

Hastra's NAV-accreting Solana token backed by US consumer auto loans, built on Figure's YLDS → wYLDS stack. The higher headline yield over its HELOC sibling is compensation for a materially riskier credit book.

Protocol
Hastra, incubated by Figure Technologies and Provenance; Figure (Nasdaq: FIGR) originates and services
Base asset
YLDS, the first SEC-registered yield-bearing stablecoin (Figure Certificate Company, Investment Company Act §28), wrapped to wYLDS
Yield source
Figure's Democratized Prime Auto pool: near-prime / non-prime US consumer auto loans
Collateral
The financed vehicle, depreciating, recovered through repossession
Launched
~30 July 2026, a day-one asset with no track record yet
Rails
Agora origination, Figure Forge onchain delivery, Kamino lending, Chainlink Data Streams pricing, RockawayX and Sentora curation

Why we hold it

  • It is the counterweight to ONyc. The core position is catastrophe risk. Auto loans go bad when people lose jobs or run out of money, which has nothing to do with hurricanes. Holding a second asset that fails for entirely different reasons is the point of holding it at all.
  • The risk is legible. US auto credit has decades of loss data, mapped seasoning curves and standardised servicing and repossession practice. We know how this asset class behaves in a downturn because it has been through several. What is new here is the onchain wrapper, not the credit behaviour.
  • There is a public company's balance sheet behind the stack. Figure is Nasdaq listed, built the Provenance chain, and issues YLDS, an SEC registered face amount certificate under the Investment Company Act. Capital sits in a regulated instrument before it is lent, and the counterparty is an audited public company rather than an anonymous protocol.

How it works

  • USDC locks into wYLDS, which earns Figure's treasury-backed base yield; staking wYLDS into Hastra's auto vault mints AUTO.
  • Staked capital routes programmatically into Figure's auto lending pool; borrower interest accrues to AUTO's NAV, and unstaking realises it.

What the risk is

  • Consumer credit risk is the main one: near-prime borrowers default at materially higher rates than prime, and cyclically, delinquencies rise with unemployment, so the "uncorrelated" claim is weaker than reinsurance.
  • No equity cushion. Auto loans are commonly written at or near 100%+ LTV against a vehicle that depreciates from day one, the opposite of the ~47% LTV home-equity buffer behind its PRIME sibling.
  • Zero seasoning: launched July 2026, so no default history and unproven servicing at scale. Day-one DEX depth is thin; the real exit for size is unstaking plus loan amortisation.
  • Layered wrapper (YLDS → wYLDS → AUTO) adds smart-contract and custody dependency, and Figure is the concentrated counterparty across the whole stack.

How the position is sized

  • A deliberately smaller allocation than OnRe, in its own isolated Kamino market, priced by Chainlink Data Streams.
  • Treated as unseasoned consumer credit: conservative LTV and caps until the book shows a default history.

References

Risk disclaimer

This report was written with AI assistance and is not an offer or a solicitation. Figures are read from public sources and carry the limits set out above; net equity is a floor, not a complete NAV. Vaults are non-custodial smart contracts and may carry significant risk, including loss of the amount deposited. Levered strategies can be liquidated. Nothing here is investment, legal or tax advice, and past performance does not guarantee future results. Upshift's terms exclude US persons and sanctioned persons.