UNTY — Unity Bancorp, Inc.

Deep-value analysis — UNTY (Unity Bancorp, Inc.)

Date: 2026-09-04 · Exchange: NASDAQ · Reference price: … (yfinance, close 2026-09-03) · Market cap: …M · BVPS: … · EPS ttm: …

Regime: REFRESH over the reference thesis from 2026-07-27 (verdict then: BUY, GBL score …). Mechanical trigger: 3 new filings — the Q2 2026 10-Q (filed 2026-08-06), the 8-K from 2026-08-07 (a director’s death), and the 8-K from 2026-08-20 (dividend raised). The 2026 filings were missing from the local SEC file set; I downloaded them manually from EDGAR, since edgar_10k_downloader.py only looks for 10-Qs within the fiscal year of the last 10-K (2025-12-31) and doesn’t see 2026 quarters.

What I inherit unchanged from the July 27, 2026 report (re-verified selectively, not re-derived): the business description and moat assessment; James Hughes’s capital-allocation history through the end of 2025; the company’s own historical P/E and P/B multiples for FY2022–FY2025; the CRE sector concentration from the FY2025 10-K (the owner-occupied portfolio by industry); the governance structure and insider-ownership level. What I fully re-derived, because the new filing moved it: the entire credit analysis (the key falsifier broke), the normalized-earnings bridge, all four valuation models, the Monte Carlo simulation (entirely absent from the prior report), and the verdict.

The price move since the prior analysis is zero (… → … … over 38 days). This revisit has nothing to do with price; it has to do with what appeared in the 10-Q.

Executive summary

The thesis, in one sentence: Unity Bancorp remains an operationally excellent bank — NIM 4.56%, ROA 2.01%, cost/income 41.3%, CET1 14.19% — but the August 6, 2026 filing breaks the prior thesis’s central premise: credit deterioration is no longer isolated to a single …M CRE loan, it has spread into the residential and consumer portfolios, and the “resolution” of the flagship loan is actually 80% seller financing that doesn’t qualify as an accounting sale.

The broken falsifier. July’s thesis rested the BUY verdict on the explicit claim that the NPL jump from 0.58% to 1.17% reflected a single, well-collateralized CRE name. At June 30, 2026: nonperforming loans are …M (1.23% of the portfolio, versus …M / 1.17% at 31.12.2025 and …M / 0.67% at 30.06.2025). The increase since end-2025 comes from segments other than CRE: residential mortgages …M → …M (…), home equity …M → …M (…), residential construction …M → …M. CRE is practically flat (…M → …M), while SBA and C&I decline. In parallel, commercial loans classified Special Mention explode from …M to …M (…), and “potential problem loans” — loans with a non-pass risk rating that are still performing — nearly double, from …M to …M. Total criticized loans in the rated portfolios (SM + Substandard): …M → …M, … in six months, in a portfolio that grew 5.4%.

The …M loan wasn’t resolved, it was refinanced. Note 11 (Subsequent events) of the 10-Q: after June 30, 2026 the bank sold the nonperforming CRE loan “at par,” but financed 80% of the price to the buyer through a 3-year, interest-only balloon loan, secured by the very asset sold. Because the bank retains indirect credit exposure, the transaction doesn’t qualify for sale accounting. Only the 20% cash down payment (…M) effectively reduces nonperforming loans; the remaining ~…M of risk stays on the balance sheet in a new form. It’s the “extend and pretend” playbook, correctly and transparently disclosed — but correctly disclosed doesn’t mean resolved.

Estimated value. Four bank-adapted models plus a 20,000-scenario Monte Carlo simulation give a range of … / … / … around the … price. Median intrinsic value from the simulation: …. Probability the stock is undervalued: …% — practically a coin flip. The 5-analyst consensus (average target … …) overlaps with my optimistic scenarios, not my base ones; I don’t have a figure contradicting the market, I have a figure that refuses to assume provisions stay at 0.05% of loans indefinitely.

Verdict: HOLD / MONITOR — downgraded from BUY. Not because of price (unchanged), but because of two new, verifiable facts: (1) credit deterioration has spread beyond the perimeter assumed in the thesis, and (2) provision coverage of nonperformers fell from 184% (June 2025) to 105% (June 2026), with zero specific allowance on all …M of individually evaluated loans. With a negative median margin of safety and a broken falsifier, the thesis no longer justifies a new position. What remains is a good bank at a fair price, with credit risk that has shifted from “isolated” to “watch quarter by quarter.”