Open in interactive viewer → charts, metric popovers & call review

Oxbridge's Premiums Fell to $86,000 a Quarter. The Fix Is a Data Center Business That Hasn't Broken Ground.

A $7.3 million market cap, $19.5 million of (mostly restricted) cash, and a plan to sell tokenized reinsurance to fund 10–100 MW nano data centers by 2027.
OXBR · Earnings Call · 2026-11-09

The underwriting book has quietly gone to zero

The headline loss at Oxbridge Re is small in absolute terms — $421,000, or $0.05 a share, for the September quarter. What matters is the denominator underneath it. Net premiums earned collapsed to $86,000 from $555,000 a year earlier, and the company's own explanation is disarmingly plain: less capital being put to work. The expense ratio went to 752.3% from 146.8%. Read that again — for every dollar of premium earned in the quarter, the company spent roughly $7.50 running itself. Nine-month figures look better only for accounting reasons. The combined ratio improved to 185.8% from 288.6%, but that is because “the loss ratio decreased to 0% from 132.4% for the nine-month period ended September 30th, 2026, when compared with prior comparative period” — Wrendon Timothy, Chief Financial Officer · 2026-11-09 — no losses booked this year versus a full-limit loss last year. Take away the loss accrual and what remains is a fixed cost base sitting on a shrinking revenue stream. Effective SG&A of $522,000 in the latest filed quarter against total revenue of $623,000, itself down 10% year over year. The whole-historical shape is stark: revenue sits 93% below its 2017 high. And the keyword that dominated the prior quarter's call, underwriting losses, has fallen off hard in the current quarter's lexicon — because there is barely any underwriting left to lose money on.

Two pivots, one balance sheet, and a script that hasn't changed

The SurancePlus story, to be fair, advanced. Management reported five completed offerings for the 2026-2027 treaty year at roughly $7.1 million in aggregate gross proceeds, and — for the first time — three of those were tied to third-party reinsurance originated by HCI Group and Fortex Re rather than by Oxbridge itself. That is the genuinely new fact on the call: “we've completed five offerings for the 2026-2027 treaty year, with approximately $7.1 million in aggregated gross proceeds. Two were tied to reinsurance originated by Oxbridge” — Jay Madhu, President · 2026-11-09. A platform that can issue securities against someone else's risk is a different asset than a platform that only wraps its own. The announced minimum on the HCI offering was $5,000, which is the whole point of the retail-facing reinsurance offering model. AI GridWorks is another matter. Launched in August, it now has a team and a thesis: “We're talking about nano data centers, 10 MW-100 MW... Power first, we're looking at distribution. Primarily, we're looking at centers where we can get 2027 power” — Jay Madhu, President · 2026-11-09. There is no site under contract, no tenant, no capital committed beyond team and diligence spend, and the timing answer to two separate analyst questions was essentially the same word: 2027. Here is the tension. That is almost verbatim the description given three months earlier, minus the concrete steps: “What we are targeting is the nano data centers, 10 to 100 megawatts. That way, we have an opportunity to play in a space where we are not competing with some of the juggernauts in the space” — Sanjay Madhu, Chairman, President and Chief Executive Officer · 2026-08-13. In August the call walked through optionality, entitlements and value creation; in November the same optionality was re-described, with site selection still described as an evaluation exercise and counties still being courted. The one meaningfully new operational detail is the political read: management now says the areas it is targeting have no data center moratoriums and that officials remain receptive, a direct response to the political pushback that had been the obvious external risk since the NIMBY discussion on the prior call. The genuinely novel idea — and it is a clever one — is using the tokenization rails to defuse local opposition by reserving an allocation for the host community: “hypothetically, in a particular county, we can reserve an allocation to residents of that county” — Jay Madhu, President · 2026-11-09, with residency verified on-chain during onboarding. That converts county-level goodwill into a financing channel, and it links project financing to the securities platform instead of running two unrelated stories side by side.

There is a tremendous amount of value creation in getting — in taking land and moving it into that next step where you have your entitlements that are put into place and you also have power that's your power letters. That creates a significant upside over there.

Sanjay Madhu, Chairman, President and Chief Executive Officer · 2026-08-13

The tape has already voted, twice, against both legs

Oxbridge shares are down 38.7% over the last 90 trading days and sit 47.3% below the July peak of $1.70; the lifetime chart is a loss of 88.4% from the 2014 high near $8. Against a $7.3 million market cap, the company reported “combined cash and restricted cash and cash equivalents increased by $12.5 million to $19.5 million from $6.9 million as of December 31st, 2025” — Wrendon Timothy, Chief Financial Officer · 2026-11-09. On the surface that is nearly three times market cap in cash. The composition is the story. Restricted cash is collateral posted against reinsurance treaties; the increase reflects the release of collateral from the successful 2025-2026 treaty contracts as much as anything earned. Effective net cash of $8 million, down 34% year over year. A market capitalisation below the cash pile, on a company with liabilities at 24.3% of assets, up nearly 14 points year over year, is not a market forecasting a windfall. It is a market pricing in burn and dilution. Note also that the company's own Real World Asset keyword momentum swung sharply negative this quarter after being a top thematic driver in the June quarter — the narrative engine of the last two years is losing lift even in management's own language. And the broader tape offers no tailwind. The global data center complex is being de-rated, not re-rated: data center capacity is down roughly a third over the last 30 days across its constituent tickers, and the wider data center / AI GridWorks-adjacent cluster is among the weakest cohorts in the market over both 30 and 90 days. The tokenization cohort is not helping either — real world assets names have sold off about 24% over the last month. Oxbridge is asking investors to underwrite both a 2027 power thesis and a tokenized securities thesis during the one stretch when neither is in favour.

What would actually change the story

Three things, in ascending order of difficulty. First, a fully third-party digital reinsurance deal of real size, which would prove SurancePlus is a platform rather than a wrapper. Second, a named site with an executed utility power commitment and a county approval — the moment nano data centers stops being a category and becomes an address. Third, some evidence the expense base is being right-sized to a premium line that is now six figures a quarter. The dilution question is the one management answered most candidly, and it deserves to be the framing for the next twelve months:

These data centers are highly capital-intensive, but in the starting phases, when you're talking about land, you're talking about making sure you have the various different items that needed to be going forward, that's a small dollar amount... We're doing this sequentially, and at every different stage, we would do what's needed to be done, but we would be acutely aware of dilution.

Jay Madhu, President · 2026-11-09
That is the right instinct. The trouble is that a company with a $7.3 million equity value, roughly $8 million of effective net cash and a business line earning $86,000 of premium per quarter cannot fund even a 10 MW site from the balance sheet. It will need third-party capital — bank debt, a partner, or a tokenized raise — before it can argue the model works. Sequentially is a strategy only if the first step is small enough to survive being wrong. 3.7x price to revenue on revenue that has shrunk 93% from its peak is a reminder that the market has never valued this company on its earnings power — only on optionality. The optionality just got two new branches, and the branches are being priced at a discount.