The house always wins

In the first piece I argued that investing, gambling, and entertainment have collapsed into one category, and that the collapse is dissolving the stigma that kept money out. This is the piece about the money. Specifically: why the business sitting underneath all that flashing UI is one of the most beautiful — and I mean that as a technical compliment, not a moral one — that anyone has ever built, and why the most pattern-matching industry on earth somehow never bought it.

Venture capital has a checklist. High gross margins. Near-zero marginal cost. Recurring revenue. No inventory, no physical COGS, demand that doesn't care what season it is. Pricing power. A product that improves as it scales. When a founder walks in with all of those, the term sheet writes itself. Now read the list again and describe an online casino. It is, line for line, a perfect score — a business that prints software-grade economics off a four-hundred-year-old mechanic. And until very recently, the people who fund software wouldn't touch it.

The math is the whole business

Begin with the identity every other number in this essay descends from. A casino's revenue — its gross gaming revenue, GGR — is just the total amount wagered, the handle, multiplied by the slice the operator keeps, the hold:

What makes this magic rather than ordinary is that the hold is fixed in advance, by mathematics, not by luck. A slot's return-to-player is a number set in the software; a roulette wheel's house edge is a property of the geometry. The operator does not gamble. The operator sells access to a process with a guaranteed positive expected value to the operator, and then makes money in direct, almost boring proportion to how much volume runs through it. Revenue equals volume times a constant you chose — that is the line every SaaS founder wishes they could write, and the casino has been writing it since before the steam engine.

Now layer on the cost structure. Once the platform exists, dealing one more hand of digital blackjack costs essentially nothing — a sliver of compute and payment fees. There is no marginal unit to manufacture, no widget to ship. So the gross margin is enormous and the operating leverage is violent: every incremental dollar of handle drops a near-constant fraction straight toward profit. You can see it cleanest in the part of the industry that sells the picks and shovels. Evolution, the Swedish company that supplies live-dealer casino games to operators worldwide, runs an adjusted-EBITDA margin around 66% — it was 68% in 2024 — on roughly €2 billion of revenue.¹ Sixty-six cents of operating profit on the dollar. Most of the software industry would saw off a limb for that number. Evolution gets it dealing cards.

The operators who run the player relationship earn less than the supplier — call it 20–30% at the online-pure-play level² — because they carry the marketing and the licensing. But that is still a software margin on a product with structurally higher retention than almost anything in consumer tech, which brings us to the second beautiful property.

No season, no off-switch

A sportsbook is a hostage to the calendar. Roughly a third of a US book's annual revenue arrives in the four months of the NFL season; summer is a desert.³ Worse, its margin is genuinely at risk on any given Sunday — a weekend where every favorite covers is a weekend the book bleeds. The sportsbook is in the business of managing a bet. It can lose.

The online casino has no season and cannot lose. The player who likes slots at 11 p.m. in February likes them just as much in July, and the math holds every single night regardless of who's playing or how the games "land," because the games don't land — they grind. Revenue is a smooth, predictable line instead of a sawtooth. You can see both truths in one company: DraftKings, mostly known as a sportsbook, finally turned its first annual profit in 2024,⁴ and in the fourth quarter its online-casino arm booked $426 million — already about half of its sportsbook revenue, earned in just five states while the sportsbook operates in twenty-five.⁴ The smaller, newer, far-more-restricted casino business is the one with the cleaner economics. By my read, online casino is on the order of a quarter of DraftKings' revenue and something closer to half its profit — and it gets there from a fraction of the footprint.

Put the two businesses side by side and the contrast is stark enough that you wonder how they ever got filed under the same word.

The reason the casino's economics compound is lifetime value. With near-zero marginal cost, almost all of a player's lifetime spend is margin:

A retained casino player is an annuity. That is why an operator will pay startling sums to acquire one, and why the whole industry's growth math works: the cost is front-loaded and the margin is forever.

The ramps look like AI companies

If the economics are this good, the growth curves should be obscene, and they are. Look at Betty, a Canadian online casino aimed at women that launched in 2022. It exited 2024 at roughly a $114 million annual run rate; by the third quarter of 2025 it was running at about $334 million Canadian — call it $246 million US — and had just posted its first EBITDA-positive quarter in Ontario.⁵ Roughly tripling in a year, turning profitable, in well under three years from launch. That is not a gambling curve. That is the shape of a hot enterprise-software company.

And it gets there on its own cash flow, which is the part that should make an investor sit up. Compare the celebrated AI ramps of the same period — ElevenLabs, the voice-AI darling, reached something like $200 million of annual recurring revenue in about two years.⁶ Comparable trajectory — except ElevenLabs raised enormous venture rounds and burns to grow, while a casino like Betty funds its growth out of margin it generates on day one. Or look at the apex predator: Stake, the crypto casino, went from about $105M of gross gaming revenue in 2020 to $2.6B in 2022 to roughly $4.7B in 2024⁷ — a top-ten global gambling operator built almost entirely offshore, with no US license and almost no conventional advertising. The way Stake actually markets is its own subject — paying streamers to gamble on camera and flooding the feeds with the clips. The point here is narrower and about the P&L: this is venture-scale growth, self-funded, in a category venture mostly watched from the sidelines.

~$246M

Betty's run rate by Q3 2025 — up from $114M a year earlier, now EBITDA-positive⁵

66%

Evolution's adjusted-EBITDA margin — a software number, dealing cards¹

$4.7B

Stake's 2024 GGR — a top-ten operator, no US license, built on streamers⁷

So why did VCs skip it?

If the business is this good, the interesting question isn't "why would a VC invest" — it's "why didn't they, for decades." Three reasons, none of them about the economics.

Geography. The talent that knows how to build and run online gambling is concentrated in Europe — Malta (which hosts hundreds of iGaming companies and earns something like a tenth of its GDP from the sector), Sweden, Gibraltar, the UK.⁸ The capital that funds venture is concentrated in the United States. For a long time those two worlds barely spoke. Silicon Valley wasn't going to wire a Series A to a team in Valletta it had never met, to build a product it found faintly embarrassing.

Regulation and tax. The product is legal in a patchwork and taxed at rates that would terrify a software founder. Pennsylvania taxes online slots at an effective ~54% of revenue; New Jersey takes about 9%.⁹ Every state is its own license, its own filing, its own compliance build. That friction is real, and it scared off generalist investors who are used to a product you can ship everywhere at once.

The vice clause. This is the quiet one, and probably the biggest. The people whose money fills venture funds — public pensions, university endowments, sovereign wealth funds — frequently write explicit prohibitions on "sin" investments into their agreements. A VC firm whose limited partners ban gambling exposure simply cannot do the deal, however good it is. As one investor told Fortune in late 2024, firms "initially resisted gambling investments, citing LP agreement restrictions."¹⁰ The business didn't fail the returns test. It failed the reputation test, upstream, before a partner could even pitch it.

Why the walls are coming down

Every one of those three is eroding at once, which is what makes this a moment rather than a permanent state.

The convergence did the cultural work: once sports betting is a primetime ad and "investing" shares an icon with "betting," the reputational tax that powered the vice clause starts to look quaint. And the screens were always a bit of a fiction — Norway's trillion-dollar sovereign fund already holds stakes in Evolution, Flutter, and MGM even as its ethics council opened a 2025 review of exactly those holdings.¹¹ The big money is in the public names already; the formal prohibition is a fading habit, not a wall. Geography stopped mattering when the frontier of the industry moved to crypto, where the builders are global and native to the rails. And the friction of regulation turned into a moat for whoever's willing to do the work — the very thing that scared amateurs away protects the professionals who show up.

The clearest tell is that the disruption has started from the inside. In September 2025, BetHog — a crypto casino founded by two of FanDuel's co-founders — launched what it billed as the world's first AI-powered blackjack dealer, a synthetic dealer running 24/7 with no studio and no human shift.¹² Sit with what that attacks. Evolution's whole 66%-margin empire is built on human live dealers in physical studios — its one genuine cost center. An AI dealer is a software-margin attack on the highest-margin business in gambling. When founders start aiming AI at the casino's last remaining cost, the casino has officially become a technology category.

The mispricing — and what's hard

Here's where the opportunity actually lives, and where I want to be careful. Because gambling carries legal and reputational risk, the market prices its cash flows at a discount — unregulated offshore operators trade for a fraction of what a "clean" regulated operator commands, even when the offshore cash flows are larger.¹³ That is a textbook mispricing born of stigma, and the convergence is exactly the force that closes it. The AGA estimates that unregulated operators still capture about a third of all US gaming revenue, and that illegal online slots and table games alone pull in roughly $18.6 billion, up nearly 40% in three years.¹³ A vast, fast-growing, deeply profitable market trading at a discount because it's disreputable — that's the trade.

But I won't sell it without the other side, because the discount is not pure free money — it's compensation for genuine risk. An offshore operator can be de-banked, license-stripped, geo-blocked, or prosecuted; "software margins" quietly omit the cost and the peril of operating a product half the world is trying to regulate or ban. The valuation framing people throw around — offshore at a fraction of revenue, regulated at several times it — is stylized; real public operators cluster closer together than the slogan suggests, and the very highest multiples belong to the suppliers, not the casinos.¹³ The vice premium is real, but it is earned by sitting in legal jeopardy, not conjured from nothing.

And the deepest tension is the one I flagged in the first piece and won't stop flagging: this magnificent P&L is, mechanically, the sum of other people's losses. The margin is the edge, and the edge is paid by players who, on average and by design, lose. You can admire the engineering of the machine and still believe we should be careful where we point it. Both things stay true through all three of these essays.

The house always wins — that's not a warning, it's a balance-sheet description. What's new is that the people who fund houses are finally allowed to notice. The last question is the biggest one: where in the world is all this money actually pooling?