Gemini Space Station, Inc. (GEMI) Earnings Call Transcript & Summary

Thanks, operator, and good morning to all of you for being here on Gemini's Fourth Quarter and Full Year 2025 Earnings Call. I'm Ryan Todd, and I serve as Head of Investor Relations at Gemini. With me today are Gemini's founders, Cameron and Tyler Winklevoss, along with our Interim CFO, Danijela Stojanovic. Our fourth quarter and full year 2025 financial results came out yesterday. On this call, we might deliver forward-looking statements, and those could differ materially from actual results, resting as they do on management's current expectations, forecasts and assumptions. Details about the risks, uncertainties and other factors that might lead to differing results appear in our SEC filings. Today's discussion will likewise cover certain non-GAAP financial measures. You can find reconciliations to the most directly comparable GAAP financial measures in the shareholder letter posted on our Investor Relations website and on the SEC's website. Non-GAAP financial measures ought to be viewed alongside GAAP measures, not as a replacement for them. We'll begin with prepared remarks and afterward move to questions. So with that, I'll hand things over to our founders, Cameron and Tyler.

Appreciate that, Ryan. This is Cameron speaking. For Gemini, 2025 turned out to be an extraordinary year. We made the leap into public markets, becoming a publicly traded company on September 12, following more than 10 years as a private enterprise. Bitcoin's price stood at $115,000 on that date. From there, it dropped to $60,000 and later climbed back to roughly $70,000, the level where it sits now. Worth remembering that the cyclical nature of this space ranks among the toughest things crypto builders and investors face. And worth remembering that to get past those cycles, your building has to extend past them too. Our origins were as a Bitcoin company. Then we evolved into a crypto company; at present we are transforming into a markets company. If Gemini's first decade involved constructing a bridge to the future of money, these days we are constructing a bridge to that same future within markets, through a super app. Our initial step into people's everyday financial lives past buying, selling and holding crypto came with the Gemini credit card, which posted robust growth during last year. During 2025, card sign-ups increased by almost 15x while credit card revenue hit $33.1 million, a 185% rise year-over-year. A large number of these Gemini credit card holders interact with Gemini several times each day to earn crypto rewards whenever they use the Gemini credit card. December ushered in a fresh chapter for Gemini when we launched Gemini Predictions. In our view, prediction markets will match or exceed the scale of today's capital markets. They present a deep and unlimited chance to harness the wisdom of the crowds alongside the power of markets, yielding distinctive foresight about what lies ahead. By investing in obtaining a designated contract market DCM license from the CFTC so we could launch our own prediction marketplace, we have placed ourselves as an early mover on this new and thrilling frontier. For more than 10 years we have been building and running regulated marketplace infrastructure: sequencers, matching engines, order books, real-time settlement, post-trade reporting, custody infrastructure and more. A big chunk of what we're really good at is reflected here. Our prediction markets are brand-new instruments, and they run on infrastructure that we already have the expertise to build and run ourselves. So rather than bring in an outside partner or license technology from somebody else, we decided to develop it in-house. That decision also means we've chosen to put money into building the special operational skills required to create and settle thousands of contracts daily — a fresh and exciting problem, and one that gets more complex all the time, since we expect the number of these markets to keep growing dramatically going forward. In short, the reason we created Gemini predictions from scratch is that our goal is to control and run our prediction markets ourselves, from start to finish, over the long haul. Our faith in markets is strong. Bitcoin functions as a store of value, and it exists thanks to market forces. Market-based systems make for the strongest economies. Over time, markets represent truth, and we think we're only now learning how to bring them into the world around us. Whether the subject is politics, economic indicators, business, tech, culture, or sports, prediction markets outperform traditional pollsters, pundits, and news outlets at forecasting what lies ahead — both more quickly and with greater accuracy. That represents a deep shift in where the world gets its truth, and just as deep an answer to declining faith in institutions and the epistemic crisis that has followed. The printing press gave rise to the fourth estate, the public press. The Internet gave rise to the fifth estate, the decentralized public press. Now prediction markets are giving rise to the sixth estate. Decentralized information, paired with the integrity and accountability that markets provide < > in the game. Much like money, markets are an innovation and a technology that keep evolving thousands of years after their first invention. Markets have kept expanding and evolving — starting with bond markets that emerged in the Italian city-states back in the 1100s, moving on to Amsterdam's first stock exchange in the 1600s, and then, by the 2000s, electronic trading pushing out the human open outcry that had defined Wall Street's trading pits. Just when you thought the money experiment had reached its final, stable form, along came Bitcoin. Just when you thought markets had stopped maturing, prediction markets surged. Gemini exists to help build and shape a new age of money. Now we have a similar chance to help build and shape a new age of markets. And running alongside all of it is AI's meteoric rise. When technology, money, markets, and AI converge, our view is that each will magnify the others in powerful and unprecedented fashion, creating fresh economic activity — and we occupy a rare spot at the middle of all of it, able to assist in constructing and defining it as well. What this Promethean vessel of fire might produce could render everything accomplished in these areas up to now seemingly old-fashioned. For a long while, we've held the conviction that the day will inevitably arrive when customers who are machines surpass customers who are people. Opening a bank account is beyond a machine's ability, yet linking into protocols and employing crypto to turn into a sensible economic participant is simple for it. Crypto might have been created by people, yet it functions more as a currency for machines than for humans. This shift is still in its early stages. Take one illustration: during the first ten years, three API protocols existed — REST, WebSockets, and FIX. Now a fourth is being introduced: Model Context Protocol, or MCP, which is an open-source API interface designed specifically for AI agents like large language models, or LLMs. We think AI will transform who our customers are, and it is already transforming who makes up our workforce and how we operate. Not long ago, the contribution of software engineers could differ by a factor of ten, or 10x. Excellent engineers used to deliver ten times the results of merely good ones. With AI, that whole dynamic has been transformed, pushing the same principle up by at least one more order of magnitude — so someone who was a 10xer becomes a 100xer. What matters here is that we're seeing this jump in productivity show up for every single engineer who adopts AI as part of their workflow. The identical logic holds just as strongly for work that isn't engineering at all. Doing more with less has never rung so true or seemed so within reach, and as we see it, we're only at the very beginning of this curve. One thing worth calling out: the way AI acts as a multiplier — for Gemini and for our own team alike — has only just started to appear. It was only at the close of last year that AI agents focused on coding and software development experienced what you could call a splitting of the atom moment. Although various parts of our technology organization had been trying out AI within their workflows for some time already, it wasn't central to what they did. To illustrate, late last summer, during the middle of our IPO roadshow, just 8% of the code written and shipped to production involved AI; by December, though, the future showed up. Models reached an inflection point, and paired with the internal tools we created for < > management, AI has now become too powerful for Gemini not to use. As of today, AI plays a role in over 40% of the production code changes we make, and we anticipate that figure will rise to close to 100% before too long.

If you aren't using AI at Gemini before long, you might as well walk into the office carrying a typewriter rather than a laptop. Because of that, our workforce has shrunk by about 30% from where it stood at the beginning of 2026. Our view is that a leaner company that puts the right tools to work doesn't merely operate more efficiently — it moves more quickly. Gemini's story began in America back in 2015. From that point on, we pushed our operations out into over 60 countries. Winning in those overseas markets turned out to be difficult for a range of reasons, and we ended up overextended, dealing with organizational and operational complexity that pushed our costs higher and held back our speed. On top of that, the demand in those regions wasn't there to make them worthwhile. The plain truth is that the world's strongest capital markets are in America, and for Gemini, America has always been the place that matters. On top of that, the goals laid out by the current SEC and CFTC give us encouragement, as do their efforts so far toward making the super app a reality in America and bringing about a new golden age of markets. So we reached the conclusion that the moment had come to concentrate and go all in on America. Doing so lets us develop deeper, more impactful ties with both new and existing customers. With that in mind, alongside shrinking our workforce, we have cut back the regions where we do business by pulling out of the U.K., EU and Australian markets. Our expectation is that this will bring total expenses down in step with the drop in headcount and give a meaningful boost to how quickly we reach profitability, even with the crypto market as it currently stands — simplify, consolidate, then accelerate. Being a public company is something we genuinely enjoy, which may sound a bit odd given how our share price has performed across the 6 months since going public. Even so, instead of feeling discouraged, we feel driven. And though watching your stock fall is never pleasant, we value the feedback loop. It pushes us to face squarely what's working and what isn't, and it sharpens us; it's difficult, but it's unquestionably the right difficulty; in our eyes this feedback loop ranks among the biggest advantages of public company life, since growers who lose a race hand you priceless input about the changes required to win. Lost races and the priceless lessons they teach are what pave the road to the Olympics. So we embrace the feedback and relish the challenge. Gemini 1.0 came to a close in 2025, and Gemini 2.0 begins in 2026. That kicks off with our move to become a markets company through Gemini predictions and by applying the same infrastructure to run our perpetual futures contracts once those contracts get the green light in the U.S. And it carries on with our intention to roll out U.S. equities as the following stage of our platform, which gives our customers a way into the biggest, most liquid markets anywhere in the world. All in all, what we've built are the underpinnings and components of a super app, one where people can take care of both their present and upcoming financial requirements in a single spot — and something amazing is on the way.

Thanks, Cameron and Tyler — it's good to be talking with all of you. Ahead of getting into the figures, a quick note: earlier this year I took on the interim CFO position, having worked as Gemini's Chief Accounting Officer starting in May of 2025. My involvement has been hands-on with the company's financial reporting, the IPO process, and the previous 2 quarters as a public company. The wider finance team is entirely intact, and nothing has been disrupted in either our financial reporting or how we execute operationally. I'll start with several key points from the quarter, then go deeper into the results. One: revenue rose on a sequential basis even though the crypto trading backdrop in Q4 was considerably weaker. Two: meaningful diversification carried on across the business. Services revenue posted more than a doubling versus a year earlier and now accounts for more than 1/3 of our revenue. And three: the restructuring steps announced earlier this year repositioned the company with a substantially reduced cost base as we head into 2026. Now to the results. Net revenue in the fourth quarter came to $56.4 million, which is 13% higher than the $49.8 million recorded in Q3. That increase happened against a tougher market setting. Volatility in the crypto market was the largest factor behind the shift. Bitcoin dropped almost 47% below its October peak, and conditions like that placed genuine strain on trading volumes and transaction fees. The credit card business expanded throughout, offsetting part of that pressure, though Q4 was a more difficult macro quarter than Q3. Let me go through the main components. Transaction revenue reached $26.7 million, a slight increase over Q3's $26.3 million, on spot volumes of $11.5 billion versus $16.4 billion in Q3. Retail volumes totaled $1.6 billion, while institutional came to $9.9 billion. As a reminder, fees are earned from retail and institutional customers alike, with rates that differ by order type — instant orders sitting at the top of the range and active trader orders lower. Even with volumes falling, transaction revenue held up fairly well. That shows better fee economics in both retail and institutional trading, plus a mix shift within retail trading toward order types carrying higher fees. Services revenue for the quarter totaled $26.5 million, a 33% rise sequentially from $19.9 million in Q3. This category keeps expanding rapidly and stands as one of the most significant structural changes in our business. A few items deserve mention. Credit card revenue was $16 million, up 87% from Q3's $8.5 million; we brought in almost 30,000 new card sign-ups during the quarter, versus 64,000 in Q3, and receivable balances climbed to $219.8 million. Staking revenue was $5.1 million, 13% lower than Q3's $5.9 million, mostly a reflection of weaker crypto asset prices over the quarter. Still, adoption of staking keeps growing across the platform, including via auto staking features tied into the credit card rewards program. During Q4, Card Auto staking rewards were available for a whole quarter for the first time, and we also rolled out the Solana card in October. That functionality perfectly illustrates how multiple products can engage customers naturally by giving them an organic way to stake. A customer selects a reward that can be staked. Then, with each card transaction, it is staked automatically, and they become a staking customer without doing anything further. At the close of the quarter, staking balances stood at about $509 million. The adjustment to our staking fee rate, which we implemented in Q3, likewise completed its first full quarter. Now let me discuss expenses. Q4 total operating expenses came to $171.7 million, basically unchanged from Q3. Compensation and headcount costs fell to $72.3 million, down from $82.5 million during Q3, which mirrors a decrease in stock-based compensation expense. Stock-based comp for Q4 totaled $36 million. We ended the quarter with 650 employees, versus 677 in Q3. It's important to note that those figures do not yet include the workforce reduction of roughly 30% that took place in early 2026. That effect begins to appear in Q1 of 2026, and we expect the full run rate savings to show up by Q3 and afterward. Total headcount as of March 1 stood at about 445. Sales and marketing reached $39 million, an increase from $32.9 million in Q3, which mirrors continuing expansion and momentum in the credit card portfolio plus greater cardholder spending, and that produced higher crypto rewards during the fourth quarter. As we have consistently stated, marketing is treated as a variable line, and we calibrate it against what we observe in acquisition performance and growth opportunities. For the entire year, sales and marketing amounted to $97.1 million, or $52.5 million when credit card rewards and promotions are excluded, which stayed within the $45 million to $60 million range we had previously guided to. Transaction processing expenses came in at $7.3 million, lower than Q3's $8.6 million, which mirrors reduced trading volumes during the quarter. Transaction losses were $6 million, down from $7.7 million in Q3; this figure includes a $2.8 million provision for credit losses on the card, broadly consistent with the prior quarter. On the whole, credit quality throughout the card portfolio remains stable as the book scales. Technology and infrastructure totaled $22.3 million, higher than Q3's $20.3 million, primarily reflecting greater cloud infrastructure and software licensing costs as the platform grew. General and administrative came to $24.9 million, up from $19.3 million in Q3, mainly due to increased professional services and continuing public company operating costs. For the full year, tech and G&A reached $154.6 million, consistent with our guidance range. Now, briefly, to full year metrics: we had roughly 601,000 MTUs as of December 31, which is 17% higher year-over-year, reflecting ongoing growth in engagement as users take up more products across the platform. Net revenue for the full year was $174 million, versus $141 million in 2024, an increase of 24% year-over-year. Transaction revenue for the year was $98 million, while services and interest revenue came to $76 million, making up a significant and expanding share of our overall revenue base. This move into services represents an important structural transformation, lessening our reliance on trading. Revenue from services and interest exceeded the $60 million to $70 million guidance we gave during our third quarter earnings call; the main driver was card flows that outperformed expectations, with over 116,000 new card sign-ups over the year following card addition launches like the XRP card; we also experienced expansion across a number of other services categories. Custodial fee revenue rose 25% year-over-year, fueled by greater average crypto assets under custody; in addition, we recorded roughly $4.8 million in advisory revenue tied to services rendered for a strategic customer, plus $1.2 million from new on-chain offerings, such as integrations and token listing services. As we keep growing the platform, we anticipate rising chances to generate monetization through various services as users interact with more products outside of trading. For the full year, total operating expenses reached $525 million, compared with $308 million in 2024. The increase from the prior year stemmed mostly from 3 main factors: first, stock-based compensation connected to the IPO, which includes the Q3 bonus accrual settled in equity; second, the substantial marketing investments we undertook after becoming public to fuel card growth; and third, ongoing spending on technology, compliance and public company infrastructure costs. These investments were intentional, and the restructuring steps we announced are meant to reset the company's cost structure moving forward.

For the entire year, adjusted EBITDA came in at a loss of $258 million, and that figure includes $33.4 million in net realized plus unrealized losses. Under GAAP, our net loss for the full year totaled $582.8 million; worth emphasizing is that a big chunk of that net loss stems from noncash items; among them are $178.5 million in fair value losses tied to our prior related party instruments along with mark-to-market adjustments on crypto assets, plus $85 million in stock-based compensation expense connected to equity awards granted alongside our IPO. In our view, adjusted EBITDA offers a helpful lens on how the business is performing underneath; even so, where adjusted EBITDA landed is not acceptable to us, and since the year ended we have taken steps meant to fix that. Now a quick word on the balance sheet. We closed out the year holding roughly $252 million in cash and cash equivalents. The biggest cash outflow during the quarter came from the $117 million repayment of the Galaxy loan, which we finished in Q4 and which took that obligation off our balance sheet. Because of that, we head into 2026 with a cleaner balance sheet and less debt. After the restructuring moves we announced earlier this year, we anticipate a meaningful drop in our normalized operating cash losses. Looking ahead, we're concentrated on closing the distance to profitability by managing costs with discipline and growing revenue from higher-margin services. At year-end, the card warehouse facility showed $154.4 million outstanding versus $188 million in pledged receivables, backing a capacity of $250 million. As our receivables book expands, we will put additional funding capacity in place to support the growth we anticipate; regarding restructuring costs, the $11 million in pretax charges tied to the Gemini 2.0 plan will fall almost entirely within Q1 of 2026 and should be cash charges. They account for the U.K., EU and Australia wind down along with the headcount reductions. When exactly some of the international components happen will hinge on local consultation requirements, though we anticipate the plan as a whole will be substantially finished by midyear. We expect these steps to streamline the organization and bring down our operating cost base as we move forward. Before getting into the full year outlook, allow me to share what we've observed in Q1 2026 thus far. Through February, trading volume reached roughly $5.3 billion, lower than Q4 levels as trading activity more broadly kept softening. On the card, payment volume has topped $330 million alongside over 150,000 open card accounts; and on predictions, about 15,000 users have traded since launch spanning more than 12,000 listed contracts. Across the platform, total monthly transacting users stood at approximately 606,000. As ever, we caution against extrapolating from partial quarter activity; with that caveat in mind, let me explain how we're approaching fiscal year 2026. For the time being, we are not offering total operating expense guidance for the year. Since our cost structure is still being reworked and the broader economic picture remains hard to predict, we believe it makes more sense to address each major expense line on its own. The restructuring moves we made earlier in the year started showing up in our cost structure during Q2. Compared with where we ended the year, our headcount is down roughly 30% from its highest point. Since 2025 compensation assumed a full year at staffing levels from before the restructuring, the decline measured year over year looks smaller than the actual reduction in headcount; for compensation, leaving out stock-based comp and restructuring charges, we anticipate a drop of 15% to 20% versus 2025. For stock-based compensation, we expect a total of $100 million to $115 million in 2026. Following our September listing, 2025 contained just 2 quarters of stock-based compensation at post-IPO levels. In absolute dollars, the full year number comes in higher, though the quarterly run rate is settling down now that the IPO-related grant cycle is normalizing. Technology and G&A should land somewhere between $155 million and $190 million. The bottom of that range mirrors the normalized base after restructuring. How wide the range is comes from variable costs that move with card and trading activity, and we intend to tighten it as visibility improves over the course of the year. Marketing expenses, not counting rewards and promotions, should come in at 10% to 15% of revenue, subject to market conditions and the opportunities we identify in our highest returning acquisition channels. Turning to revenue, our credit card product is still the main driver of acquisition and growth. It is still early to make predictions, yet with over 15,000 users since December, the initial traction strikes us as encouraging, and it sits at the center of where the company is headed. Although 2025 ranked as the priciest year in the company's history, owing to our IPO, the card investments and international expansion, the steps taken since then are meant to make sure 2026 looks very different financially. All told, we believe the organization we carry into 2026 is leaner, more focused and ready to deliver better operating leverage as our business keeps scaling. Taken together, we expect these dynamics to produce an improvement in adjusted EBITDA in 2026 as we run with a more disciplined cost structure and a more diversified revenue base. In short, 2025 was a year of major transformation for Gemini. We went public, scaled our credit card program, expanded and diversified revenue through services, launched prediction markets and took decisive steps to reset our cost structure. We enter 2026 with a simpler organization, a lower expense base and a more durable business model. The way we see it, the core story of Gemini today is straightforward. The business is becoming less dependent on crypto trading volumes and increasingly driven by recurring and diversified platform revenue; and with that, we will now turn to questions. Thanks, everyone.

James Yaro from Goldman Sachs has our opening question. He's asking whether we could bring everyone up to speed on what's behind the recent exits among our executives, and where that leaves things relative to the new strategy we've laid out.

Appreciate that question. Look, things looked completely different back in the summer. Then we went public in September, and at that moment one Bitcoin cost roughly $115,000. Obviously, from there the market fell sharply. On top of that, when it comes to launching a super app here in America featuring predictions, a route ahead now exists for doing so. And given how AI has hit an inflection point, we've concluded that a leaner, flatter company powered by AI can let us move more quickly, while remaining, naturally, heavily founder-led; so in our view the team and the org design we have fit both the present and the future.

Matt Coad from Truist is up next, and his question is this: even with the tough conditions in the crypto market, you keep gaining momentum and expanding your user base. In your view, what's behind that growth in users? And what's your strategy for cross-selling prediction markets into this big and expanding base of users?

Appreciate the question. I suppose I'll take this one first, and then perhaps pass it along to Cameron or Tyler so they can add a bit more regarding predictions. So the ongoing expansion we're seeing in our user base has us quite encouraged, especially considering the wider market environment. A major factor behind this, I believe, is that we keep experiencing substantial user acquisition via our credit card program, along with deeper engagement that comes from new products we're rolling out and from diversifying our revenue base, like predictions. So we'll turn it over and see whether Cameron or Tyler would like to say a little more about predictions.

When Gemini first got going back in 2015, Bitcoin was our whole business. People would show up and use us to purchase, trade, and hold Bitcoin. As time went on, we turned into a crypto company, and we brought in extra money features such as staking, letting users put their assets to work with us. Then came the Gemini credit card, which has turned into a regular piece of many customers' financial lives as they look to earn crypto back whenever they swipe. We'll keep introducing new products so that users find more reasons to engage with us as time passes. Eventually, plenty of these activities are going to stay independent from crypto cycles. I think we're eager to watch the engagement around prediction markets, our credit card, and the other offerings we plan to add to the Gemini app, so that people feel no need to look elsewhere.

Adam Frisch from Evercore is up next, and his question is whether we can help lay out how stand-alone card economics stay positive on a sustainable basis — in particular, how much each of these contributes relatively: optimizing rewards, reduced costs to acquire, normalization in provision and credit, and broader funding capacity that's cheaper.

I'll take that question. Appreciate it, Adam. We feel genuinely good about how far we came during Q4 — we got the card to roughly break-even. Growth on the card side has happened very fast, and in our view there's an obvious route to making money once the portfolio seasons. We tend to think about a handful of main levers. Number one, on the top line: interchange is fueling solid gains as spending rises. Also worth flagging, with respect to how big the receivables book is, interest income is still not earning its full potential. So as that portfolio seasons and matures, interest income turns into a significant tailwind. Turning to expenses, we have multiple levers, honestly. Rewards represent our biggest cost right now, though that was deliberate and pushed upfront to fuel adoption and truly get the credit card established — and that approach paid off. At the beginning of '25 we had about 30,000 open accounts, and as of March 1 we're now above 150,000. Consider this: the Bitcoin card has existed for only around 9 months, XRP for roughly 5 months, so this program is still in its early days. Rewards sit squarely under our control, and over time we intend to fine-tune them. On top of that, we've been quite happy with the organic sign-up trend even on a smaller spend base; we continue to average comfortably above 100 sign-ups each day, which is over twice what it was 12 months ago. Beyond that, bank fees are improving as we grow, and that should mirror stronger underlying economics; from the credit angle, results are moving the right way too. Loss rates appear to be leveling off and should keep getting better as the book matures; last, funding. Funding costs are now entering the model, and we anticipate them growing more efficient both as the portfolio expands and as financing options widen. Expanding the funding facility is genuinely a major step; over a longer horizon, we see chances to bring down the cost of capital and broaden funding sources as the portfolio grows. So when you pull it all together, we're already close to break-even before provisions, and the route to lasting profitability really comes from a mix of portfolio seasoning, cost optimization and efficiencies born of scale. No single lever has to carry everything. It's the incremental gains in each one of these areas that we think will push the card business into steady profitability.

Michael Cyprys from Morgan Stanley is up with the next question. By the close of February, 15,000 users had engaged with Prediction Markets. What does that mean in terms of revenue? From that point, where's the growth opportunity? And when peers have more active users than you do, how do you compete?