Flex Raises $70M Led by Halo Fund to Scale AI-Native Private Banking for Global Business Owners

Every year, the owners of the world's middle-market companies surrender somewhere between two and five per cent of every dollar they move across a border to a system that was designed for banks and was never designed for them. It is not lost to fraud. It is not the product of a badly negotiated banking relationship. It evaporates in the space between the invoice and the settlement, inside FX markups that never appear as a line item, inside correspondent bank deductions taken silently while the money is in transit, and inside working capital trapped for three to five days in a clearing architecture that nobody has owned end-to-end since SWIFT was designed in 1973. Flex believes that gap, and the far larger gap sitting beside it, is the last unclaimed category in business finance. The San Francisco company announced this morning a $70 million Series B1, led by Ryan Smith and Ryan Sweeney's Halo Fund, with participation from Portage, Wellington, Crosslink Capital, 53 Stations, Titanium Ventures, Spice and Florida Funders, to scale an AI-native private bank for high-net-worth business owners and to launch Flex Global, a cross-border layer that settles owner payments on stablecoin rails across more than 100 countries. Reuters, citing a person close to the deal, put the valuation at roughly $1.2 billion, more than double the mark the company carried six months ago. Total capital raised now stands at $180 million in equity and $300 million in debt.' The pitch is unusual for a fintech company in 2026, because it contains almost no productivity promise at all. Flex does not claim to make a controller faster, a bookkeeper smarter, or an accounts-payable clerk more efficient. It claims to become the institution the owner banks with, on both sides of a balance sheet that every incumbent insists on treating as two unrelated customers, and to be measured in products attached and dollars moved rather than in seats sold. The Most Expensive Gap in Business Finance The numbers behind Flex's thesis come from a decade of research that every regional bank in America has read, internalized, and never acted on. The National Center for the Middle Market counts roughly 200,000 US middle-market companies employing 48 million people and generating about one third of private-sector GDP. Flex sizes the population of owners it serves at 350,000 to 400,000 in the United States and around three million globally, and puts the payroll they drive at 40 per cent of the American private sector. Whichever number you prefer, the shape of the finding is the same: this is one of the largest concentrations of economic activity in the developed world, and it is served by the thinnest software stack of any customer segment in financial services. The structural problem is simple to state and brutal to solve. Every institution in the owner's financial life is a system of record, and none of them is a system of execution across the whole of that life. The regional bank records the operating account. The card issuer records the spend. The private banker, if the owner is large enough to have one, records the household. The lender records the loan. Then the owner's actual financial life enters the wild, and nothing in the stack knows that the same person is simultaneously running three entities in two currencies, guaranteeing a facility with a personal asset, paying a Polish supplier out of one account and a school fee out of another, and being underwritten twice by two institutions that will never share a file. The macro condition making this market addressable right now is hidden inside a curve that has only sharpened with time. Correspondent banking relationships have contracted by roughly a fifth since the mid-2000s as large banks exited corridors they judged too small to be worth the compliance exposure, which means the owner who wants to pay a vendor in Warsaw is being served by a shrinking number of intermediaries with growing pricing power. Banks still process 92 per cent of B2B cross-border flows overall, but their share falls to 76 per cent on small and mid-sized transactions, which is the precise band the middle-market owner lives in. Traditional bank cross-border pricing runs 2 to 5 per cent all in once the FX markup, the wire fee and the correspondent deductions are counted, against 0.3 to 0.8 per cent from a specialist fintech and between five and fifteen basis points on a stablecoin rail. These are not conditions that get fixed by adding another point solution to the stack, because the problem is precisely that the stack already contains too many point solutions and no single layer accountable for the owner's financial outcome. Founder-Market Fit, Measured in Products Per Customer If any founding team has earned the right to make that diagnosis, it is, plausibly, this one. Zaid Rahman started what became Flex in 2022 as a construction-focused fintech called Flexbase, and did the unfashionable thing for a company founded in that vintage, which was to spend the first product cycles not on the demand-generation layer but on the credit layer, in a segment that no venture-backed fintech wanted and that the bear market of 2023 made even less fashionable to want. The result, three years in, is a platform that is not one product with adjacent features but five product lines running against the same customer: private credit, a business finance stack across banking and cards and expense management, a personal finance stack that manages the household, a payments layer covering receivables and payables, and an agentic back-office operating system with AI agents running underwriting, cash management and reconciliation. The number that closed this round is the one that describes what happens when those five things sit on a single customer record. The average Flex customer now uses four or more products. That is the hardest metric in business fintech to move, and the reason it is hard is that most fintechs bolt products onto a relationship they never owned in the first place. Publicly traded fintechs a decade into their existence, having spent hundreds of millions of dollars on acquisition, are only now reaching three or four products per customer. Flex reached it in three years. The trajectory between funding announcements is the data point that closed the round. In March 2025, Flex raised $25 million at just under a $250 million valuation on roughly $1 billion of annualised payment volume. In December 2025, it raised a $60 million Series B led by Portage at roughly $500 million, on $3 billion of volume. It now sits at a reported $1.2 billion on more than $10 billion of annualised payment volume, with revenue up 3x in the seven months since the last round, volume up 4x year on year, and a nine-figure annualised revenue run rate. Valuation has moved 4.8x since March 2025. Payment volume has moved 10x. That is a company that has repeatedly outrun its own price, and the multiple has compressed on every round rather than expanded, which is the opposite of the pattern that gets fintechs into trouble. One Execution Layer, Both Sides of the Owner's Balance Sheet Flex connects to the owner's operating reality rather than replacing it, interprets the obligations attached to a business and to the household that owns it, executes across both, and follows the money from the moment it is earned to the moment it is spent personally. No rip-and-replace of the accounting system. No demand that the owner abandon the regional bank on day one. That deployment posture is itself a strategic weapon, because it sidesteps the migration friction that has killed most business-banking momentum and lets the platform prove itself against live cash flow from week one rather than month thirteen. The business credit card, with sixty days of interest-free float on purchases, is the gateway. Adoption of the card opens banking, then expense management, then bill pay and accounts payable automation, then treasury, then working capital, and then, once the platform has seen enough of the business to underwrite it properly, private credit. In December 2025 the company extended the same architecture into the household with Flex Elite, an invite-only consumer card positioned directly against Amex Centurion, which is the point at which the strategy stopped being a business-finance strategy and became a private-banking strategy. Today Flex adds the international dimension. Flex Global ships stablecoin payment rails and wallets across more than 100 countries with cross-border settlement in minutes, institutional US dollar accounts for foreign business owners who need access to the world's reserve currency, multi-currency accounts across 76 countries supporting 32 currencies from the dollar to the renminbi to the rupee to the peso, private credit solutions in more than 20 countries, and cards issued to businesses operating across entities and geographies on a single platform. Flex carries state across the boundary every competitor treats as a wall. The card knows what the loan knows. The loan knows what the household knows. The cross-border payment knows what the treasury balance knows. Most platforms in this category are a front-end onto a workflow. Flex is an execution layer with a balance sheet behind it, and those are different companies with different ceilings. Why "Stablecoin Rails" Is Not The Full Story Here The phrase "stablecoin rails" has been flattened into a category label over the past eighteen months, so it is worth showing what the underlying claim actually means. Stablecoins are the rail. The product is a private banking relationship that happens to settle on the cheapest rail available. The clearest way to see this is to consider what a settlement rail is worth to the companies that have gone out and bought one. Stripe paid roughly $1.1 billion for Bridge in February 2025. Mastercard, the second-largest payment network in the world, agreed in March 2026 to acquire BVNK, a London company whose entire product is stablecoin pay-ins, payouts, FX and custody across more than 130 countries, for up to $1.8 billion including a $300 million earn-out. Coinbase had reportedly been close to buying the same company for around $2 billion before talks collapsed in November 2025. Zerohash last raised $104 million at roughly a $1 billion valuation. S&P Global Market Intelligence counted at least fourteen stablecoin-related acquisitions announced across 2025 alone. The world's largest payment companies have concluded, repeatedly and expensively, that this capability is worth between $1 billion and $2 billion to own rather than to rent. Flex now runs it in-house, inside a company the market has priced at $1.2 billion. That is the correct frame for Flex Global, and it is not the frame the announcement uses. Stablecoin settlement is not a revenue line for Flex, because the revenue on a cross-border payment does not sit in the settlement rail; it sits in the FX spread wrapped around it and in the deposit float resting on either side of it. What the rail buys is the customer conversation. The moment a Flex customer signs a supplier in Warsaw or Mexico City, that owner is going to ask somebody how to pay them, and the only strategic question is whether the somebody is Flex or Airwallex or Stripe. Flex is now the one answering, and it is answering at a price the incumbent physically cannot match: fifty basis points against the bank's three hundred, minutes against days, and the customer still keeps eighty per cent of what the bank was taking. Both sides win. That is the definition of a durable product, and it is why a defensive launch reads as an offensive one on the P&L. The Two TAMs? Understanding the Addressable Market Here is where the analysis gets interesting, because Flex's market can be sized two completely different ways, and the gap between them is the whole investment story. Sized conventionally, Flex lives inside business banking and spend management, competing for a budget line against Ramp, Brex, Mercury and a half-dozen point-solution vendors that have collectively raised billions of dollars to chase slivers of the same spend. Respectable, growing, crowded with scale players defending their turf. If Flex is fighting for software-line-item dollars, it is competing on a category map that already has incumbents on every border. But that is not the market Flex is claiming. Its category framing, a private bank scored on the owner's whole financial life rather than on seats sold, points at the pool of value the owner is currently surrendering, and the arithmetic of that pool is an order of magnitude larger than the software budget it appears to be selling against. The evidence is in the take rate, and it is the number that should have led every piece of coverage of this round. Flex moves $10 billion a year and earns a nine-figure revenue against it, which puts its implied take rate somewhere between 1.0 and 2.5 per cent of every dollar crossing the platform. Airwallex, moving $287 billion, earns 0.45 per cent. Mercury, moving $248 billion, earns 0.26 per cent. Flex monetises a dollar four to seven times harder than the global payment platforms it now competes with, and the reason is structural rather than promotional: it earns commercial card interchange on the spend, net interest margin on the deposit, and a lending spread on the credit, against a customer who has attached four products rather than one. That is not a payments business with a banking feature. It is a bank with a payments distribution channel, and the two are valued very differently by everyone except, at the moment, the market. The stablecoin data reinforces the point rather than undermining it, provided you read it honestly. McKinsey and Artemis Analytics stripped trading, internal transfers and automated contract loops out of the on-chain data and found that of roughly $35 trillion in raw annualised stablecoin volume, about $390 billion represents genuine end-user payments, of which business-to-business flows account for $226 billion, or roughly 58 per cent, growing 733 per cent year on year. $226 billion against a global B2B payments market of roughly $160 trillion is a penetration rate of 0.14 per cent. That is not a warning; that is a runway. B2B stablecoin payments are the fastest-growing category in the entire payments industry and they have captured roughly one seventh of one per cent of what they address. Visa's own stablecoin settlement run rate doubled to $7 billion between December 2025 and April 2026, across nine chains, with more than 130 stablecoin-linked card programmes now live in over 50 countries. The rails have arrived. The volume has not. Flex is standing at the base of a curve that is bending, which is precisely where a company wants to be found. There is precedent for this exact manoeuvre. Stripe did not compete inside the payment-processing software market when it launched; it competed for a percentage of online transaction volume, an opportunity orders of magnitude larger than the software budget it appeared to be selling against. The middle-market owner's financial life today looks structurally similar to online payments in 2010, fragmented across a regional bank, a card issuer, an FX broker, a private banker and a lender, with the owner's own assistant stitching the five together at the boundary. Flex is the consolidation of that stack, with the same pricing geometry quietly available to it. The SaaS Playbook: How This Business Actually Compounds For readers building or evaluating financial software, the Flex model rewards a closer look, because almost none of the classic SaaS mechanics apply in their usual form. Pricing. Seat-based pricing is incoherent for a private bank, because there are no seats sitting in it. Revenue arrives as interchange on spend, spread on credit, margin on FX and yield on deposits, which means it scales with the owner's economic activity rather than with headcount. The strategic consequence is non-trivial: the platform gets paid more as the customer's business grows, and the customer's business growing is the thing the platform's credit product is explicitly designed to cause. Net revenue retention. Land-and-expand is built into the underwriting rather than into the sales motion. An owner who takes the card is one data cycle away from a working capital line, one cycle from there to treasury, and one from there to the household. Each product the platform underwrites expands the surface of the relationship without expanding the acquisition cost, which is what best-in-class net revenue retention actually requires, and which is why four products per customer at year three is the metric that determines whether the next round prices this company as infrastructure or as software. The moat. Every dollar Flex lends teaches the underwriting engine something about a segment with no public credit market and thin bureau coverage, in a population that does not file with the SEC and does not appear in any syndicated loan database. That library is generated by doing the work, against borrower behaviour that varies by industry and region, and a competitor cannot shortcut it with a larger foundation model, because the data is not on the public internet and never will be. Layer on the fact that Flex sees the company ledger and the owner's personal ledger simultaneously, which no bank in the world does because no bank in the world is organised to, and the switching cost compounds in both directions at once. That is the honest answer to the "why is this AI-native" question, and it is a data claim rather than a model claim. Airwallex launched an agentic finance platform called T:0 three weeks ago. Ramp raised $750 million at a $44 billion valuation in June substantially on the strength of an accounting agent. Every serious platform in this category ships agents now, which means agents are table stakes. The differentiation is not the agent. It is the proprietary corpus underneath it, and Flex is the only company in the category holding both halves of an owner's balance sheet in the training set. The valuation. At a nine-figure run rate, $1.2 billion implies somewhere between 4.8 and 10 times revenue. Ramp trades at 29 to 44 times revenue. Airwallex trades at roughly 8.5. Mercury trades at roughly 8. Brex sold to Capital One at roughly 5.5. Flex is growing at roughly 200 per cent, which is faster than any of them, and prices below all of them that remain independent. Whichever point of the revenue band you plug into the surface, the answer comes out the same way, which reads less like a stretched round and more like what a preemptive round looks like when a single conviction buyer prices it rather than an auction. Competitive Landscape: Four Rings Around the Same Owner Flex enters a market with incumbents on every side, none of whom currently does what it does in one platform. The first ring is spend management for venture-backed companies. Ramp, at a $44 billion valuation on more than $1 billion of annualised revenue, is the definitional platform in the category and has extended from cards into treasury, procurement and accounting automation. Brex, its rival for eight years, sold to Capital One for $5.15 billion in January 2026. Both are excellent, and both were architected for a company with a finance team and a board, which is precisely what the middle-market owner does not have. The second ring is startup and SMB banking. Mercury, at a $5.2 billion valuation on $650 million of annualised revenue and more than 300,000 customers, has been profitable for four years and is pursuing a national bank charter. Its take rate of 0.26 per cent tells you exactly where it sits: deposits and payments at scale, with credit and the household deliberately left outside the perimeter. The third ring is global cross-border. Airwallex, at an $11 billion valuation on $1.3 billion of annualised revenue, serves 676,000 businesses across more than 85 licences and local rails in 120-plus countries, and has just pivoted its narrative toward autonomous finance and agentic commerce. It is the most formidable company on this map and the one whose corridors Flex Global now overlaps. Its limitation is the mirror image of Ramp's: extraordinary rails, sold overwhelmingly to e-commerce merchants, marketplaces and platform businesses, with no private credit arm and no interest whatsoever in the owner's household. The fourth ring is the incumbent the customer actually uses today, which is a regional bank for the business, an Amex card for the spend, and, if the owner is large enough, a private banker at a wirehouse for the household. They own the relationship and have owned it for thirty years. What they do not have is software, and what they will not do is underwrite a $12 million distribution business and the owner's second home out of the same file. The competitive question is not whether the others can see the customer. It is who gets to the referenceable four-products-per-owner proof first, because in business finance the compounding runs through the relationship rather than through the feature list, and the platform with the deepest attach today holds the structural advantage in the buyer's evaluation tomorrow. Britain, and the Regulatory Detail Everyone Gets Backwards Flex is opening in the United Kingdom, and the regulatory reading here is considerably more favourable than the received wisdom suggests. The UK Interchange Fee Regulation caps consumer credit-card interchange at 0.30 per cent and consumer debit at 0.20 per cent. Most American card fintechs looking at Europe read that sentence, conclude the economics do not survive the Atlantic, and stay home. They stop one line too early, because commercial cards are exempt from the cap. Flex's gateway product is a business card. In the United States, a commercial card-not-present transaction on Visa's corporate and purchasing schedule carries 2.70 per cent interchange. In Britain, commercial interchange is uncapped and, while it prices materially below American levels, it remains a real and unregulated economic engine, which means the pillar that drives the entire attach motion crosses the border substantially intact. That is an uncommon position, and it is less an accident of luck than an accident of architecture: a company that had led with a consumer card would have found roughly 87 per cent of its interchange legislated away at the border. Flex led with the business. The Investor Thesis: Why Halo Fund, Why Now Halo Fund, formally the Halo Experience Company, is a $1 billion vehicle launched in April 2025 by Ryan Smith and Ryan Sweeney. It is hard-capped, expects to make roughly twenty investments over three years; it is a technology fund that uses live sports and entertainment as a distribution surface for its portfolio. Smith owns the Utah Jazz and the Utah Mammoth. Sweeney is a general partner at Accel. The check gets written about as access to an audience. The reading that matters is the biography. Ryan Smith co-founded Qualtrics in Provo, Utah, in 2002, and bootstrapped it for ten years before any Silicon Valley firm would look at it, because it was a profitable software company in a state the Valley's capital allocation model did not recognise as a place where software companies came from. He then sold it to SAP for $8 billion, took it public, and watched it go private again at $12.5 billion. He is, in every respect that matters to this deal, a middle-American operator running a real business with real revenue who was ignored for a decade by the institutions whose entire job was to find him. He is the customer. Ryan Sweeney is the Accel partner who eventually found him. And the rest of Sweeney's book is the part almost nobody covering this deal is going to read correctly, because it is not a fintech book in the conventional sense. It is Atlassian, which is the definitional bottoms-up, multi-product, land-and-expand company in enterprise software, and it is Braintree and Venmo, which is payments consolidation. Sweeney has already run the two plays Flex is running simultaneously, in two separate decades, and he ran them to the two largest outcomes of his career. The Portage check is the specialist's confirmation. Portage led the December Series B and returned for this one, which is a stronger signal than any new name on a cap table, because the fund that has held a board seat for seven months is the fund with the least excuse for being wrong. The 53 Stations check is the one that gives the whole thing away. 53 Stations is a $190 million fund with a single limited partner, and that limited partner is The Pritzker Organization, which is to say one of the oldest operating fortunes in America, built across Hyatt and a portfolio of exactly the kind of unglamorous mid-market industrials that Flex was designed to bank. The investor is the customer archetype. Wellington, meanwhile, is a crossover, and crossovers appear on cap tables when somebody at the table has started thinking about a listing. Now notice who is not here. No a16z. No Sequoia. No Founders Fund. There is no marquee Sand Hill Road lead in this round and there has not been one in any round, and that is not a gap in the cap table; it is the strategy, written down. Flex's customers are in Ohio and Texas and Utah and Warsaw. So, deliberately, is its money. What Has to Go Right Honest analysis requires naming the hard parts, and Flex has three worth naming. The first is the composition of the revenue line. "Nine figures" spans a wide range, and the mix inside it determines whether this company gets valued as software or as specialty finance. Interchange revenue, deposit spread and lending spread are three different assets with three different multiples attached, and the market's willingness to keep paying up will depend on Flex eventually disclosing a net revenue split rather than a gross run rate. The company has every incentive to do it, because on the evidence available the disclosure would help it. The second is the credit book, which is what a $300 million debt facility actually buys. Take a representative structure for a non-bank lender funded by a warehouse line: a blended asset yield in the mid-teens, a cost of funds in the mid-to-high single digits, and servicing and origination costs of roughly two points. On those assumptions, the book carries about four points of loss cushion before the spread goes to zero. For scale, US commercial bank credit-card delinquency was 2.9 per cent in the first quarter of 2026 and charge-offs ran at 3.8 per cent. Flex is not lending to a bank-card population; it is lending to owners with collateral, personal guarantees and audited revenue, and its loss experience should be structurally better. But the underwriting model has been trained through a benign cycle, and it is now being extended into twenty-plus jurisdictions where Flex has no recovery infrastructure and no relationship with the local courts. That is the number the next four quarters will settle. The third is the regulatory exposure attached to settling owner money on stablecoin rails. The GENIUS Act was signed in July 2025, the statutory deadline for regulators to finalise implementing rules falls on 18 July 2026, four days after this announcement, and the Act itself does not take effect until the earlier of 120 days after final rules or 18 January 2027. As of early July the Federal Reserve Board, one of the primary payment stablecoin regulators, had not published its proposed rule, and the Treasury had made no comparability determination for USDT, the largest stablecoin in the world. Any platform settling customer money on rails whose governing statute is still in rulemaking carries a non-trivial exposure that does not exist for platforms confining themselves to fiat. The counter-argument, and it is a strong one, is that the firms waiting for perfect clarity will be integrating in 2027 against a competitor with eighteen months of live corridor data, and in payments the corridor data is the product. Final Thoughts The most valuable financial software categories have always been built where the money already is and nobody has been watching it. Stripe claimed the online transaction. Ramp claimed the corporate card. Mercury claimed the startup deposit. Airwallex claimed the cross-border corridor. The middle-market business owner, the person whose company and household are the same balance sheet expressed twice, has never had an institution built around him. If Flex's thesis holds, the prize is not a software budget line carved out of the spend-management category. It is the entire financial relationship with 350,000 American owners and three million globally, priced at a take rate four to seven times what the global payment platforms extract, compounding through a lending data set no competitor can assemble because no competitor is looking at both halves of the ledger. That is a structurally larger opportunity than the comparable set suggests, and it is the opportunity that explains why an Accel partner with Atlassian and Venmo in his book and a Chicago family office built on mid-market industrials wrote checks into the same round on the same day. Series B1 announcements are easy to make and hard to interpret, but this one comes with an unusually clean test. Either products per customer goes up as the base globalises, and the blended take rate holds while Flex Global scales, or it does not. In a fintech market exhausted by narratives about AI and stablecoins, that kind of falsifiability might be the most valuable thing this funding announcement actually communicates. Don't forget to like and share the story! Vested Interest Disclosure: HackerNoon has reviewed the report for quality, but the claims herein belong to the author. #DYOR.