SoFi: investing in the ability to respond

SignalThe position is taken→Examines how present investment decisions can expand, preserve, narrow or foreclose the positions an organisation can occupy later.

SoFi was built in student finance. In 2020, US policy abruptly weakened much of the reason borrowers had to refinance eligible federal student loans: payments were paused and interest was set to zero. By April 2022, SoFi said refinancing volumes had been running at less than half their pre-Covid level for two years. Yet it was still forecasting adjusted net-revenue growth of about 45 per cent for the year.¹

SoFi had not predicted a three-year disruption to its founding market. But by then, the original product no longer defined everything the company could do. The customer relationship had widened, SoFi had bought technology it already knew as a customer, and it was close to becoming a national bank. The shock exposed something that had been building for years: more than one way to respond.

The relationship outlived the first product

SoFi’s first model combined a borrower relationship, lending capability and somebody else’s capital. In its Stanford pilot, roughly $2 million from 40 alumni funded loans to nearly 100 students. The founders also wanted the relationship to extend beyond the transaction through advice, connections and an alumni community.²

The funding model changed quickly as SoFi scaled, but the relationship proved more durable. The company added mortgages, personal loans, saving, spending and investing. By Q2 2026, SoFi reported 15.8 million members and 24.4 million products, with 51 per cent of new products opened by existing members, up from 35 per cent a year earlier.³ A relationship created around one financial need had become useful when another need appeared.

That matters because the relationship is not only a route to revenue. It keeps SoFi close to changing demand. The strategic value comes when what the company sees from that position changes what it decides to do next.

Learning what mattered enough to own

Galileo was already a substantial payments-infrastructure company when SoFi bought it in 2020 for about $1.2 billion. It was profitable, processing more than $53 billion of annualised payment volume, and serving fintechs including Robinhood, Monzo, Revolut, TransferWise and Varo. SoFi also knew the technology from the inside: SoFi Money was already a Galileo customer.⁴

SoFi was therefore assessing a capability it already used inside its own business. Galileo would continue serving external institutions, while ownership was expected to accelerate SoFi’s own technology roadmap. SoFi was not simply buying a software revenue stream; it was changing its relationship with a capability it had learned to value — from customer to owner — while keeping an external market for the capability itself.

Two years later, SoFi wanted another layer: a modern multi-product banking core and ledger beneath the payments capability. Its relationship with Technisys moved from a possible commercial arrangement, to consideration of a minority investment, to a decision to acquire the company. Anthony Noto, SoFi’s chief executive since 2018, has described that progression as SoFi learning more about what the capability could become when combined with Galileo and SoFi itself.⁵

The roughly $1.1 billion Technisys deal still had to justify the capital committed. SoFi published expected revenue synergies, cost savings and an illustrative mid-teens standalone return. But management was also buying control over infrastructure it believed would support faster product development, more real-time decision-making and products that had not yet been specified.⁵ The investment case therefore covered both what could be modelled and what ownership could make possible next.

A different capability changed the economics

Technology was only one part of the position SoFi was building. It could lend before becoming a bank, using warehouse facilities and capital markets. The national bank charter changed how that lending capability could be funded and what economics were available to it. After SoFi Bank was created in 2022, member deposits could fund more lending directly and loans could be retained where that made sense.⁶

By Q2 2026, deposits had reached $45.5 billion and represented more than 90 per cent of average liabilities. SoFi estimated that its average deposit rate was 156 basis points below comparable warehouse funding, equivalent to roughly $713 million of annualised interest-expense savings.⁶ The national bank charter did not give SoFi the ability to lend. It changed the economics and choices around a lending capability it already had.

Then the context changed again. In 2025, the Office of the Comptroller of the Currency reaffirmed that national banks could undertake specified stablecoin, distributed-ledger and crypto-custody activities and removed an earlier supervisory non-objection requirement.⁷ SoFi had not pursued a charter years earlier because it knew that rule would change. But when it did, the charter sat alongside payments technology, a cloud banking core and a regulated operating environment. A capability acquired for one reason had become relevant to another market.

Capability creates possibilities, not guarantees

The external technology business is the harder test. Galileo and Technisys gave SoFi technology it could use internally and sell externally, but the external market has not developed in a straight line. SoFi recorded goodwill impairments against both businesses in 2023 after concluding that growth had been slower than expected at acquisition. A large customer later transitioned off the platform. In Q2 2026, Technology Platform revenue was $84.5 million, up 13 per cent sequentially but down 23 per cent from a year earlier, with contribution margin at 14 per cent.⁸

Large enterprise customers can take a long time to implement, migrate and reach mature transaction volume. Noto had been making that point since the Technisys acquisition, years before the large customer transitioned off the platform; in Q1 2026 he pointed to 13 partners producing revenue that had produced none a year earlier.⁹ That timing matters when interpreting the current result: it helps explain why new client wins do not immediately replace mature revenue, but the external business still has to rebuild growth and margins.

At the same time, SoFi is using the acquired technology inside its own operations. SoFi Money is being moved onto the SoFi Technology Solutions cloud-native core, and Big Business Banking was built on it. SoFi intends to offer the same underlying capability to other banks and brands.¹⁰ The internal business can therefore become a live operating environment for technology that also has an external market. Internal deployment broadens the return case for the capability. It does not settle the external one.

The market changed the plan

Big Business Banking shows the learning loop in real time. SoFi initially approached crypto marketplaces and market makers looking for partnerships. Noto says the response it heard repeatedly was different: could SoFi itself become their bank and provide API-driven fiat-and-crypto banking? That market feedback changed the proposition. SoFi could respond because it already had the regulated bank and technology core on which Big Business Banking was then built.¹⁰

SoFiUSD sits inside the same story. SoFi has launched the bank-issued stablecoin and Big Business Banking, and is working with Mastercard to enable SoFiUSD as a settlement option across its network.¹⁰¹¹ The activity is real; what remains unproven is its eventual scale and economics. What is already observable is that capabilities bought or obtained for earlier purposes could be mobilised when the market exposed a different need.

The process is still happening. Noto said Big Business Banking, small-business lending, SoFiUSD and the relaunch of crypto were among initiatives that were not in SoFi’s original 2026 plan. Management chose to invest in them during the year rather than convert all of the revenue upside into higher near-term profit.¹² The plan changed because what SoFi was seeing had changed.

More than one way to use the same lending capability

The Loan Platform Business brings the case back to SoFi’s origins. The bank charter did not make external capital obsolete; it gave SoFi another funding route. Through LPB, SoFi can combine borrower demand with its acquisition, pricing, underwriting, origination and servicing capabilities while institutional partners provide the capital. SoFi earns fees without putting every loan on its own balance sheet.¹³

That creates a choice the original company did not have. A loan retained by SoFi Bank can generate interest income over time but consumes SoFi capital and leaves the bank with the credit exposure. A loan originated through LPB can generate fee income while using less of SoFi’s own balance-sheet capital. In Q2 2026, SoFi originated $3.1 billion of personal loans on behalf of third parties or through referrals and reported $143.3 million of adjusted net revenue from LPB.⁸

The echo with 2011 is useful because the company in the middle has changed so much. SoFi again connects borrowers with somebody else’s capital, but now it also has a national bank, deposits, a much larger customer relationship, more sophisticated underwriting and servicing, and technology infrastructure it owns. The value is not a return to the original model. It is having more than one economically useful configuration available when demand appears.

The ability to respond still has to earn a return

SoFi is not a case for owning every capability or turning every investment into a platform. Technology Solutions shows why: some parts of the thesis are still being pressure-tested, while some newer value pools are already producing activity and revenue, but their eventual scale and economics will take time to prove. Adaptability is valuable only if the organisation can convert what it learns into economically worthwhile action.

What makes SoFi useful is the movement between participation, learning and investment. A relationship created for one financial need becomes useful for others. Experience as a technology customer changes the judgement about what is important enough to own. Regulatory permission acquired for banking economics later matters in a different market. External capital reappears in a more sophisticated form because the company has more ways to deploy the same lending capability.

The strategic achievement is not owning the pieces. It is recognising when they can be used differently, finding more valuable ways to combine them and being able to act while the opportunity is still forming. Whether SoFi’s individual bets justify their cost remains open. The capacity to keep changing what the company can do is already visible.

Notes

Footnotes

1. US Department of Education / Federal Student Aid, FY 2023 Annual Report, records that the federal student-loan payment pause began effective 13 March 2020 and lasted 42 months. SoFi’s April 2022 guidance update said refinancing volumes had been below 50 per cent of pre-Covid levels for roughly two years; the latest extension reduced 2022 adjusted net-revenue guidance from $1.57 billion to $1.47 billion and adjusted EBITDA guidance from $180 million to $100 million, while SoFi still expected about 45 per cent adjusted net-revenue growth. Sources: Federal Student Aid — FY 2023 Annual Report; SoFi — April 2022 guidance update. ↩

2. SoFi’s 2012 account describes the Stanford pilot as roughly $2 million funded by 40 alumni for nearly 100 students and presents the alumni/student relationship as broader than financing alone. SoFi was founded in 2011 by Mike Cagney, Dan Macklin, James Finnigan and Ian Brady. Source: SoFi — launch account. ↩

3. SoFi Q2 2026 results report 15.8 million members, 24.4 million products and 51 per cent of new products opened by existing members. Source: SoFi — Q2 2026 earnings release. ↩

4. Galileo’s October 2019 $77 million Series A was its first institutional financing. Galileo described itself as profitable and processing more than $26 billion in annual payments volume. SoFi’s April 2020 acquisition announcement records the $1.2 billion transaction, more than $53 billion of annualised payment volume by March 2020, SoFi Money’s existing integration and the intended internal/external strategic relationship. Sources: Galileo — 2019 Series A; SoFi — Galileo acquisition announcement. ↩

5. SoFi’s February 2022 Technisys acquisition announcement and acquisition-call transcript outline an approximately $1.1 billion all-stock transaction, expected revenue and cost synergies, and an illustrative mid-teens standalone return. Noto describes evaluating alternatives, including build, a commercial relationship and minority investment, before concluding that ownership created more value; he also describes faster innovation, lower cost and future products as part of the rationale. Sources: SoFi — Technisys acquisition announcement; SoFi — Technisys acquisition call transcript. ↩ ↩²

6. SoFi announced the Golden Pacific Bancorp acquisition in 2021 for approximately $22.3 million and planned a $750 million capital contribution. The OCC approved SoFi Bank in January 2022. Q2 2026 results report $45.5 billion of deposits, more than 90 per cent of average liabilities, and SoFi’s estimate of roughly $712.6 million in annualised interest-expense savings versus warehouse funding. Sources: SoFi — Golden Pacific acquisition; OCC — SoFi Bank approval; SoFi — Q2 2026 earnings release. ↩ ↩²

7. The Office of the Comptroller of the Currency’s March 2025 guidance reaffirmed national-bank authority for specified crypto-asset custody, stablecoin and distributed-ledger activities and removed the previous supervisory non-objection requirement. Source: OCC — Interpretive Letter 1183. ↩

8. SoFi’s Q3 2023 filing records $124.5 million of goodwill impairment for Galileo and $122.7 million for Technisys and states that Technology Platform growth had been slower than expected at acquisition. Q2 2026 Technology Platform revenue was $84.5 million, down 23 per cent year on year and up 13 per cent sequentially; contribution margin was 14 per cent. The Q2 release also records $143.3 million of LPB adjusted net revenue and $3.1 billion of third-party personal-loan originations / referrals. Sources: SoFi — Q3 2023 filing; SoFi — Q2 2026 earnings release. ↩ ↩²

9. In SoFi’s Q1 2026 earnings call, Noto said 13 Technology Platform partners were generating revenue that had generated none a year earlier and that this revenue would scale over time rather than arrive instantly. Source: SoFi — Q1 2026 earnings-call transcript. ↩

10. In SoFi’s Q2 2026 earnings call, Noto described Big Business Banking as built on the SoFi Technology Solutions cloud-native core, SoFi Money’s migration to the same core, the intention to offer the platform externally, and the market conversations that led SoFi to provide API-driven fiat-and-crypto banking. Source: SoFi — Q2 2026 earnings-call transcript. ↩ ↩² ↩³

11. SoFi announced SoFiUSD as a fully reserved, bank-issued stablecoin in December 2025. In March 2026, SoFi and Mastercard announced work to enable SoFiUSD settlement, with Galileo clients and their issuing banks among the first expected to be offered the option. These are operating and implementation developments; eventual commercial scale remains unproven. Sources: SoFi — SoFiUSD launch; SoFi and Mastercard — SoFiUSD settlement announcement. ↩

12. In SoFi’s Q2 2026 earnings call, Noto said Big Business Banking, small-business lending, SoFiUSD and the relaunch of crypto were among initiatives not in SoFi’s original 2026 plan. Management said it chose to invest in those opportunities rather than flow all revenue upside through to near-term EBITDA. Source: SoFi — Q2 2026 earnings-call transcript. ↩

13. SoFi’s 2026 Loan Platform Business materials and Q1 2026 earnings call describe LPB as connecting borrower demand with institutional capital, generating fee revenue while retaining servicing rights, and contrast its capital-light economics with loans retained on SoFi’s own balance sheet. Sources: SoFi — LPB agreements; SoFi — Q1 2026 earnings-call transcript. ↩