Meta Muse is the company’s new consumer agent grounded in dense customer data profiles to better understand goals and desires. It can work behind the scenes on your behalf to complete a task, finish a piece of work or schedule an upcoming week. To ease inevitable privacy concerns from some (as they give all of their data to Anthropic and OpenAI with no issue), data and credentials are stored on a secure virtual machine with a separate dedicated agent to verify every single request for risk before approving.

While this all sounds nice, there’s encouragingly some promising early demand news pointing to Meta being on a great path. The early Meta Muse usage is blowing past internal expectations for Meta, with cohorts 10Xing internal company consumption goals. Meta already had the ubiquitous distribution full of trillions of interactions to reimagine and add value to. That makes scaling new offerings a lot easier, but this is still great to see.

And now? Meta has a promising new agent offered as a standalone app that’s shattering their usage goals & in the top 5 on the App Store. They’ve already built this directly into WhatsApp. If things keep going this well, they will be big parts of all other core Meta apps too. So many high margin revenue opportunities will be layered into their existing experience over time.

Need to book an appointment? Send an email? Schedule your life? Or fill out a form? That’s what Muse is for, with intuitive ways to monetize heavier usage via $20 & $100 monthly tiers. These early learnings should also provide highly valuable data to inform that roadmap and give Meta a better chance of shaping it in ways that more quickly appeal to users (the benefit of getting to debut compelling innovation across several different scaled products).

So? Meta boasts crystal clear evidence of using AI to make core app engagement & monetization better. Just listen to any of their last 8 earnings calls and how consistently they’ve used AI to generate more engagement and higher price per ad. On top of that, it’s now building a paid model business as its LLMs shoot up leaderboards & undercutting competition on token costs. And for icing on the cake? It’s successfully scaling the beginning of its army of agents while it leads in a wearables category that could become the new medium for how we engage with this next-gen software. So many ways to make this hefty AI infrastructure spend productive. Meta has a very strange way of showing they can’t monetize or win in AI… I say sarcastically. I can’t help but feel grateful that investors decide to doubt Mark Zuckerberg every 2-3 years. I’m happy to take advantage of those deals, knowing this world-class founder has a better shot at succeeding than virtually anyone else.

News circulated around social media about an AI agent bypassing Shop Pay (Shopify’s checkout accelerator) to buy a product through a Shopify store. In other words, the agent chose to use a vaulted Mastercard on the website instead of going through Shop Pay. The ironic thing is that people make these decisions all the time. Checkout options have many convenient choices from various checkout accelerators with pre-loaded information like the one in question. It’s not just Shop Pay. Checkout is hyper competitive and always has been. Also note two important things:

Luckily, Shopify doesn’t need remotely close to 100% market share, and some of the same ingredients that helped it lead with people are still relevant in the agentic commerce era. While agents can more easily fill out info, these agents also occasionally hallucinate and fail as they collect and enter needed information. Every failure means a lost sale for the merchant and a lost positive outcome for any potential agent vendor to monetize. A structured, authenticated, one-shot checkout is the more reliable path for a machine for the same reasons why it’s more convenient for a person. That’s still what is best for conversion rates and Shop Pay should do fine amid that reality.

World-class fraud systems and a massive base of pre-validated identities in the scaled Shopify ecosystem are also needed assets for agents to most effectively do work. Agents to a merchant are a stranger, without familiar usage history or patterns. This makes digital wallets an important tool for confirming there’s actually a human tied to this agent request, which is also good news for wallets representing large batches of people like Shop Pay.

It’s also important to point out that the companies building the agents that people are worried about are doing so with Shopify. Shopify was named a partner for Anthropic’s commerce agents launch, which directly uses Shopify Catalog just like OpenAI does.  These two companies know that shorter workflows mean cheaper costs and while the savings per transaction is tiny, that adds up at global commerce scale. This naturally motivates them to want to work with the checkout providers that will enable the lowest hallucination rates thanks to everything already discussed.

I think SHOP is going to keep doing very well in checkout, but also believe the plan B is acceptable too. The unit economics associated with a shopper/agent using Shop Pay to make an order on a Shopify merchant site are better than anything else. The company can use that advantage to fund rewards and incentives, absorb a modest margin hit, and make Shop Pay the option an agent optimizing on a consumer's behalf will rationally pick. They can become the best option on Shopify-hosted websites because they own more of the value chain, command more of the margin and have more flexibility to share that with consumers.

The biggest risk provided here is that customer relationships go away when agents are doing all of the shopping. Even if Shop Pay is used and a Shopify merchant is selected, a customer doesn’t have as good of a chance of knowing that if they’re not spearheading the transaction. While that may be true, I do think people naturally enjoy shopping for things and will keep doing it. I don’t think golf courses should be worried that people will stop golfing because robots can do it for them. I don’t think people who like to shop will stop doing it because agents can do it for them either. Maybe sometimes for the more mundane or routine purchases, but certainly not for everything (or close to it). I do think these people will increasingly use chatbots to sharpen discovery. I just don’t think that threatens Shopify’s future.

And most importantly, Shopify’s most valuable asset and biggest reason for AI companies all needing to go through them is their massive base of merchant data. They have 14% of U.S. e-commerce market share (and rising) and the mountain of product listings and descriptions that go alongside it. That is their gold.

Outside of checkout-specific defensibility:

They have world-class assets that AI is heavily reliant on and give Shopify a great chance to play a big role in agentic commerce. Their catalog offering is best-in-class in ensuring accurate, current, granularly described product listings can be pushed to chatbots and agents. The higher quality and accuracy of these listings are naturally favored by agents while the large and growing merchant market share Shopify enjoys means these agents will need to go through the stores it powers to find the products its people want.

And finally, the army of small brands on Shopify make up a sizable portion of its volume, while successful new business creation is a vital cog in its long-term growth engine. For these smaller companies, keyword marketing budgets are a main bottleneck standing in the way of reaching more people and connecting great products with interested consumers. In the age of AI, that bottleneck shifts. These agents can easily find everything and are motivated to find the absolute best option due to being graded on outcomes. There will be sponsored listings that pop up from this, but that’s an obvious extension for Shopify’s advertising suite and another easy performance marketing outlet it can pursue for its merchants.

Quick Take:

All in all, I think the commerce landscape will keep evolving. I think channel usage will keep changing and loud risks will keep emerging. I just think Shopify will be the company setting the innovation curve, staying one step ahead of everyone else and remaining a big piece of the overall commerce space. As long as Tobias Lütke is leading this roadmap, Shopify should continue to do very well. This news does not change my feelings on the company.

No Capital Raises:

Lapointe reiterated that “under the current operating plan” SoFi does not see any need for capital raises “at this point in time.” Some were worried with that language, but leadership can’t promise no capital raises ever again. No matter how responsibly they manage their balance sheet, how affluent their typical borrower is or how effective they are in pricing risk, they are a creditor tied to economic cycles. There are macroeconomic possibilities that are highly unlikely, but would still force many, many institutions to seek external funding and shore up their balance sheets. It would not be wise to pigeonhole themselves with a promise no bank could possibly know if they can keep. 

And also one more reminder about the fact that SoFi’s capital raises did not hurt net income or tangible book value per share. All of them allowed SoFi to retire enough highly expensive debt to pocket interest expense savings needed to at least (or in some cases more than) offset share count growth. The reason we all can’t stand equity raises is because it eats into the profit each share is entitled to. That didn’t happen in this case. These decisions haven’t been harmful for shareholders like they normally are because of the unique nature of SoFi’s previous capital structure and it earning a bank charter. That charter has allowed it to shift most of its funding to much cheaper deposits. Now that this balance sheet optimization process is largely done, future capital raises are even less likely in my mind. There aren’t  additional opportunities to raise without hurting profit per share, which SoFi hasn’t shown a willingness to do under Noto’s leadership.

Guidance Notes:

Lapointe confirmed what most of us have been thinking. Because SoFi is now assuming 1-2 rate hikes in its guidance, that not coming to fruition would mean potential upside to those forecasts. They don’t need significant brightening in any trends or successful product launches or any other kind of positive surprise to hit their current 2026 forecast even if rates actually are hiked twice. For 2028, Lapointe was bluntly asked if SoFi remains confident in 30% revenue growth through 2028. Here’s what he had to say:

“We feel confident. We have large and mature businesses that are growing extremely well. At the same time, we have less mature businesses, like I said at the top of the hour, that are starting to scale but not contributing meaningfully. We have strong confidence in our ability to grow revenue.”

Finally on the previously offered long-term 25%-30% return on tangible common equity (ROTCE) goal, I think SoFi could get there quickly if it wanted to. There’s a clear path if it chooses to prioritize near-term profitability over long-term compounding and building the best possible business that it can. That’s just not the correct decision for maximizing shareholder value, so it’s not the decision they’ll make. They will continue to pursue growth with that 30% incremental EBITDA margin target, striking a compelling balance between top-line success and still solid margin expansion.

Remaining Personal Loan Runway:

There’s ample opportunity to refinance a boatload more 25% APY credit card debt to 12% personal loans for prime borrowers. Creditors have no incentive to refinance this highly profitable product for customers. SoFi can easily do it in a highly profitable manner. That has yielded explosive market share gains for years, and they don’t think they’re done yet. That includes their prime credit bread and butter. While there has been a push from SoFi into lower credit quality buckets to cater to some loan platform business (LPB) capital market partners, Lapointe explicitly said SoFi doesn’t need to do this to find growth. There is plenty of opportunity remaining among prime borrowers… whether that’s from refinancing or taking share from charter-less digital lenders with inferior cost structures and an inability to sustainably compete on price.

Going forward, alongside ongoing personal loan growth, LPB volume will continue to feature more small and medium business (SMB) demand as well as mortgages. They have significant committed capacity from partners and are ramping these businesses meaningfully enough to begin materially fulfilling all of it. Lapointe called this the next leg of LPB growth, but again was careful to say the personal loan side still had plenty of room to run. When considering all of this, SoFi expects LPB volume growth to accelerate through the second half of the year. That likely also is tied to just wrapping up two quarters of hefty balance sheet originations, as SoFi took advantage of its large capital ratio cushions and locked in multiple years of visible, margin-rich revenue to gain confidence in funding its product roadmap.

A Quote on the Current Health of SoFi’s Members:

"Our members are very healthy and performing in line with expectations... everything's performing in line to better than expected from a consumer credit perspective. We're also seeing really good spend behavior across our entire membership. We reported $28B in annualized spend across our debit and credit products, and we don't see that slowing down here in Q3. Overall, our consumer remains extremely healthy, spend behavior is really robust, and losses and delinquencies are performing in line to better than expected." — CFO Chris Lapointe

Big Business Banking & The Tech Platform:

SoFi was pulled into Big Business Banking (BBB) by large financial services players seeking a nationally chartered bank that could handle their fiat and crypto payment flows on a single regulated platform. Providing this unified capability opens two large opportunities for the company. It should attract a lot more deposits to juice net interest income. It should also provide several fee-based revenue streams related to the 24/7, lightning-fast and relatively cheap SoFi Exchange Network (SEN).

It's no secret that the Tech Platform has been the most disappointing piece of this business. Everywhere else you look, trends look elite and then some. I think this provides SoFi with another avenue for earning a piece of the enterprise banking opportunity. This time, it's from a payment settlement point of view, sitting in the flow as the bank and stablecoin issuer rather than just selling the APIs outright. And since SoFi Tech Solutions supplies the  infrastructure behind Big Business Banking, its success provides more proof of concept to hopefully create a catalyst for faster growth than what it has delivered over the last couple years. The overall company continues to remain in 30% top-line compounding mode while this segment underwhelms. It’s exciting to think what that could do to the overall picture.

As a reminder, the cost savings SoFi’s consumer-facing business enjoys from owning its tech stack provides a layer of differentiation in a highly competitive banking industry that I find highly compelling. It helps build a sustainable moat in a place where that is rare. That makes the Galileo and Technisys purchases worth it to me regardless of enterprise sales momentum. But still, I’d love to see that part of the business doing better, and it sounds like they think that will happen in 2027. We’ll see.

More Notes: