STABO.io

Money Behind AI #05: CoreWeave Rebounds 8%. Is the Neocloud Shakeout Just Beginning?

It has been a wild month for AI infrastructure stocks.

CoreWeave, one of the best-known GPU cloud providers, lost more than 30% of its share value in October before rebounding nearly 8% on Tuesday. Its peers Nebius and Applied Digital also rallied sharply.

So, what is making investors nervous?

The AI boom is expensive. Very expensive.

Unlike traditional software companies, neocloud providers such as CoreWeave, Nebius, Lambda and Crusoe need enormous amounts of capital to purchase GPUs, build data centres and secure electricity before they can generate revenue.

CoreWeave alone expects to spend between $35 billion and $39 billion on capital expenditure in 2026. It also reported $104.2 billion in contracted revenue backlog as of June.

That sounds promising. But much of the infrastructure is financed through debt, and the returns depend on sustained demand for AI computing capacity.

Not everyone is convinced the economics will work.

Research firm Gartner has warned that as many as 80% of neocloud providers could cease operating independently by 2030.

Elon Musk added to the debate this week, predicting significant consolidation and failures across the sector.

Yet the market bounced back on Tuesday, supported by renewed confidence in AI demand and spending.

Perhaps both sides have a point.

AI computing demand could continue growing rapidly, while many of the companies supplying that capacity struggle to survive.

The internet boom created enormous demand for infrastructure, but not every infrastructure provider became a winner.

Could AI be heading towards a similar consolidation?

The next chapter of the AI boom may be less about who can build the most GPU capacity, and more about who can build it profitably.

#MoneyBehindAI #CoreWeave #Neocloud #AIInfrastructure #ArtificialIntelligence

Money Behind AI #04: Agents Are Becoming Buyers

Payment systems were built around a person approving each purchase. AI agents now spend money as part of finishing a task, and every step they take can carry a price.

Agents Turn Tasks into Transactions

Consider a small Hong Kong AI company running a research agent for a logistics customer in Germany. Each morning the agent buys shipping data at USD 0.50 a query, translates supplier documents at USD 2 each and rents extra compute at USD 4 an hour.

Over a day that can add up to around 2,000 payments to a dozen suppliers in four or five countries. Each one needs authorisation, settlement and a record someone can reconcile.

When software does the buying, there are more payments and each one is smaller.

Small Payments Expose Fixed Costs

Imagine each USD 0.50 query carries a fixed fee of USD 0.30. More than half the purchase goes on moving the money.

Suppliers usually respond with subscriptions or minimum top-ups of USD 500. When an agent finds a useful new data source mid-task, it waits while someone opens and funds an account. Converting between euros, US dollars and yen adds a little more time and cost to every payment.

Permission Becomes Part of the Payment

Finance teams will want to know who approved each purchase, so an agent’s spending needs rules set in advance. Those rules might be a daily limit of USD 1,000, a list of approved suppliers, a USD 10 cap per transaction and a log of the task each payment served. Wallet and payment infrastructure providers are building tools that keep these controls alongside the funds.

An agent’s spending limit is a business policy written in code, and the payment rail has to be able to read it.

Payment Rails Are Adapting to Machine Speed

Card networks have announced tokenised credentials for agents. Coinbase’s open x402 protocol, launched in May 2025, revives the web’s long-unused “402 Payment Required” status code. With it, an API can quote a price and an agent can pay in stablecoins within the same request, without opening an account first. Bryce Ferguson of Turnkey recently described agents as looking for low-friction ways to complete tasks. Local regulation still applies, and suppliers still need to convert funds into their home currency.

For our Hong Kong company, the likely answer is a mix: bank or card payments for large monthly contracts, and an option such as x402 for the long tail of small, pay-per-query purchases.

Payment infrastructure that settles small amounts quickly, carries spending rules and reconciles cleanly will shape how much work companies trust their agents to do.

That is the story behind Money Behind AI.

#AI #Fintech #CrossBorderPayments #Stablecoins #SaaS #GlobalBusiness #X402

This week’s most useful stablecoin launch may be a risk score

On 19 October at Money20/20 in Las Vegas, Mastercard introduced Stablecoin Security. The service gives financial institutions a risk score on a stablecoin payment before it settles. It combines Mastercard’s payment intelligence with blockchain analytics, and checks where the funds came from, whether they touch sanctioned or restricted parties, and whether there are signs of a scam. The score is delivered through an API, so it can feed the fraud and compliance systems a bank already runs.

Settlement on a blockchain can take seconds. Approval inside a regulated institution usually takes longer, because someone has to be comfortable with where the money has been. On public ledgers that question is often complicated: a BIS study cited by Mastercard found that nearly 60% of stablecoin transfers in 2025 were complex, multi-step transactions.

A score before settlement gives compliance teams evidence they can review, record and act on, within the same workflows they use for card and bank payments. The decision stays with the institution, which now has more information to base it on.

The service launches first in North America, with Borderless.xyz and Keyrails as early partners. Mastercard expects worldwide commercial availability in 2027.

For many banks and corporate treasurers, the path from testing stablecoins to using them in production runs through the compliance team. Tools like this make that path shorter.

#stablecoins #crossborderpayments #fintech #compliance #stabo

STABO is helping the platform settle creator payouts

Most conversations about cross-border payments focus on moving large sums.

The more revealing test is often a small one.

Moving USD 5.

When the Fee Is the Payment

Consider an influencer platform managing hundreds of micro-creators across Nigeria, South Africa, India, Brazil and the Middle East. Many are paid USD 5–10 per post.

Bank and card rails carry fixed costs per transaction. At USD 5,000 these barely register. At USD 5 they can swallow the payment.

Assume an international transfer fee of around USD 5. Paying 300 creators USD 5 each means USD 1,500 of earnings, and another USD 1,500 in fees.

Even a USD 0.50 fixed charge is 10% of a USD 5 payment, before currency conversion or transfers that can take several business days.

The Workaround Shapes the Marketplace

So platforms set minimum payouts or minimum campaign sizes. Creators must build up a balance before they are paid.

It solves the fee problem. But it quietly changes who gets hired.

A brand ends up working with the same ten creators again and again, because everyone needs enough volume to justify a payout.

Consider the alternative: one hundred creators posting once, instead of ten creators posting ten times.

Same budget. Very different reach.

More voices, more communities, and much earlier signals about which creators actually perform.

The payout threshold was designed to manage fees. It ends up shaping the marketplace.

When Small Payments Become Cheap

If a USD 5 payment can move for a small fraction of its value, the minimum becomes optional. A creator can do one USD 5 campaign, get paid, and leave.

Stablecoins are one potential alternative. They can make small cross-border payouts cheaper and quicker to settle.

They are not a universal solution. Creators still need a reliable way to convert into local currency, and regulation, identity checks and tax obligations vary by country.

STABO is currently working with one such influencer platform on creator payouts.

The exact economics will differ by platform, route and provider.

Cheaper payments don’t just reduce the cost of a transaction. They change which transactions, and which business models, are possible in the first place.

#AI #Fintech #CrossBorderPayments #Stablecoins #SaaS #GlobalBusiness

Money Behind AI #03: Why AI Companies Need Better Treasury Infrastructure

Treasury used to be simple for most software companies.

Money came in, salaries and hosting got paid, someone checked the bank balance. Done.

AI companies do not have it that easy.

They can sell globally from day one, but their biggest expenses—model APIs, GPU compute and cloud infrastructure—are often denominated in USD. Meanwhile, revenue may arrive in BRL, EUR, HKD or whatever currency their customers use.

So the money is there. It is just not always in the right currency, account, or place at the right time.

A normal week for an AI company

Imagine a Hong Kong AI startup serving customers across Brazil, Germany and Southeast Asia.

Customers pay in different currencies, through different payment methods, and on different settlement schedules. But the company still needs to pay its model providers and compute suppliers in USD.

Finance has to collect the funds, convert currencies, manage FX costs and make sure everything arrives before supplier payments are due.

None of this is new. Traditional international businesses have dealt with the same problems for decades.

What makes AI different is how quickly these challenges can emerge.

A small AI startup can serve customers in dozens of countries without having offices in any of them. Its compute expenses can rise directly with usage, and some suppliers require prepaid credits or upfront commitments.

More customers mean more revenue, but also more immediate demand for liquidity.

A company can be growing rapidly and still struggle to pay its next compute bill on time.

What treasury actually needs to solve

In practice, it comes down to four things:

Collections: Customers should be able to pay easily, without the company opening bank accounts everywhere.

Visibility: The team should see what has arrived, what is pending, and where each balance sits.

Liquidity: Available funds should move into the right currency and account before a supplier bill is due.

Reconciliation: Payments should match invoices, even when fees or FX conversions change the final amount.

Many smaller companies still handle these through separate bank portals, spreadsheets and manual transfers.

That approach becomes harder to sustain when a business sells globally, incurs usage-based costs and needs to manage multiple currencies at once.

The gap is infrastructure

Large multinational companies have managed multi-currency treasury for years. They have dedicated teams, banking relationships and systems built for it.

Now much smaller AI companies are encountering similar challenges far earlier in their growth.

What they need is not just another bank account.

They need collections, balances, currency conversion and settlement to work together, so they can understand and use their money as one position rather than a collection of disconnected accounts.

Stablecoins can help, particularly for cross-border B2B payments where both sides can access compliant digital-asset rails. They can make certain transfers faster and reduce dependence on traditional banking hours.

But stablecoins are not the entire solution. Regulation, fiat conversion, supplier payment preferences and genuine working-capital gaps still matter.

At STABO, we're building infrastructure that connects fiat and stablecoin payment rails—helping businesses collect globally, manage balances across currencies and settle funds where they're needed.

Because for an AI company, growth is not just about generating more revenue.

It's also about making sure the money can keep up with the business.

#AI #Fintech #CrossBorderPayments #Stablecoins #SaaS #GlobalBusiness