How Developers Pay for Cloud, GPUs and Hosting With Crypto (2026 Data)

Every guide to "paying for a VPS with crypto" lists the same three hosts that take Bitcoin at checkout. That is not how most developers actually do it. The bigger pattern, visible only from the card side, is developers loading crypto onto a prepaid Visa or Mastercard and paying the vendors that will never add a crypto button: Vercel, Cloudflare, RunPod, Railway, AWS.

This post is built on that card-side view. We looked at 12 months of anonymized, aggregated SolCard purchases at 25 developer-infrastructure vendors, about 32,000 settled transactions from roughly 5,000 cards. No accounts, no identities, just merchant names, transaction counts and ticket sizes. Three things stood out: GPU rental is already a fifth of developer infra spend, it bills far more often than anything else, and the whole category doubled between the first and second half of the period.

Card networks do not attach a clean category to a merchant, so we grouped vendors by hand into seven lanes. Shares are of the transactions we could attribute to one of the 25 vendors; long-tail hosts we could not identify are excluded.

Hosting is the widest lane by a distance: more cards touch a VPS or shared-hosting vendor than any other kind of infrastructure, and Hostinger alone is the single most-used vendor in the dataset. That lines up with who funds a card with crypto in the first place: freelancers and small teams in markets where a $5 to $40 monthly hosting bill is the hard part, not the code.

The number that surprised us is in the fourth column. A hosting customer makes about four purchases a year. A GPU-compute customer makes ten, and at the largest vendor in the lane the figure is closer to fourteen.

This is a billing-model artefact, not a spending-habits one. Vast.ai and RunPod meter by the hour and run on prepaid balances, so a developer training a model or running inference tops up in small increments, over and over, rather than paying one invoice at month end. Each top-up is a card authorization. For anyone funding that card with crypto, it means the card has to clear a stream of small charges reliably, not one big one.

The lane is also young. GPU compute did not exist as a category in our 90-day spending breakdown from July; twelve months of data show it at a fifth of all developer infra purchases, second only to hosting.

At the other end of the table, AWS, Google Cloud and Azure account for under 5% of transactions but carry an average ticket around $319, roughly seven times the hosting average. These are monthly usage invoices, billed once, for real workloads.

That combination, few charges and large ones, is exactly the profile a prepaid card is worst at. A $5,000 monthly load limit is generous for a Vercel Pro seat; it is a ceiling for a production AWS bill. Developers on hyperscalers are the ones most likely to keep a crypto card as a secondary payment method for one account rather than the primary rail.

Ranked by transaction count over the 12 months. Distinct cards are shown so you can separate "many small charges" from "many people".

Cloudflare is the interesting one: fourth by transactions but second by reach. It is the vendor most developers touch at least once, usually for a domain or a $20 Pro plan, and then rarely again. Vast.ai is the mirror image: a small group of cards generating a very high volume of charges.

Two Solana-native names, Helius and Alchemy, sit inside the top 25 with QuickNode. Given that most crypto-card deposits arrive on Solana, it is not surprising that developers building on the chain pay their RPC provider with the same balance.

Comparing the two halves of the 12-month window, developer-infrastructure purchases on crypto cards grew 107%, from roughly 10,400 transactions in the first six months to 21,500 in the second. Growth was broad, but not even: GPU compute grew more than fourfold, blockchain RPC, app platforms and hyperscalers roughly doubled, hosting grew by half, and GitHub barely moved.

Some of that is SolCard growing. Most of it, judging by the mix, is the AI build-out reaching individual developers: more people renting GPUs by the hour, more people deploying agents on Railway or Vercel, more people paying for inference through OpenRouter or Cursor. Those are the same people who often get paid in USDC and would rather not route it through a bank to pay a $20 invoice.

• Check the billing model before you pick a card. Metered vendors (Vast.ai, RunPod, most inference APIs) generate many small authorizations. A card that declines on the tenth charge is worse than one with a slightly higher fee. Keep the balance topped above your typical burn.
• Know the load limit. On SolCard's virtual card the limit is $5,000 per month, with a $10 issuance fee and 5% per top-up. That covers any hosting, platform or GPU bill in this dataset comfortably; it does not cover a large hyperscaler account. Verified users on the Platinum tier pay 0% on top-ups.
• Watch for prepaid-card policies. GitHub is the smallest lane here, and its billing documentation states that prepaid cards are not accepted. Read each vendor's payment page before you rely on any card for a production account.
• One card, one line item. Hosting, domains and app platforms are all small, recurring, card-billed charges. Paying them from one crypto-funded balance gives you a single expense line and one balance to watch, which is most of the appeal.

If you want the step-by-step for a specific vendor, start with how to pay with crypto, then the vendor guides for Hostinger, Cursor and the AI tools developers pay for most.

• Source: anonymized, aggregated SolCard card transactions, settled purchases only, over the 12 months to September 2026. No user identifiers, account data or individual transactions were used; cells with fewer than 25 transactions were dropped.
• Vendor matching: merchant names were matched to 25 known infrastructure vendors by pattern. Hosts we could not identify are not counted, so shares are of identified infra spend, not of all developer spending.
• Tickets: average ticket is computed on USD-denominated transactions only, to avoid mixing currencies.
• Sample: about 32,000 transactions from roughly 5,000 cards. Large enough to rank lanes and vendors; not a market-share estimate for the vendors themselves.
• Scope: this reflects one crypto card's customers, who skew toward freelancers and small teams in markets with limited card access. A corporate cloud bill paid by a finance team looks nothing like this.

Only a handful of hosts accept crypto at checkout directly. The common route is a crypto-funded prepaid Visa or Mastercard, which works at any vendor that takes cards: in our data that includes Hostinger, Hetzner, Contabo, OVH, DigitalOcean, Vultr and Linode.

Yes, and it is one of the fastest-growing uses of crypto cards. Vast.ai and RunPod bill by the hour from a prepaid balance, so a card funded with USDC or SOL is topped up in small increments; GPU-compute customers averaged about ten purchases a year in this dataset.

By transactions: Hostinger, Vast.ai, Vercel, Cloudflare and GoDaddy. By number of people: Hostinger, Cloudflare, Vercel, Railway and GoDaddy.

Not directly. AWS, Google Cloud and Azure invoice a card monthly; a crypto-funded card can be that card, but the average ticket in our data is around $319 and a prepaid card's monthly load limit is the practical constraint.

GitHub does not accept cryptocurrency, and its billing documentation states that prepaid credit and debit cards are not accepted as a payment method. Check GitHub's current policy before relying on any prepaid card for a Copilot or Pro subscription.

Yes. It is first-party, anonymized, aggregated SolCard transaction data covering 12 months, not a survey or a market model. See the methodology section for what was included and what was not.