• AI visibility

Guide · AI search

How can a cloud provider win enterprise deals when buyers ask AI first?

By being the provider AI answers name when architects ask about workloads, costs, sovereignty and migration, and by making the proof behind that answer easy to check. Cloud is a concentrated market where three hyperscalers hold most of the spend, so an AI answer that reaches for the default names is a real risk for everyone else. It is also an opening: buyers now ask much more specific questions than “which cloud is best,” and specific questions have specific answers.

By Underneath Team Guide Updated

The short version

  1. Enterprise spending on cloud infrastructure services reached $143 billion in the second quarter of 2026, and Amazon, Microsoft and Google held 28%, 20% and 15% of it, according to Synergy Research Group (opens in a new tab).
  2. Customers are large and committed: in Flexera’s 2026 survey (opens in a new tab), 76% of large enterprises spent over $5 million a month on cloud, and Canalys (opens in a new tab) counts more than $360 billion of multi-year commitments to the top three providers.
  3. Those commitments steer software buying too: Canalys expects sales through hyperscaler marketplaces to reach $85 billion by 2028, up from $16 billion in 2023.
  4. There is room for challengers: Gartner (opens in a new tab) expects sovereign cloud spending of $80 billion in 2026 to shift 20% of current workloads from global to local providers.
  5. A new kind of buyer is arriving: at Vercel, fewer than 3% of deployments were triggered by AI coding agents at the start of 2026, against more than half by June, eCommerceNews reported (opens in a new tab).

Who signs for cloud infrastructure, and how much is one account worth?

A platform team, finance and procurement buy it together, and one customer can be worth millions a month.

The market is big and still accelerating. Synergy estimates that quarterly cloud infrastructure revenue grew 43% year on year in the second quarter of 2026, the highest rate in eight years, with the top three holding 67% of the public cloud segment. Gartner (opens in a new tab) forecast $723.4 billion of public cloud spending in 2025, growth of 21.5%.

A cloud contract is rarely signed off by one person. In Flexera’s 2026 survey, 71% of organizations ran a cloud center of excellence and 63% had a FinOps team that manages cloud costs. In practice that means:

  • Architects and platform engineers decide what is technically possible and what they will support.
  • FinOps and finance decide what is affordable; 85% of Flexera’s respondents called managing cloud spend a key challenge.
  • Security and compliance decide what is allowed, which increasingly includes where data may live.
  • Procurement turns the decision into a multi-year commitment.

Most enterprises already run more than one cloud. Flexera found AWS (83%) and Azure (79%) in active use for enterprise workloads, and 73% of organizations operating hybrid environments. Gartner predicts 90% of organizations will adopt a hybrid approach through 2027. So the commercial question is less “which cloud” than “which cloud for this workload.”

What a customer is worth shows up in public filings. DigitalOcean, a challenger that serves developers and growing companies, reported $1,125 million of annual run-rate revenue in its second quarter of 2026 results (opens in a new tab). Its customers spending over $1 million a year grew 73%, it signed its first nine-figure annual commitments, and its weighted average contract life rose from 1.6 years to over 3 years. A won workload grows, and the contract around it gets longer.

Cloud software vendors, the second group this article covers, sell into the same budgets. Canalys reports that enterprises increasingly spend their existing cloud commitments on third-party software bought through hyperscaler marketplaces, and that CrowdStrike and Snowflake were among the first to claim $1 billion of cumulative marketplace sales. Data platforms such as Snowflake have their own buying path, covered in how data platforms get shortlisted.

Where does AI already sit in a cloud buying decision?

At the research stage, in Google’s results, and increasingly inside the coding tools that provision infrastructure.

No public survey isolates cloud buyers. The closest evidence covers business software: in G2’s August 2025 survey (opens in a new tab) of more than 1,000 software buyers, 87% said AI chatbots were changing how they research. G2 sells visibility on its review platform, so it has an interest in that finding.

Google’s own results already answer most technology searches with AI. The B2B software and technology keywords in our study of 1,248 US searches, the group closest to cloud platforms, showed an AI Overview on 96.0% of searches, more often than any of the other seven industries.

The newer development is that software, not only people, now makes infrastructure choices. At Vercel, a hosting platform, the share of deployments triggered by AI coding agents rose from fewer than 3% to more than half in about six months, according to eCommerceNews’s report of the company’s figures. When an agent sets up a project, the database, hosting and services it reaches for become the default. Database makers already feel this, as our guide to how AI helps pick the database explains. A reasonable expectation is that the defaults agents pick will matter more for developer-led cloud products each quarter. No published study yet shows how agents choose.

What do architects and FinOps teams ask AI assistants about cloud?

Mostly workload questions about architecture, cost, location and migration, rarely a general “which cloud” question. We wrote the sample prompts below to show the shape of those workload questions; they are not logged queries from real cloud buyers.

StageIllustrative prompt
Architecture“What is the best way to run GPU inference for a mid-sized fintech without a hyperscaler contract?”
Alternatives“Alternatives to AWS for a SaaS company that wants simpler pricing and EU hosting”
Cost“Is it cheaper to run Kubernetes on Azure, Google Cloud or a smaller provider for 200 nodes?”
Sovereignty“Which cloud providers offer sovereign cloud regions in Germany that meet government requirements?”
Migration“How hard is it to move from Heroku to a container platform, and which ones do teams choose?”
Marketplace“Can I buy this observability tool through AWS Marketplace and count it against our commitment?”

The answer to each of these is rarely written from one page. For a GPU or Kubernetes question, Google says AI Overviews and AI Mode may use a “query fan-out” technique (opens in a new tab), firing off multiple related searches across subtopics and data sources. OpenAI’s help pages say ChatGPT search rewrites a question (opens in a new tab) into one or more targeted queries and passes them to its search partners. An architecture question, we infer, sets off searches for documentation, pricing, benchmarks and practitioner discussion all at once.

How does an AI answer turn into cloud consumption and commitments?

Through two paths: a developer signup that grows into a commitment, and an enterprise shortlist that ends in a contract.

The developer path. An engineer or a coding agent asks how to run something, gets an answer that names a platform, signs up and starts a small workload. If it works, usage grows. DigitalOcean’s numbers show where that leads: revenue from customers spending $1 million or more a year now makes up 23% of its total and grew 214%. The AI answer sits at the very first step, where the platform is chosen.

The enterprise path. An architect asks a workload question, the answer names a few providers or tools, the team runs a proof of concept, and procurement folds the winner into a multi-year agreement. For cloud software vendors there is a third step: the purchase is often routed through a hyperscaler marketplace so it counts against an existing commitment. Canalys expects more than 50% of marketplace sales to flow through channel partners by 2027, so a partner may be the one asking the assistant.

On both paths the revenue rarely shows up as a click from an AI answer. It arrives later as a signup, a sales conversation or a marketplace order, which is why attribution is hard; we cover this in what lost clicks mean for pipeline.

Why does an AI assistant name one cloud provider and skip another?

No platform documents how it picks providers; studies show strong defaults toward famous names that specific evidence can overcome.

Documented by the platforms. Both Google and OpenAI say their AI answers search the web and link what they used. Neither says how it decides which hyperscaler, neocloud or cloud tool ends up in the answer.

Observed in studies. Assistants lean toward names they know. In controlled tests by Chu and Hou (opens in a new tab), three AI models always picked a real brand over invented ones when products were otherwise identical, but a rating edge of just 0.075 stars was enough for an unknown name to win half the time. In our brand study of four assistants, a brand named on ten times as many independent sites had 4.7 times the odds of being recommended. Our hidden-searches study caught ChatGPT searching for a named publication, ranking or award in 43.8% of its answers and for reviews in 46.2%. What that pull toward famous names means for challengers to the big three is the subject of do AI assistants favor big brands.

What cloud buyers check. Architects verify claims in public: documentation, pricing calculators, status pages, compliance attestations, region lists, benchmarks and practitioner forums. Flexera found 53% of cloud leaders named security and compliance as their top challenge for cloud-based AI work. Security vendors meet that scrutiny head on, as our guide to cloud security shortlists in AI answers shows.

Our inference. For a challenger, the advantage that wins an evaluation (a lower price for a workload, a region, a certification, a simpler pricing model) is the same thing an assistant needs to see stated plainly and confirmed elsewhere. A reasonable expectation is that a specific, verifiable difference helps a challenger more than general claims to be “developer-friendly.” Nobody has yet tested this for cloud providers.

What does a cloud provider lose when AI answers leave it out?

A missed workload can mean a missed multi-year commitment, though no study has measured the loss directly.

  • Commitments lock in choices. With more than $360 billion of multi-year commitments to the top three, Canalys notes that customers increasingly spend that money on software bought through those providers’ marketplaces. A vendor not considered when a workload is placed may wait years for the next opening, we infer.
  • Sovereignty is a new opening, with a deadline. Gartner forecasts sovereign cloud spending to grow 35.6% in 2026, with Europe projected at 83% growth. Buyers asking where they can host regulated workloads are forming new shortlists now.
  • Neoclouds are moving fast. Synergy counts nine neocloud companies, providers built for AI workloads, among the top 40 cloud providers. The challenger field is getting more crowded, not less.
  • Agents pick defaults. If more deployments start in AI coding tools, a platform that agents do not reach for loses the first step of the developer path, our inference from Vercel’s figures.

What does GEO look like for a cloud provider or cloud software vendor?

It makes your workload-level advantages easy for AI assistants to find, confirm and repeat. No provider, large or small, can be promised a slot in the answer.

  1. One precise identity. State which workloads you are best for, where you host, and how you price, the same way on your site, documentation, marketplace listings, analyst profiles and partner pages. Mixed messages produce vague answers; see how to fix wrong brand information in AI answers.
  2. Readable documentation and pricing. Keep pricing, region lists, compliance attestations and limits on plain, crawlable pages rather than behind sign-in or in scripts an assistant cannot render. We explain the rendering problem in when AI agents can’t read your site.
  3. Independent proof. Benchmarks run by others, analyst coverage, migration case studies with named customers, and conference talks by practitioners give assistants something independent to cite. Our guide to how brands build authority for AI search covers how that third-party proof builds up.
  4. Honest comparison content. Workload-by-workload comparisons and migration guides answer the questions buyers actually ask; ranked lists on other sites still carry more weight, as we show in best-of lists.
  5. Developer and agent paths. Clear quick-starts, templates and setup instructions are what a coding agent or a developer follows, so they deserve the same care as the homepage.
  6. Workload-by-region tracking. Ask each workload and region question repeatedly in ChatGPT, Google AI Overviews and AI Mode, Gemini, Perplexity, Copilot and Claude, and log the answers; how many prompts to track covers how many to run.

What remains unmeasured about AI answers and cloud workload decisions?

Nobody has measured whether appearing in AI answers moves a workload, or a commitment, toward one provider.

  • No cloud-specific buyer survey on AI use. The surveys we found cover software buyers in general, and the platforms that publish them have commercial interests. Surveys of software buyers more broadly, and their limits, are weighed in how B2B SaaS companies generate revenue from AI search.
  • Agent choices are a black box. Vercel’s figures show agents deploying, not how they choose a provider, and one platform’s data may not generalize.
  • Market figures move quickly. Synergy’s quarterly shares and Gartner’s forecasts are estimates and are revised often.
  • The tie to consumption revenue is unproven. Evidence linking AI visibility to business results is the thinnest of all; does AI visibility drive business results sets out what exists.

How can a cloud provider see which workloads AI answers send elsewhere?

Run the workload questions behind your recent wins and see which providers the answers name, and why.

Pick the five or six workloads where you win deals today, and the regions and compliance needs that come with them. Put those questions to ChatGPT, Gemini, Copilot, Perplexity and Google’s AI features several times each, noting who is named, which sources are cited, and whether your pricing, regions and certifications come out right. The gaps usually point to missing independent proof or unreadable documentation, not missing blog posts.

For an outside view, talk to us about a workload-level audit. We will map where AI answers place you on the questions that lead to proofs of concept and multi-year commitments, and which gaps most likely cost you enterprise deals. Our generative engine optimization service page sets out how the work after that audit is organized, workload by workload, from readable pricing and region pages to independent proof.

Frequently asked questions

Can a smaller cloud provider be named next to AWS, Azure and Google?

Yes, for specific workloads. In controlled tests a small, clear advantage beat a famous name, while generic questions favor the leaders.

Do hyperscaler marketplace listings help AI visibility?

Untested. They matter commercially, with marketplace sales forecast to reach $85 billion by 2028, and a clear listing gives assistants one more consistent description.

Should cloud pricing pages be public?

For the facts buyers compare, yes. Assistants can only repeat prices they can read, and FinOps teams check costs before any proof of concept.

Does sovereign cloud demand change what buyers ask AI?

Likely. Gartner expects sovereign cloud spending to reach $80 billion in 2026, which makes location and compliance questions more common.

Do AI coding agents count as buyers?

Increasingly they choose the first platform. At Vercel, agents triggered more than half of deployments by mid-2026.

Sources

Free strategy call

Some questions are easier to answer about your own business.

Bring the one that matters most. On a free 30-minute call we’ll take a first look at it and send you a short written read afterward.

Ask about my businessHow we work