A laptop displays workflow software connected to cloud servers, online storefronts, and e-commerce tools.
Thunders, a Paris-based AI testing vendor, says it has joined the Microsoft for Startups Pegasus Program. Its testing platform is now available on Microsoft Marketplace. The announcement matters to Azure, GitHub and Azure DevOps shops because it changes how they can buy the tool. It does not show that the tool performs better. Here is what the claims amount to and what they leave open.

A laptop displays workflow software connected to cloud servers, online storefronts, and e-commerce tools. What was announced​

The Marketplace route is the practical change. Organizations can procure Thunders through their existing Microsoft agreements, with eligible purchases counting toward their Microsoft Azure Consumption Commitment (MACC). In plain terms, a team with a prepaid Azure commitment may be able to put a testing tool on that budget without opening a new vendor contract.

The word "eligible" matters. Not every Marketplace listing automatically counts toward a commitment, and the announcement doesn't give the offer's pricing, plans or eligibility terms. Check the live listing and your own Microsoft agreement before you plan a purchase around MACC drawdown.

A timing note​

The story was published on 16 November 2026, but the news is older. A Thunders press release carried by openPR is dated 4 November 2026. Thunders' own company timeline places the selection and Marketplace availability earlier, in September 2026. That timeline also says Thunders is one of three French startups in the cohort, though that count is Thunders' own. Treat the mid-November coverage as a late retelling, not a fresh launch.

What Pegasus does and doesn't mean​

Microsoft describes the program as invite-only, helping qualified startups accelerate enterprise customer acquisition through curated introductions and hands-on go-to-market support. That is a sales and distribution benefit for the startup.

Pegasus is not a certification of product quality. A trade write-up of a similar Pegasus announcement, from another startup, makes the same point: participation does not itself guarantee customer contracts or adoption. Microsoft's quote in the Thunders story is a welcome message, not a performance review.

What the product claims to do​

Thunders says its "AI Test Agents" create tests from plain-language descriptions of user journeys. It says they run those tests across web, mobile, API, desktop and enterprise environments, such as SAP, and update them as applications change. It also says each run records screenshots, console output and the agent's reasoning, and that it can be deployed as SaaS, hybrid or self-hosted.

Some of this is backed by Thunders' own documentation, which is still vendor material:

  • The help center says Thunders AI can create web, mobile and API tests. It can also mix app types in one test case, such as web plus API.
  • The same documentation does not establish the desktop and enterprise-system coverage claimed in the announcement. Buyers with SAP or desktop estates should ask for a proof of concept.
  • Release notes show separate Cloud and Self-Hosted release streams. Thunders says self-hosted releases moved to calendar versioning in step with Cloud, with a recent self-hosted build numbered 2026.11.12.
  • The documentation warns that enabling Web Browse lets the AI interact with the target page. That can change page state or submit forms, so Thunders advises caution on production environments.

No benchmarks were provided for test accuracy, maintenance savings or release speed. The "tests update themselves as the UI changes" claim is a company claim, not a demonstrated result.

The Microsoft stack​

Thunders says it runs on Azure, using Azure Container Apps and Azure Kubernetes Service for test execution, Azure SQL Database for test data, and Azure Monitor and Application Insights for observability. It says its AI functions use Microsoft Foundry and its analytics use Microsoft Fabric. It lists integrations with GitHub, GitHub Copilot and Azure DevOps.

I could not independently verify these architecture details. Because they come from the vendor, they also say nothing about how deep the integrations go. Questions worth asking include:

  • Where does test data live, and in which region?
  • Do the Azure DevOps and GitHub integrations trigger runs from pipelines, or only report results?
  • What does the self-hosted option need from your own Azure subscription?

The DORA statistic in context​

The announcement leans on the 2024 DORA finding that a 25% rise in AI adoption was associated with an estimated 7.2% drop in delivery stability. Google Cloud's summary of the report confirms the number. It also gives the surrounding picture: the same analysis estimated a 1.5% decrease in delivery throughput, alongside gains in documentation quality (7.5%), code quality (3.4%) and code review speed (3.1%). DORA says the data point to the basics, such as small batch sizes and robust testing.

That supports a case for stronger testing in general. It is an association from survey-based modeling. It doesn't show that AI coding tools cause instability, and it doesn't show that any particular testing product reverses the trend.

The Marketplace growth figures​

The story cites Omdia's forecast that cloud marketplace sales will rise from USD $30 billion in 2024 to USD $163 billion by 2030. That is a projection, not realized sales. Thunders' own figure of more than USD $500 billion in commitments to the largest cloud platforms is attributed only to the company, and I couldn't check its definition or time frame. Treat both as marketing context.

Company background​

Company information is vendor-supplied. Thunders' own timeline says founders Karim Jouini and Jihed Othmani previously built the expense-management company Expensya, which it says Medius acquired in 2023 for over $100 million. It also says Jouini worked at Microsoft as a software engineer, and that Thunders has ISO 27001 and SOC 2 Type 2 certification dating from July 2025. It names Paris, Tunis and Boston as its offices. The customer list in the announcement (Allianz, SNCF Connect, PwC, Sopra Steria, LegalPlace) and the "co-sell ready" status are company statements I couldn't corroborate independently.

Practical takeaways for IT and dev teams​

  1. Check the listing. Confirm the Marketplace offer's plans, price and MACC eligibility with your Microsoft account team or procurement.
  2. Test the coverage claim on your own stack. Run a trial against your actual mix of web, API, mobile and any SAP or desktop workflows.
  3. Choose a deployment model deliberately. SaaS, hybrid and self-hosted carry different data-residency and operations burdens.
  4. Keep humans in the loop. AI-generated tests still need review. The evidence trail (screenshots, console output, reasoning) is useful for audits, but only if people actually read it.
  5. Don't read the badge as endorsement. Pegasus membership signals a go-to-market relationship with Microsoft, not that Microsoft has validated the product.

Marketplace procurement makes it easier for Azure-committed enterprises to try a new class of AI testing tool. Whether it earns a place in the pipeline depends on a pilot, not on the Pegasus badge.

 

References

  1. Thunders joins Microsoft for Startups Pegasus programme - IT Brief Asia IT Brief Asia 2026-11-16T09:30:00+00:00
  2. Thunders Joins the Microsoft for Startups Pegasus Program as AI Puts Pressure on Enterprise Software Testing openpr.com
  3. Announcing the 2024 DORA report | Google Cloud Blog cloud.google.com