Risk Management Strategies

Azure's Gateway Outage Hit South India and Jio India Central: When Your App Is Up but Your Office Can't Reach It, Which Policy Pays?

Azure's gateway services failed for about 5 hours 45 minutes across 18 regions, including South India and Jio India Central, while the applications behind them largely kept running. That failure shape falls between a cloud SLA credit and a cyber policy's dependent system failure trigger.

Tarun Kumar Singh
Tarun Kumar SinghStrategic Risk & Compliance SpecialistAIII · CRICP · CIAFP
9 min read

Listen to this article

Audio version • 9 min read

cloud outagedependent business interruptioncyber insurancesystem failure coverhybrid cloud

Last reviewed: October 2026

What failed between 30 September and 1 October

From 20:30 UTC on 30 September 2026 to 02:15 UTC on 1 October 2026, Azure's gateway networking layer failed across 18 regions, two of them Indian: South India and Jio India Central. That is roughly 5 hours 45 minutes, and some third-party trackers logged closer to 6 hours 17 minutes, according to the shattered.io incident write-up updated on 2 October 2026.

The services involved were the ones that sit at the edge of a customer's Azure estate rather than inside it:

  • ExpressRoute Gateway and VPN Gateway, the private and encrypted links from on-premises networks into Azure
  • Azure Firewall, Application Gateway and Web Application Firewall, which inspect and route traffic in and out
  • Azure VMware Solution, which many enterprises use to lift on-premises VMware estates into Azure with minimal change

The impact centred on hybrid connectivity. Workloads running inside Azure largely kept running. What broke was the path between a branch, a plant or a back office and those workloads. Microsoft's preliminary finding pointed to a "correlation" with infrastructure OS servicing activity, which it paused. As of 2 October it had not published a detailed post-incident review and had announced no blanket service-credit programme.

A day earlier, on 29 September, a separate Azure incident in Sweden Central caused failures in Azure OpenAI and AI Foundry services for about 5 hours 55 minutes. Two multi-hour Azure events in three days is a fair reminder that cloud dependency is now a recurring operational exposure, not a once-a-decade scenario.

Up but unreachable: why this failure shape is awkward

Most outage planning, and most insurance wording, assumes a binary. Either the system is down or it is up. The gateway incident produced a third state: the application could be up, the database could be up, the cloud provider's compute status could show green, and a bank branch, a claims office or a factory's MES client on a hybrid link could still be unable to reach any of it.

For a business on a hybrid setup, that third state is operationally identical to an outage. Core banking screens time out at the branch. A GCC's delivery floor in Bengaluru or Hyderabad cannot reach the client environment it supports. A manufacturer's plant-floor systems lose the link to the ERP instance that releases production orders and dispatch notes.

For a claim, it is not identical at all. The questions an adjuster asks are: what system failed, whose system was it, was it "unavailable" in the sense the policy uses, and for how long? Every one of those answers is murkier when the failed component is a network gateway owned by the cloud provider, the impaired traffic was originating from your own premises, and the provider's own status history may describe the event as degraded connectivity in some regions rather than a service outage.

The SLA credit is not your recovery

The first instinct after a cloud outage is to look at the service level agreement. That instinct should be calibrated. Cloud SLAs typically compensate with service credits calculated as a percentage of the monthly fee for the affected service, claimed by the customer within a set window, and capped at that fee. They do not compensate for lost revenue, idle staff, overtime, penalty clauses with your own customers, or regulatory consequences.

For this incident the position is thinner still. Microsoft announced no blanket service-credit programme, so any credit depends on each customer showing that its own gateway resources breached the published availability commitment for that specific service, and filing within the provider's claim window. Even a successful claim returns a fraction of the monthly gateway or firewall bill. For an enterprise whose branches were cut off for six hours, the gap between that credit and the actual loss can run to orders of magnitude.

So the SLA credit is a commercial remedy, not a recovery mechanism. The financial loss, if it is going to be recovered at all, has to come from insurance, and specifically from the business interruption sections of a cyber insurance policy.

Where a cyber policy might respond, and where it might not

Indian cyber wordings usually split first-party business interruption into two triggers. One responds to a security failure such as ransomware or a network intrusion. The other, offered as system failure cover, responds to an unplanned, non-malicious outage of computer systems. Both then split again by whose systems failed: your own, or a dependent business or service provider's. The cyber business interruption cover piece walks through these modules in detail.

The Azure gateway incident was, on the preliminary finding, non-malicious and linked to the provider's own servicing activity. So the relevant trigger is dependent system failure: an unplanned outage of a third-party provider's systems that interrupts your business. That is the right door. Whether it opens depends on three points in the wording: who counts as a provider, what counts as a failure, and the waiting period, which the next section covers.

Who counts as a provider

Some wordings define the dependent business or service provider broadly, as any entity providing IT services to the insured under written contract. Others list named categories, or require the provider to be scheduled. If your policy lists "cloud hosting providers" but your ExpressRoute circuit is delivered jointly through a telecom carrier and Microsoft, check whether both sit inside the definition. Some wordings also carve out telecommunications and internet infrastructure, which a network gateway failure can be argued to fall into.

What counts as a failure

Many system failure definitions require an "unplanned and unintentional outage" or "interruption" of the provider's computer system. An insurer could reasonably ask whether a gateway that failed while applications stayed up meets that test. A wording that also covers degradation or impairment of access, not just total unavailability, is far easier to claim under for this kind of event.

The waiting period and the clock problem

Cyber business interruption cover almost always carries a waiting period (sometimes called a time deductible) before loss starts to count. Waiting periods of 8, 12 or 24 hours are common, and dependent-provider extensions are sometimes set higher than first-party cover. Against a 5 hour 45 minute incident, a 12 hour waiting period produces no recovery at all, regardless of how much revenue was lost inside those hours.

This is the single most important number to check after this event. Three practical points follow:

  1. Know which clock applies. Some policies start the waiting period at the onset of the outage, others at the point the insured discovers it or notifies the insurer. A late-night UTC incident, which began at 2:00 am IST on 1 October, can easily lose an hour or two between onset and recognition.
  2. Ask whether the waiting period is a deductible or a qualifier. In some wordings, once the waiting period is exceeded, loss from the start of the outage is payable. In others, only loss after the waiting period counts. The difference is large for short events.
  3. Price a shorter waiting period against a realistic outage profile. The incidents cited here ran about six hours each. A 6 or 8 hour waiting period costs more premium, but a 24 hour one turns the cover into catastrophe-only protection that may never respond to the outages you actually see.

Proving income loss when nothing was technically down

Assume the trigger is met and the waiting period is exceeded. You still have to prove the loss. Business interruption in cyber policies is typically measured as the reduction in net profit plus continuing operating expenses, and extra expense incurred to reduce the loss, over the indemnity period. That requires evidence that the interruption caused the loss, which is harder when your own monitoring shows healthy applications.

Build the evidence chain from the edge inward:

  • Connectivity telemetry: on-premises router and firewall logs showing tunnel or circuit drops to Azure, with timestamps.
  • Provider records: the Azure Service Health notices and resource health history for your specific gateway resources, exported promptly, since portal retention is limited.
  • Business throughput: branch transaction counts, policies issued, tickets handled or production orders released by hour, compared against the same weekday in prior weeks.
  • Extra expense: overtime, emergency failover to internet VPN or an alternate region, and penalty or service credits paid to your own customers.

A loss adjuster will want the gap between "the cloud was up" and "we could not trade" to be closed by your data, not by narrative. Teams that already map business processes to the network paths they use, as covered in cloud dependency business interruption for SaaS firms, can produce this within days. Teams that cannot will find the claim stalls on causation.

What Indian banks, insurers, GCCs and manufacturers should check now

Indian hybrid estates have their own concentration: two of the 18 affected regions were Indian, and data-localisation choices push many regulated entities towards in-country regions. Physical events add to the picture. In June 2026 a fire at STT GDC's New Delhi data centre disrupted cloud services, and Google warned of elevated latency for traffic from Delhi, Chennai and Mumbai (Business Today, 24 June 2026). Connectivity failures and facility failures can both leave applications intact and users stranded.

A short review list by segment:

  • Banks and NBFCs: map which branch and ATM-switch paths ride ExpressRoute or VPN Gateway into a single Azure region, and test whether your cyber policy's provider definition captures the carrier and the cloud gateway together.
  • Insurers and intermediaries: check whether policy issuance and claims intake count as "business" for income loss, or whether the wording only measures revenue from customer-facing sales.
  • GCCs: confirm whether loss is measured at the Indian entity or the parent, since a GCC billed on cost-plus may show little direct profit impact even when the parent loses heavily.
  • Manufacturers: test whether plant systems can keep running for six hours with no cloud link, and if not, whether non-damage business interruption wording or a cyber system failure extension is the better fit.

For each, record the waiting period, the provider definition, whether degradation counts, and the indemnity period, then compare them against the September and October incidents. The adjacent question of multiple vendors sharing one cloud estate is covered in dependent system failure cover for AI provider concentration. This post's point is narrower: even with a single provider and fully healthy applications, the connectivity layer can be the failure that matters.

Engineering and contract steps that reduce the uninsured gap

Insurance will not close the whole gap, so the risk owner should treat the policy as one layer and fix the rest in architecture and contracts.

On the engineering side, the gateway incident argues for a secondary path that does not share the failed component: a second ExpressRoute circuit terminating on a gateway in a different region, or an internet-based VPN or SD-WAN fallback that can be switched on by runbook in minutes. Test it under load, not just at go-live.

On the contract side, review the outsourcing and cloud agreements for notification commitments, root cause disclosure timelines and audit rights. A provider that has not published a post-incident review a day after the event leaves the customer without the documentation an insurer or regulator will ask for. Writing a disclosure timeline into the contract, and keeping your own logs regardless, protects both the claim and the board report.

Finally, take the result back to the underwriter. Insurers price dependent system failure partly on how concentrated and how recoverable the insured's dependencies look. Showing a tested failover path and a mapped dependency register is the most credible argument for a shorter waiting period and a wider provider definition at the next renewal.

About the Author

Tarun Kumar Singh

Tarun Kumar Singh

Strategic Risk & Compliance Specialist

  • AIII
  • CRICP
  • CIAFP
  • Board Advisor, Finexure Consulting
  • Developer of the Behavioural Underinsurance Risk Index (BURI)

Tarun Kumar Singh is a seasoned risk management and insurance professional based in Bengaluru. He serves as Board Advisor at Finexure Consulting, where he advises insurance, fintech, and regulated firms on governance, growth, and trust. His work spans insurance broker regulatory frameworks across India, UAE, and ASEAN, IRDAI compliance and Corporate Agency model reform, VC governance in insurtech, and MSME insurance gap analysis. He is the developer of the Behavioural Underinsurance Risk Index (BURI), a framework applying behavioural economics to underinsurance and insurance fraud risk.

Frequently Asked Questions

Will Microsoft compensate my business for the 30 September 2026 Azure gateway outage?
Only through service credits, if at all. Microsoft announced no blanket credit programme, so each customer has to show its own gateway resources breached the published SLA and file within the claim window. Credits are a share of the affected service's monthly fee and do not cover lost revenue or extra expense.
Does a cyber policy cover an outage that was not a cyber attack?
It can, if the policy includes system failure cover and extends it to dependent or third-party service providers. A non-malicious cloud outage falls under dependent system failure, not the security failure trigger. Some policies sold as cyber cover include only the security trigger, so check the wording.
Our applications in Azure never went down. Can we still claim?
Possibly, but it depends on how the policy defines a failure. Wordings that cover degradation or impairment of access to the provider's systems are easier to claim under than those requiring a total outage. You will also need connectivity logs and business throughput data to show the interruption caused the loss.
What waiting period should we negotiate for dependent system failure cover?
Set it against the outage durations you actually see. The two Azure incidents in late September 2026 each ran about six hours, so a 12 or 24 hour waiting period would have produced no recovery. A 6 to 8 hour waiting period costs more premium but responds to realistic events.

Related Glossary Terms

Related Insurance Types

Related Industries

Related Articles

Sarvada Intelligence

Ready to see Sarvada in action?

Explore the platform workflow or start a product conversation with our underwriting automation team.

Explore the platform