Picture a large organisation’s cloud estate from the outside. AWS handles the core applications because that’s what the platform team standardised on five years ago. Azure crept in through Microsoft 365 and — well, it was already there, so the Active Directory migration made sense. GCP appeared when the data science team benchmarked it for ML workloads and quietly won. And on-premises? That’s still running the things compliance won’t let anyone move, plus a few legacy systems whose migration was deprioritised three roadmaps in a row.
Nobody planned this. Nobody didn’t plan it either. It just… accumulated. Like a sock drawer that somehow became three sock drawers.
This is the quiet reality behind the phrase “hybrid and multi-cloud strategy.” For most enterprises, the architecture wasn’t designed — it was inherited. And the interesting question now isn’t whether to operate across multiple clouds. It’s whether the operating model has any idea what’s actually running where.
The Numbers Tell a Familiar Story
Ninety percent of enterprises now run hybrid or multi-cloud strategies. That’s not a trend — it’s just the condition of operating at scale in 2025. What’s more revealing is how those environments are actually structured. The Flexera State of the Cloud Report found that the most common multi-cloud implementation isn’t elegant workload portability — it’s applications siloed on different clouds, often because each got there through a different team, a different acquisition, or a different conversation from two years ago.
Here’s the one that always gets a quiet wince in finance reviews: only 30% of organisations actually know where their cloud budget is going. Not roughly — exactly. The other 70% are, to varying degrees, guessing. Enterprises waste somewhere between 28% and 35% of total cloud spend on idle resources, over-provisioned compute, and forgotten infrastructure from projects that ended but whose cloud footprint didn’t. One company discovered $850K annually in abandoned resources. A hospital network found over $1 million in duplicate patient imaging stored across two clouds for “redundancy” — without a policy governing either copy.
The cloud bill arrives monthly. The visibility, apparently, does not.
The Lock-In Problem Has a Twist
Multi-cloud was partly sold on the promise of avoiding vendor lock-in. The theory was sound: spread workloads across AWS, Azure, and GCP, and no single provider has you over a barrel when contract renewal comes around.
The reality is more entertaining. Moving workloads between clouds is genuinely hard — data gravity, proprietary services, and egress fees create a kind of soft lock-in that no multi-cloud strategy entirely solves. And the operational complexity of running across three hyperscalers often costs more in management overhead than the negotiating leverage it theoretically creates.
This is the tension worth sitting with: multi-cloud’s biggest actual benefit isn’t portability — it’s resilience and best-of-breed optionality. The Azure outage in July 2024 grounded airlines and disrupted hospitals. The AWS us-east-1 outage in December 2024 took Netflix, Spotify, and thousands of other services offline. Organisations with workloads distributed across clouds had options during both events. Those with everything in one place did not. That’s the genuine value of multi-cloud — and it’s different from the marketing version.
The Repatriation Plot Twist
Here’s a wrinkle that doesn’t appear in most cloud strategy decks: roughly 21% of workloads that moved to the cloud have since moved back, at least partially. Databases, high-performance compute, and certain AI/ML workloads are increasingly returning to on-premises or private cloud environments — not because cloud failed, but because cost and latency economics work differently at scale than they do on a migration business case.
This is exactly the hybrid architecture conversation — keeping latency-sensitive and cost-heavy workloads closer, reserving public cloud for the elasticity and scale it genuinely does well. The organisations navigating this thoughtfully are the ones treating workload placement as a continuous decision, not a one-time migration conclusion.
The Orchestration Layer Is the Actual Strategy
This connects directly to the composable architecture thread I explored recently. Multi-cloud is composability at the infrastructure layer. The value isn’t in picking three clouds — it’s in building the governance layer that spans all of them.
Sixty-one percent of large enterprises now use multi-cloud security tools. Fifty-seven percent use multi-cloud FinOps platforms. That’s the market telling you where the actual leverage sits: not in the infrastructure, but in the visibility and control layer above it. FinOps — treating cloud spend as a financial discipline with real accountability — has moved from a specialist practice to a board-level conversation. The organisations getting the most from hybrid and multi-cloud aren’t the ones with the most sophisticated infrastructure. They’re the ones that know exactly what’s running, what it costs, and who’s accountable for both.
The lens worth applying: cloud strategy isn’t really a technology question anymore. It’s a governance question that happens to involve technology.
In your organisation, if you drew a map of every cloud environment and the decision that created it — how many would trace back to a deliberate strategy, and how many would trace back to a convenience that quietly became permanent?
Let’s keep learning — together.
harish.balasubramanian 
Share your thoughts