Cloud Computing Watch: Enterprise Cloud Infra Hits $129B, Amazon Announces AWS Interconnect Availability & Cloudflare's 13-Incident
The Reliability Crisis Nobody's Punishing: Outages Pile Up as Cloud Growth Keeps Accelerating
Reliability Is Degrading, Growth Isn't Slowing, and the Two Trends Are Directly Connected
Cloudflare's public status page logged 13 separate incidents between August 7 and 14 alone, spanning R2 storage, Durable Objects, Workers KV, and Workers AI. AWS's US-West-2 region went down for roughly 80 minutes this month — its fourth incident in four months. Yet Cloudflare posted $696.1 million in Q2 revenue, up 36% year-over-year, adding more than 80,000 paying customers in the quarter. AWS's market position hasn't visibly moved either. The uncomfortable pattern across this cloud cycle: reliability is degrading in real time while growth keeps accelerating, and the two trends share a common root cause rather than being unrelated.
Forrester's own 2026 predictions flagged the mechanism directly: hyperscalers are diverting investment away from legacy x86 and ARM environments to prioritize GPU-centric AI data centers, leaving aging general-purpose infrastructure to falter under growing complexity — a trade-off Forrester expects will produce at least two major multi-day hyperscaler outages before year-end. This edition covers that reliability-growth tension alongside the accelerating multicloud interconnect trend and Oracle's continued outperformance of every other major cloud provider on growth rate.
01 — The Cloud Lead
Outages Are Piling Up, and the Market Genuinely Doesn't Seem to Care Yet
Cloudflare's 13-incident stretch this month is the most visible evidence of a broader pattern: cloud infrastructure reliability is measurably worse in 2026 than in prior years, even as usage keeps climbing. AWS alone has now logged four separate incidents in four months, the latest an 80-minute US-West-2 outage. Between August 2024 and August 2025, AWS, Azure, and Google Cloud together experienced more than 100 service outages combined — and 2026's pace shows no sign of slowing that trend.
What's genuinely striking is the complete absence of customer flight. AWS still leads global cloud infrastructure spending with a 28-30% share as of Q1 2026, per Synergy Research Group, with no visible dent from any of its 2026 incidents. Cloudflare added 80,000+ paying customers during the exact window its outage count spiked. The explanation is structural, not sentimental: migrating cloud providers is costly and slow enough that even a bad reliability month doesn't change switching-cost math, which is precisely why hyperscalers can currently under-invest in legacy reliability without near-term financial consequence.
02 — Hyperscaler Cloud Watch
Enterprise cloud infrastructure spending hit $129 billion in Q1 2026 alone, up 35% year-over-year, driven largely by AI workload demand. AWS holds the largest absolute share at 28-30%, with Azure at 20-22% and Google Cloud at 12-14% — but the percentage-growth story runs in the opposite direction of the share story: Azure and Google Cloud are both growing meaningfully faster than AWS (roughly 40% and 63% year-over-year respectively, against AWS's 19%), meaning smaller bases are closing the gap even as AWS retains the largest footprint in absolute terms.
The more striking outlier sits outside the traditional "Big Three" framing entirely: Oracle's fiscal Q3 2026 results showed total cloud revenue (IaaS plus SaaS) of $8.9 billion, up 44% year-over-year, with the infrastructure piece alone reaching $4.9 billion, up 84% — expanding three to four times faster than AWS on a percentage basis. Oracle's growth is increasingly a database-and-multicloud story rather than a general-purpose infrastructure one, and its regional footprint (53 regions, more than Google Cloud's 48) reflects that enterprise-database-first strategy directly.
03 — Multicloud & Interconnect
AWS announced general availability of AWS Interconnect – Multicloud, providing simple, resilient, private connections to other cloud providers, with Google Cloud as the first launch partner and Microsoft Azure and Oracle Cloud Infrastructure both confirmed to join later in 2026. The move directly replaces the "do-it-yourself" multicloud networking enterprises previously had to assemble themselves, and reflects a broader recognition among hyperscalers that customers are adopting multicloud strategies regardless of whether any single vendor prefers it.
Oracle's own new private, high-speed interconnect with AWS, launching in US East (N. Virginia) later this year, is a parallel move from the opposite direction — Oracle explicitly courting AWS-centric enterprise customers who still depend on Oracle databases, removing the last major friction point keeping those workloads separated. Enterprise customers "increasingly refuse to bet on a single cloud provider," and this new generation of native interconnects is the infrastructure response to that refusal, transforming multicloud from an operational headache into a supported, first-class deployment pattern across all four major providers.
04 — Enterprise Cloud Adoption
Forrester projects at least 15% of enterprises will seek private AI deployments atop private cloud infrastructure in 2026, a direct response to rising AI costs, data lock-in concerns, and the operational risk the reliability data above makes concrete. Vendor-control incidents are reinforcing the same anxiety from a different angle — Salesforce's decision to shut down third-party access to the Slack API, cutting off customers' ability to use their own Slack data on other platforms, is exactly the kind of lock-in enterprises are increasingly trying to architect around. The combination of visible reliability degradation and vendor-control tightening is pushing a meaningful minority of enterprises toward genuine infrastructure diversification rather than simply accepting single-vendor dependency as the cost of doing business.
05 — Cloud Economics
Downtime costs Global 2000 companies an estimated $400 billion annually, according to Oxford Economics research for Splunk — roughly $9,000 per minute, or $540,000 per hour, with revenue recovery taking a reported 75 days on average after a major incident. That cost is being absorbed almost entirely by customers rather than showing up as pricing power lost by providers, which is precisely why hyperscalers face limited near-term financial pressure to redirect capital back toward legacy reliability investment. From the provider side, the calculus is straightforward and, per Forrester's framing, entirely rational: AI infrastructure represents the future of cloud computing revenue, and missing that opportunity would be far more costly than the reputational and customer-relations cost of absorbing a rising outage count on general-purpose infrastructure.
06 — Capital & Competition
Oracle's aggressive interconnect strategy toward both AWS and Azure is paying off in growth terms that none of the traditional Big Three can currently match, and it's forcing a competitive response: AWS and Google's own multicloud collaboration, initially framed around customer convenience, is also explicitly described internally as a "competitive bulwark against Microsoft Azure in the race for AI-native cloud dominance." Salesforce's support for that same AWS-Google initiative signals enterprise software vendors are actively choosing sides in the multicloud interconnect wars rather than remaining neutral — a form of platform alignment with real competitive consequences as more SaaS vendors build native integrations favoring specific interconnect partnerships.
Three Cloud Signals
- Reliability and growth have decoupled, at least for now. Rising outage counts across every major provider are producing no visible customer defection, because switching costs remain higher than the cost of absorbing periodic downtime — a dynamic that only holds until a genuinely catastrophic, multi-day incident tests it directly.
- Multicloud interconnect has gone from workaround to first-class product. AWS, Google, Oracle, and (soon) Azure are all building native private interconnects rather than leaving multicloud plumbing to third-party tooling — validating that enterprise multi-provider strategies are now permanent, not transitional.
- Growth-rate leadership has shifted away from the traditional Big Three. Oracle's 44-84% growth rates, and Azure/Google's faster percentage growth against AWS's larger base, both show the cloud market's competitive momentum no longer maps cleanly onto absolute market share.
The CODEW Take
Cloud computing is running two parallel infrastructure strategies simultaneously, and the tension between them is becoming visible in real time. Capital is flowing overwhelmingly toward GPU-centric AI infrastructure, where growth and pricing power are strongest; general-purpose legacy infrastructure is being left comparatively under-invested, and reliability is degrading as a direct, predictable consequence. That trade-off has not yet cost any major provider market share, because enterprise switching costs remain high enough to absorb a rising outage count — but Forrester's prediction of at least two major multi-day outages in 2026 is a warning that the current arrangement has a breaking point, not a permanent equilibrium.
The company best positioned right now is arguably Oracle, growing 3-4x faster than AWS by leaning into multicloud interconnect and database-centric enterprise relationships rather than competing head-on for general AI infrastructure share. What to watch next: whether Forrester's multi-day-outage prediction actually materializes, and whether the first genuinely catastrophic incident of this cycle finally moves the customer-retention numbers that a dozen smaller outages so far have not.
Reviewed by Erwin Castro
on
Thursday, August 20, 2026
Rating: