AWS removed the 30-day minimum retention requirement for transitioning objects into S3 Standard-IA and One Zone-IA, so lifecycle policies can shift infrequently-accessed data out of Standard within hours instead of waiting a month. For workloads with unpredictable access patterns — logs, media libraries, backup staging — this closes a gap that pushed teams toward third-party tiering tools or manual scripts. It’s a small mechanical change with an outsized effect on realized savings, since the old 30-day floor was the single biggest reason IA tiers underperformed on paper cost models.

Lambda functions can now reference source code directly from a customer-owned S3 bucket instead of Lambda silently copying it into a managed one, removing the old code-size ceiling and cutting activation latency after updates by skipping the copy step. It’s a small plumbing change, but it’s consistent with AWS’s pattern this year — Annotations, Vectors, now this — of using S3 as the connective tissue between storage and every other service, which raises the switching cost of leaving it.

S3 Standard and EBS gp3 list pricing remain unchanged for a second straight quarter even as enterprise SSD contract costs continue rising on AI-driven NAND and HBM demand. AWS can absorb the squeeze on a small storage line within a much larger compute bill; independents like Wasabi and Backblaze already couldn’t (see Industry, below). The read for buyers: don’t assume the flat list price is permanent — treat Q4 enterprise agreement renewals as the next checkpoint where the gap could show up.

Microsoft extended native Azure Blob Storage connectors into Microsoft Foundry, Ray, Anyscale, and LangChain, letting enterprise data feed Foundry IQ for grounding and fine-tuning without a separate ETL hop. The strategic logic mirrors AWS’s S3-as-AI-substrate push: the more frameworks read Blob natively, the more each new AI workload defaults to Azure storage rather than a neutral data lake.

The .NET and Java Blob SDKs (service version 2026-06-06) added support for Blob Smart Tier alongside directory-scoped SAS tokens, giving developers programmatic control over automated hot/cool/archive placement instead of only console- or Actions-based rules. Combined with last month’s Storage Actions GA, Microsoft now offers the same tiering automation at both the policy layer and the code layer — a deliberate belt-and-suspenders push to get customers off manually-managed tiers.

Azure Storage Mover’s direct S3-to-Blob cloud-to-cloud transfer — no self-hosted agent required — is now showing up in enterprise migration playbooks (SAP RISE/HANA teams among the first documented cases), using Azure Arc’s multicloud connectors to authenticate against AWS directly. It’s the same move Azure made for Google Cloud Storage sources in July: hyperscalers are increasingly building first-class tooling to pull data off competitors, not just to move it around inside their own walls.

Nearline multi-region storage rose from $0.010 to $0.015/GB-month, while Archive multi-region pricing fell from $0.0040 to $0.0024/GB-month in the US and EU (to $0.0030 in Asia). Lifecycle rules written before 2026 that push aging data into multi-region Nearline now cost 50% more to execute and may no longer beat leaving data in Standard; the same rules pushing straight to Archive got meaningfully cheaper. This is a rare case where the fix isn’t a new tool, it’s re-running the TCO math on policies nobody has touched since they were set.

Rapid Bucket — Google’s zonal object storage class built on Colossus, delivering 15+ TB/s aggregate throughput and 20M QPS with sub-millisecond latency — is moving from Next ’26 showcase to production adoption, with checkpoint restores reported 5x faster than standard object storage. It’s Google’s clearest answer yet to the argument that object storage is too slow for GPU-bound training loops; the tradeoff is zonal-only placement, which pushes customers toward tighter compute/storage co-location than they may be used to.

Starting August 26, 2026, Object Lifecycle Management’s Delete action will remove empty folders in hierarchical-namespace-enabled buckets once they meet a rule’s conditions — closing a long-standing gap where lifecycle cleanup left orphaned empty folder structures behind. Minor on its own, but relevant for any team running HNS buckets with automated retention who assumed folder structure was untouched by lifecycle rules.

Wasabi raised its pay-as-you-go rate from $6.99 to $7.99/TB-month across all regions, citing rising storage hardware, energy, and data-center costs tied to global infrastructure constraints. It’s the clearest signal yet that the 15-year secular decline in cloud storage pricing ended in 2025: AI-driven HBM demand has reallocated wafer capacity across the memory and storage supply chain, and even the low-cost independents built to undercut AWS/Azure/GCP on price no longer can.

Backblaze raised B2 pay-as-you-go pricing to $6.95/TB-month while eliminating API transaction fees entirely — a bet that a simpler, all-in rate is worth more to customers than the lower headline number plus metered fees. Combined with Wasabi’s July increase, both major independent object storage vendors have now raised list prices within the same year, removing the pricing gap that used to be their main differentiator against the hyperscalers.

Analyst forecasts put the cloud storage market growing from roughly $179B in 2026 to $514B by 2031 (a ~23% CAGR), with 94% of enterprises now using some form of cloud storage and 54% running three or more providers simultaneously. The driver isn’t just data growth — it’s generative-AI pipelines turning storage from a back-office line item into a first-order competitive resource, which is also why pricing power has shifted back toward providers after a decade of steady declines.