April was about putting the numbers on the table and making it easier for teams to verify them.
We ranked every major block storage provider on the same workload. We opened a free ClickHouse beta. We shipped a stack of platform improvements around security and IaC. We launched a public changelog. And our team came back from CloudFest 2026 with a clear read on where enterprise cloud spend is heading next.
Here's what moved.
The Complete Block Storage Comparison, 2026

We pulled every major provider into one table and ranked them by IOPS-per-dollar on the same 1 TB / 20,000 IOPS workload. The cost spread between the cheapest and most expensive: 15×. For the same performance contract.
That's the headline. The story underneath is that most teams are paying for a tier they don't need and the gap is widening as hyperscalers raise prices.
Run the comparison yourself: https://nirvanalabs.io/blog/aws-ebs-alternatives-in-2026
Google Cloud Hyperdisk Balanced (and HA) vs Nirvana ABS

Google announced Hyperdisk Balanced improvements at Next '26 — up to 160K IOPS, 2.4 GiB/s per volume, sub-millisecond latency. The engineering is real. The pricing model isn't quite what the headline suggests.
Three things worth understanding before you buy:
- "Up to 160K IOPS" is a provisioned maximum, not a baseline. Free baseline is 3,000 IOPS. Every IOPS above that is metered separately at $0.005/mo. ABS includes 20,000 sustained IOPS in the storage price.
- Hyperdisk Balanced HA costs 2× standard. Smart engineering — but storage, IOPS, and throughput all double. 1 TB at 20K IOPS lands at ~$334/mo on HA vs $96/mo on ABS.
- Your VM has its own IOPS ceiling. Hitting high IOPS often requires a larger VM, which compounds compute cost. Nirvana compute runs 26–34% cheaper than GCP N2 at every equivalent config — and 20K IOPS works on any instance.
Full-stack cost at 8 vCPU / 32 GB / 1 TB / 20K sustained IOPS:
- Nirvana: $283/mo
- GCP Hyperdisk Balanced: $451/mo
- GCP Hyperdisk Balanced HA: $617/mo
That's 37% cheaper than HB and 54% cheaper than HB HA. SOC II Type 2 compliant, trusted by 50+ production clients including BitGo, Fireblocks, Nansen, Keyrock, and Monad.
If you're running PostgreSQL, ClickHouse, Elasticsearch, vector databases, or trading systems - run the numbers.
Full breakdown: https://blog.nirvanalabs.io/google-hyperdisk-balanced-vs-nirvana-abs/
What Is a Cold Read? (And Why Most Benchmarks Hide It)

Most cloud storage benchmarks look great because their caches warm up in seconds. They aren't measuring your disk anymore — they're measuring how well caching hides latency.
A cold read is what happens when data isn't in cache and has to come straight from storage. It's the moment after a restart. After a failover. After a working set grows past RAM. After a node scales out and shows up empty.
These aren't edge cases. They're the moments your system is judged on.
We dropped the OS page cache before every query in our ClickBench run. With caching out of the picture:
- ~10.5× median speedup of ABS over gp3 across analytical queries
- ~14.1× on IO-heavy queries
- gp3 ballooned to tens of seconds. ABS stayed sub-second.
Cold read performance answers one question: how fast does your system recover when things go wrong?
Read the explainer: https://blog.nirvanalabs.io/what-is-a-cold-read/
ClickHouse Free Beta
We opened a free beta for self-hosted ClickHouse on Nirvana Cloud, in our Silicon Valley datacenter.
What's included:
- Accelerated Block Storage with 20K baseline IOPS
- No IOPS tax, no egress fees
- Managed Kubernetes (NKS) so you can bring workloads as-is
If your queries live and die on IOPS, this is a chance to test on infrastructure that doesn't throttle you for using what you paid for.
Beta spots are limited. → https://nirvanalabs.io
Product Update: Equivalent Code, API Keys, SIEM, and More
A big shipping month for the platform.
- Equivalent Code panel — see the IaC equivalent of every action you take in the UI
- API key permissions — granular scoping, smaller blast radius
- SIEM integration — pipe audit logs straight into your security stack
- JIT provisioning controls — tighter access without the toil
- GitHub sign-in — fewer credentials to manage
- Volume attach / detach — lifecycle ABS volumes from any instance
Details: https://blog.nirvanalabs.io/product-update-equivalent-code/
Public Changelog Is Live
We made it easier to track what we ship. Every product change, in one place.
→ https://nirvanalabs.io/changelog
CloudFest 2026: What We Took Away
Our own @csnirvana spent three days at CloudFest in Rüst, Germany. A few stats from the conference floor that hit hard:
- 80%+ of cloud storage is still on HDDs — confirmed independently by Western Digital and Seagate. Disaggregated, high-IOPS storage isn't a niche; it's the biggest gap in the market.
- 75% of companies are reprioritizing data security because of agents accessing their data (AvePoint). The emerging framework: treat AI agents like digital contractors.
- The neocloud thesis is real. Per Ditlev Bredahl, neoclouds will capture 50% of remaining CSP business. In 2025, 95% of CSPs had zero AI infra offerings — while AI was 35% of industry revenue.
- HPE Private Cloud AI: 14 days vs 12–18 months DIY. The enterprise floor is rising fast.
TL;DR: storage is the unsexy kingmaker, agentic governance is where enterprise spend is heading, and you don't need hyperscaler scale to build a profitable cloud business anymore.
The Bigger Picture: Cloud Costs Are Going Up. Ours Aren't.
Cloud infrastructure costs spiked across the board in 2026, driven by the AI buildout:
- Hetzner: 30–50% increases on April 1
- OVHcloud: hardware costs up 15–35%, customer prices 5–10%
- Alibaba Cloud: up to 34% on GPU-intensive services
- AWS: EC2 ~15% increase
- Microsoft Azure: 10–12%
Meanwhile, the move off public cloud is accelerating. 86% of CIOs are planning to move workloads off public cloud (Barclays CIO Survey). 37signals saved ~$2M/year leaving AWS. GEICO is repatriating 50%+ of 600+ applications after Azure costs went 2.5× over budget.
Nirvana's bet: flat ~$95/TB/mo, 20K IOPS included, zero egress. Same number next quarter.
Other Posts Worth Re-reading
- The Rise of Neoclouds — $20B in projected neocloud revenue by 2026. The three waves driving the shift: blog.nirvanalabs.io/rise-of-neoclouds-specialized-cloud
- The other 10% — The cloud was built for web apps, APIs, SaaS, and static sites. Databases, vector search, HFT systems, analytics — these break when performance dips. Nirvana rebuilt the cloud from the metal up for that 10%.
- Stop overprovisioning — Orgs are wasting 24% of annual cloud spend on overprovisioning and poor management. There's no reason for that.
Nirvana: Modern Cloud for Real-time Workloads
Ultra-fast Block Storage with High IOPS, powering blockchain, AI and real-time systems.
Learn more at Nirvana Labs
Nirvana Cloud | Pricing | Blog | Docs | Changelog | Twitter | Telegram | LinkedIn
Related Posts

Nirvana Instance Types are live: Compute Families Built for Your Workload
Nirvana Instance Types are live: Compute Families Built for Your Workload

Google Cloud Hyperdisk Balanced, Hyperdisk Balanced HA vs Nirvana Accelerated Block Storage (ABS)
Google Cloud (GCP) Hyperdisk Balanced, Hyperdisk Balanced HA vs Nirvana Accelerated Block Storage (ABS)

Product Update: Equivalent Code, API Key Permissions, SIEM, GitHub Auth, Volume Management, and More
Developer workflows, fine-grained security controls, and infrastructure flexibility for your cloud. TL;DR We've shipped a wave of features since the Organizations launch focused on making Nirvana faster to build with, more secure to operate, and more flexible for teams managing production infrastructure: * Equivalent Code Panel → see the equivalent code to create any resource as you build it in the dashboard * API Key Permissions → scope keys to specific projects and resource types with re

