What’s Now Generally Available?
June 2025, Microsoft released three new Azure VM families under the storage‑optimized “L‑series v4” umbrella:
- Laosv4 (dense storage, AMD EPYC Genoa)
- Lasv4 (AMD EPYC Genoa)
- Lsv4 (Intel Xeon Emerald Rapids).
Each boasts local NVMe SSD storage capacities up to 23 TB, major boosts in CPU, networking, remote storage performance, along with the introduction of Microsoft’s Azure Boost SSDs/Boost platform.
Key Innovations with Azure Boost
- Sub‑second servicing: Azure infrastructure updates now roll out with minimal disruption—up to 10× faster.
- Built‑in NVMe encryption: Hardware‑based encryption at rest.
- Enhanced network/Disk IOPS: Azure Boost enables substantial improvements to NVMe local and remote storage throughput.
Breaking Down the Families
Laosv4 (New dense‑storage line, AMD Genoa):
- vCPU range: 2–32
- Memory: 8 GB per vCPU
- Local NVMe: 720 GB per vCPU — density (~3× more storage/vCPU than Lasv3)
- Performance gains: ~20% uplift in CPU perf; ~3× better remote storage and networking/vCPU; ~50% lower $/TB.
- Ideal for: Scale‑out scenarios—caching, Elasticsearch, distributed file systems, big data, relational/NoSQL databases, data warehouses.
Lasv4 & Lsv4 (General storage‑optimized):
- vCPU range: 2–96
- Memory: 8 GB per vCPU
- Local NVMe: 240 GB per vCPU; up to 23 TB at top end
- CPU boost: Up to 20% over v3 series
- IOPS: Max ~6.6M on largest sizes
- Processor platforms:
- Lsv4: Intel 5th‑Gen Xeon (Emerald Rapids) — includes Intel® AMX and Total Memory Encryption (TME)
- Lasv4: AMD Genoa — includes Secure Memory Encryption, AVX‑512, VNNI, BF16 for AI workloads.
Real‑World Customer Impact
Customers like Silk, Sprinklr, Elastic, and Microsoft Azure Data Explorer report strong results:
- Silk: Saw >50% improvement in cost/performance using Laosv4
- Sprinklr: Transitioned smoothly from L16as_v3 to L16aos_v4; fewer CPUs yet equal or higher performance with improved disk utilization
- Elastic: Up to 30% higher Elasticsearch throughput on Laosv4 vs. Lsv3
- Azure Data Explorer: Gains in speed and cost-efficiency for telemetry, logs, beam analytics
Regional Availability
- Laosv4 & Lasv4: North Europe, West Europe, East US, East US 2, Central US, South Central US, West US, West US 2, West US 3, and UK South
- Lsv4: Australia East, Canada Central, Central US, Central India, East US, East US 2, Germany West Central, Japan East, Korea Central, South Africa North, South Central US, Sweden Central, Switzerland North, West Europe, West US 3
When to Choose Which
| Scenario | Best VM Family | Why |
|---|---|---|
| Ultra‑dense local storage per vCPU | Laosv4 | 720 GB NVMe per vCPU + AMD performance |
| Scale‑up moderate to high storage workloads | Lasv4 / Lsv4 | Faster Xeon or AMD, 240 GB/vCPU, up to 6.6M IOPS |
| AI inference/train on‑node | Lasv4 / Lsv4 | Support for AVX‑512, VNNI, BF16 (AMD) or AMX (Intel) |
| Encrypted memory performance | Lasv4 / Laosv4 (AMD: TSME), Lsv4 (Intel: TME) |
Get Started
- Choose your SKU:
- Provision via CLI, ARM/Bicep, or Portal
- Monitor performance: IOPS, latency, network throughput
- Right‑size to your workload’s vCPU/storage mix
Bottom Line
Azure’s L‑series v4 VMs set a new benchmark in storage-optimized computing—bringing next-gen CPUs, huge NVMe densities, hardware encryption, sub-second maintenance interruptions, and big performance uplifts.
Ideal for data-heavy workloads: databases, analytics, AI, search, caching, and more. With strong customer feedback and wide regional availability, the L‑series v4 refresh is ready for production-grade scale.
Based on Microsoft’s Azure Compute blog
P.S. Modern AI tool has been used for creating some of the content. Technical validation and proofing are done by the author.