UpCloud in 2026: an honest review for teams leaving AWS or Azure

UpCloud is a privately held Finnish cloud provider, founded in Helsinki in 2011, that runs its own KVM platform in 15 zones: ten in Europe, three in the US, one in Singapore and one in Sydney. It sells a compact managed cloud (PostgreSQL, MySQL, Kubernetes with a free control plane, S3-compatible object storage, load balancers), includes egress up to an allowance that grows with your servers, and answers support chats around the clock on every account: our production workload costs EUR 520.33 a month in Frankfurt, 48% of the AWS price. The weak points are redundancy and the contract. Each city is one data centre (only Helsinki has two), the database standby sits on another host in the same building, and the SLA pays account credit under a legal text that names no percentage and excludes the API and the control panel.
This article is one of a series comparing 24 European cloud providers with AWS and Azure; the overview puts them side by side. Everything below was checked on 15 and 16 October 2026 against UpCloud's documentation, terms, price catalogue and status page, Finnish company data and user reports, each at its source; the pricing and legal pages, which turn away automated readers, were read from Internet Archive copies of up to 7 October 2026. We did not open an account or run workloads.
At a glance
| UpCloud | |
|---|---|
| Headquarters and control | Helsinki. UpCloud Oy, privately held, not listed; owners not published. Disclosed investors: Connected Capital & Partners (Netherlands), Tesi (the Finnish state's investment company) and Inventure (Nordic venture capital) |
| Regions and zones | 15 zones, each one data centre: ten in Europe (two of them in Helsinki), three in the US, Singapore and Sydney. No region groups several zones the way AWS does |
| Workload A (starter) | EUR 17.93 a month in Frankfurt |
| Workload B (production) | EUR 520.33 on shared vCPUs, EUR 459.50 with the free single-node Kubernetes control plane. Every part is sold; there is no dedicated-vCPU plan, and the database standby sits in the same data centre |
| Egress | EUR 0.00 within an account-wide Fair Transfer allowance (1.5 TB a month for the workload A server, 7 TB per workload B server). Beyond it, bandwidth may be throttled to 100 Mbit/s, or you opt in to USD 0.01 per GB |
| Compute SLA | No percentage in the legal text; 99.999% for Premium and 99.99% for Starter on the pricing page. Account credit of 50 times the fee for the downtime, for outages over 5 minutes, claimed within 15 days |
| Included support | 24/7 engineers by chat and email, in English; incident response under 2 hours |
| Who can sign up | Businesses, by the terms: card, billing address and phone, sanctions screening, no ID document described. Prepaid; a new account starts as a 7-day trial limited to 2 cores |
What you can build
Status is general availability unless noted, from UpCloud's documentation and changelog.
| What you need | UpCloud | AWS | Azure |
|---|---|---|---|
| Virtual machines | Cloud Servers on AMD EPYC, all on shared hosts: Starter (previous-generation CPUs), Premium and Cloud Native (compute only) | EC2 | Virtual Machines |
| Dedicated hosts, bare metal | No public plan: dedicated hardware is Private Cloud, whole hosts from EUR 3,190 a month on quote. No bare metal | EC2 Dedicated Hosts, EC2 metal | Dedicated Host, BareMetal Infrastructure |
| GPUs | L4, L40S, RTX Pro 6000, H100 and B200, on demand and spot, in Helsinki (fi-hel2) only | EC2 G and P | NC and ND series |
| Storage | Block storage in three tiers, encrypted at rest at no charge; Managed Object Storage (S3 API) in six regions; File Storage (NFS) in seven zones since February 2026 | EBS, S3, EFS | Managed Disks, Blob Storage, Azure Files |
| Relational databases | Managed PostgreSQL (one to three nodes) and MySQL (8.4 LTS) | RDS | Azure Database for PostgreSQL and MySQL |
| Other data services | Valkey (replaced Managed Redis in October 2024) and OpenSearch | ElastiCache, OpenSearch Service | Azure Cache for Redis |
| Kubernetes | UKS, CNCF certified for 1.33 to 1.35. Development control plane free (one node); Production EUR 60 a month (three API servers, three-member etcd) | EKS | AKS |
| Networking | Private networks across zones, NAT and VPN gateways, firewall, load balancer for HTTP(S) and TCP | VPC, NAT gateway, VPN, ELB | VNet, NAT gateway, VPN Gateway, Load Balancer |
| CDN | UpCloud CDN, "powered by bunny.net": documented, but "Releasing soon" on the pricing page | CloudFront | Front Door |
| Access control | Subaccounts limited by server, storage, network or tag; Workspaces with roles and single sign-on, in open beta and outside the SLA | IAM | Entra ID, RBAC |
| Monitoring | Metrics for databases and load balancers, audit logs; no general logging or metrics service | CloudWatch | Azure Monitor |
| Backup | Simple Backups (daily to yearly plans), plus on-demand backups | AWS Backup | Azure Backup |
Missing outright: functions, serverless containers, queues, a key management service or secrets manager, authoritative DNS, MongoDB, Kafka, managed AI model APIs and a container registry (the price catalogue has "beta" registry items, but there is no product). No DDoS protection is described in the documentation or on the pricing page, and the SLA excludes "hostile actions of third parties, such as denial-of-service attacks".
Object Storage lacks bucket replication, website hosting, bucket logging and SSE-KMS, and Object Lock is "Coming soon". Access control is thin: subaccounts limit which resources a user can touch, with no action-level policy language, and the role-based Workspaces model "is not yet recommended for production environments".
The tooling is current: a partner-tier Terraform provider (v5.47.0, 12 October 2026), a CLI, Go and Python SDKs, Pulumi, Packer, Ansible and Crossplane. API tokens arrived only on 18 September 2025; before that, scripts used an account's username and password. On Kubernetes, node updates are yours: "UKS does not modify, update, or configure the data plane after creation".
What it costs
UpCloud list prices in EUR, VAT excluded, from the pricing calculator's catalogue and the pricing page of 7 October 2026, checked 15 and 16 October 2026, in Frankfurt (de-fra1) at 730 hours a month. Cloud Servers are billed by the hour up to 28 days a month, so a full month costs the listed monthly price; everything else is billed by the hour with no cap.
Workload A, a starter setup: one VM with 2 vCPU and 4 GB, a 40 GB disk, one IPv4 address, 1 TB of egress and 100 GB of object storage.
| Line | Choice | EUR a month |
|---|---|---|
| VM | STARTER-2xCPU-4GB, 2 shared vCPU, 4 GB, 30 GB Standard storage, IPv4 included | 12.00 |
| Disk to reach 40 GB | a second 10 GB Standard device | 0.86 |
| Public IPv4 | included | 0.00 |
| 1 TB egress | within the plan's 1.5 TB Fair Transfer share | 0.00 |
| 100 GB object storage | sold in 250 GB units | 5.07 |
| Total | 17.93 |
AWS charges EUR 112.37 and Azure EUR 116.15 for the same in Frankfurt; on AWS the 1 TB of egress alone is USD 81.00 of a USD 125.95 bill.
Workload B, a small production system: three VMs of 4 vCPU and 16 GB with an IPv4 address each, 300 GB of SSD block storage, a load balancer, PostgreSQL with 2 vCPU, 8 GB and 100 GB as primary plus standby, 1 TB of object storage with 1 million PUT and 10 million GET requests, a Kubernetes control plane and 2 TB of egress.
| Line | Choice | EUR a month |
|---|---|---|
| 3 VMs | PREMIUM-4xCPU-16GB, shared vCPU, each with 150 GB MaxIOPS storage and one IPv4 | 276.00 |
| 3 public IPv4 | included | 0.00 |
| 300 GB block storage | included: the three plans carry 450 GB | 0.00 |
| Load balancer | Production Small, two nodes | 30.42 |
| PostgreSQL HA | 2 vCPU, 8 GB, two nodes with 100 GiB each, 15 days of point-in-time recovery | 132.80 |
| 1 TB object storage | four 250 GB units; no request fees | 20.28 |
| Kubernetes control plane | Production, three API servers | 60.83 |
| 2 TB egress | within the account's 21 TB share | 0.00 |
| Total | 520.33 |
AWS charges EUR 1,082.61 for workload B on m7i.xlarge and Azure EUR 1,091.88 on D4s_v5, so UpCloud costs 48% of either. Two lines make most of the gap: AWS bills the 2 TB of egress at USD 171.00 and its HA database at USD 320.13 with storage, where UpCloud charges nothing for the egress and EUR 132.80 for the database. The free Development control plane brings the total to EUR 459.50, with one API server, "no redundancy" and nothing to say whether the SLA covers it.
The total follows the calculator, which bills the second database node at 90% of the compute and storage price; the pricing page lists the compute at EUR 60 rather than 59. Worker nodes are not included, as everywhere in this series.
Egress and IPv4. Public egress has cost EUR 0.00 per GB since 7 May 2024, within a Fair Transfer Limit "determined by the amount of cloud resources on your account": each server plan adds a fixed amount (1.5 TB for STARTER-2xCPU-4GB, 7 TB for PREMIUM-4xCPU-16GB), shared across all zones, while object storage, databases, load balancers and most GPU plans add nothing. Past the limit UpCloud "may throttle the bandwidth on the cloud services down to 100 Mbit/s", with "no excess fees" unless you opt in to paid transfer at USD 0.01 per GB (EUR 10 per 1,000 GB in the EUR catalogue). AWS charges EUR 80.30 per TB beyond its free 100 GB, Azure EUR 77.62. The first IPv4 address comes with every Starter and Premium plan; another costs EUR 3.47 a month.
Billing. UpCloud is prepaid, and the balance is a hard ceiling. When it runs out, all servers are shut down; after a month they are removed and their IP addresses released; after two months "Storages will be removed. Data will be lost permanently." Credit is non-refundable, except for a 30-day money-back guarantee on the first payment, up to EUR 500. Starter and Premium servers are billed whether they run or not. Payment is by card (PayPal, Apple Pay and Google Pay for manual top-ups only); no bank transfer or SEPA direct debit is listed. The terms of 3 July 2026 describe a post-pay model "charged monthly in arrears" for customers who register once it launches, but the documentation still describes prepaid credit only.
GPUs, per GPU-hour in Helsinki, the only zone that has them: L4 EUR 0.58, L40S EUR 1.11, RTX Pro 6000 EUR 1.65, H100 EUR 1.79 and B200 EUR 4.50, with no H200. That is the lowest on-demand H100 price in this series; AWS does not sell the H100 in Frankfurt and charges EUR 6.57 in Stockholm. Spot servers (H100 EUR 1.78) can be reclaimed with 30 seconds' notice and have no SLA. UpCloud has announced "New H100 GPU Server pricing starting from October 20th", and the same for the B200, without saying what changes; the catalogue of 16 October still showed the prices above, and the pricing page of that day still gave no new figures. Each account may run "a limited number of GPUs, based on its account history", and more takes a support request.
Price changes. On 15 April 2026 the Starter and Premium families replaced Developer and General Purpose, which can no longer be deployed. Like for like, the new plans cost less but include less disk: a 2 vCPU, 4 GB Developer server with 60 GB was EUR 18 a month, the Starter equivalent with 30 GB is EUR 12. Whether existing servers keep their old prices is not stated. Between the pricing pages of 19 May and 7 October 2026 the monthly GPU estimate moved from a 28-day cap to a 30-day month at the same hourly rates, so a month of one H100 went from EUR 1,203 to EUR 1,289, about 7% more (our inference from two page copies; we found no announcement). We found no increase in the list price of any product that can still be ordered.
Reliability
What a zone is. At UpCloud a zone is one data centre, and nothing groups zones into a region; AWS and Azure each have three zones in Frankfurt, in separate data centres kilometres apart. Helsinki is the exception, with two zones in buildings run by different operators (Equinix and Telia); whether they have independent power and network paths is not stated. Everywhere else, surviving the loss of a building means a second city. Inside a zone the redundancy is at host level: servers move automatically to a healthy host when one fails, and block storage sits on a storage network mirrored across "two separate storage backends".
The database standby. Managed PostgreSQL HA nodes "are created on physically separated backend hosts" in the same zone; no document offers a standby in another building. The docs describe "real-time replication" without saying whether it is synchronous, and give no failover time. If every node fails at once, the primary is restored from backup, with data loss "typically ... confined to either a five-minute time frame or a single WAL file". AWS and Azure put a synchronous standby in another zone and state 60 to 120 seconds for failover.
The SLA. It is part of the terms of service effective 3 July 2026, and the legal text contains no percentage: the 99.999% and 99.99% figures appear on the pricing page only.
- It covers "all unscheduled interruptions ... that last longer than five (5) minutes", calculated per affected server or service.
- The credit is "50 (fifty) times UpCloud's service fees charged for the interrupted Service-item" for the time it was down, capped at that item's fees for the previous 30 days per interruption and at 2.5 times its average monthly fee per 12 months. It is account credit, not cash.
- You claim by email within 15 days after the failure is corrected.
- Not covered: "APIs of the Service" and "the Control Panel", so an outage that stops you deploying or scaling earns nothing. Denial-of-service attacks, insufficient account credit, announced maintenance and spot servers are excluded.
- The SLA is the sole remedy, and "UpCloud's decision on the issue will be binding ... until and unless a competent court rules otherwise". Outside it, liability is capped at six months of fees.
In numbers: a EUR 92 Premium server down for 2 hours earns EUR 13.69 of credit, and a 24-hour outage reaches the cap of EUR 92. For a short outage that beats AWS's single-instance SLA, which waives each hour with more than six minutes of downtime and adds a 10% credit only below 99.5%. But AWS commits to 99.99% for EC2 across two or more zones, and Azure to 99.99% across zones with 60 days to claim; UpCloud has no second zone to spread across.
Status page and incidents. status.upcloud.com lists incidents per zone and service, with history back to August 2015. None of its 50 most recent incidents has a postmortem attached, including the one on 11 August 2026 whose update said "We are compiling a detailed post-mortem".
| Date | Scope | Duration | Cause |
|---|---|---|---|
| 22 Dec 2025 | Networking in UK-LON1, NL-AMS1 and SE-STO1, opened within 3 minutes of each other | about 3 hours 10 each | not stated (London, Amsterdam, Stockholm) |
| 20 to 21 Jan 2026 | TLS certificates failing for new load balancers, Kubernetes clusters and Object Storage, all zones | about 17 hours | certificate issuance (incident) |
| 7 Jul 2026 | Object Storage in Frankfurt (EUROPE-2) | 1 hour | not stated (incident) |
| 11 Aug 2026 | New Cloud Server deployments failing in all 15 zones; customers asked not to start, restart or shut down servers | about 2 hours | not stated; the promised post-mortem not found (incident) |
| 25 to 26 Aug 2026 | DE-FRA1 storage backend, "affecting some resources including cloud servers" | about 7 hours | not stated (incident) |
| 6 to 13 Oct 2026 | SG-SIN1 Object Storage: latency spikes and HTTP 503 errors | fixed after about 11 hours, resolved after a week | root cause "identified", not published (incident) |
Failures in one zone stayed in that zone. What crossed zones came from shared layers (certificates, deployments) and one networking event in three cities at once. API and control-panel incidents, such as those of 3, 21 and 22 November 2025, earn no credit.
Support
| Plan | Price | Service request | Incident | Extras |
|---|---|---|---|---|
| Essentials (default) | included | under 4 hours | under 2 hours | email and live chat, 24/7 |
| Advanced | on request | under 1 hour | under 30 minutes | priority queue |
| Enterprise | on request | under 10 minutes | under 5 minutes | dedicated support members, a customer success manager |
Support is in English only, by chat and email from in-house engineers, with no phone line, and help with your own software may be refused. The targets are on the support page, not in the terms. Even so, the free tier has an incident target, which AWS Basic (no technical cases) and Azure Basic (no technical response targets) do not. A one-hour target for a production outage costs the greater of USD 29 or 9% of spend on AWS Business Support+, or USD 100 a month on Azure Standard.
Getting an account
The terms sell to businesses only: the service is "not fit or intended for use by ... consumers or private persons for their personal use". Signing up takes a name, email address, phone number, card and billing address. UpCloud screens signups against sanctions lists and serves no one in Russia, Belarus, Crimea, Iran, North Korea, Cuba or Syria. None of the documents we read mentions an ID document or selfie check.
A new account is a trial (7 days by the documentation; the pricing page says "a free 30-day trial") with 2 cores and 4 GB of servers, one single-node database, one Development load balancer and one Development Kubernetes cluster, and no GPUs. A deposit of at least 10 in the account currency ends it. Port 25 stays blocked until support opens it.
Workload B on day one: not on the trial, which has a sixth of the cores it needs and no HA database. After the deposit, we cannot say: UpCloud publishes no default limits for a paid account, and support raises them on request. On AWS the default of 5 vCPUs for standard instances is below workload B's 12; Azure's 20 cores per region fit on paper.
Two clauses matter later. UpCloud may suspend an account on "reasonable grounds to believe" the terms were breached, with "no liability" for doing so. And after non-payment it may terminate and delete after "a reasonable time ... (at a minimum fourteen (14) days)", much shorter than the two months the documentation describes for an empty balance.
Compliance and data protection
| Framework | Status |
|---|---|
| ISO/IEC 27001:2022 | Certificate seen: no. 13851-01 from Inspecta Sertifiointi Oy (Kiwa), issued 20 October 2023, valid until 20 October 2026. No renewed certificate was on the site on 16 October |
| CISPE Data Protection Code of Conduct | Register entries seen: "Controlled adherence" for ten services, running to June and July 2027 |
| BSI C5, ISO 27017, ISO 27018, SOC 2, PCI DSS, SecNumCloud, HDS, TISAX, ENS | Not claimed for UpCloud itself |
The ISO certificate covers UpCloud's cloud service, with offices in Helsinki and Singapore; check for a renewal after 20 October before relying on it. The other certificates on UpCloud's data centres page belong to the colocation operators.
The Data Processing Agreement is part of the terms. You choose the data centre, and "UpCloud shall not move the Relevant Personal Data from the selected data centre unless explicitly instructed". The only sub-processors are UpCloud's four wholly owned subsidiaries in Singapore, the UK, the US and Australia, each for the data centres in its own country; the nine EU and EEA zones "are operated directly by UpCloud Oy, and no subprocessors will be utilised". Changes come with 30 days' notice and a right to object and terminate.
A header check on 15 October showed the website and control panel served through Cloudflare, and the CDN is "powered by bunny.net"; neither is named as a sub-processor.
Who controls it
You contract with UpCloud Oy (business ID 2431560-5) under Finnish law, with arbitration in Helsinki in English. Founder Joel Pihlajamaa, CTO from the start, became CEO again on 2 July 2026, succeeding Arno Schäfer, CEO only since 31 March 2025; webhosting.today reports no reason for the short tenure.
The shareholdings are not public: a Finnish Oy's owners are not in the trade register, and no source we read says whether the founder or the investors hold a majority. In January 2020 UpCloud raised EUR 18 million "of equity and debt funding" from Connected Capital & Partners, Tesi and Inventure. We found no later round, acquisition or exit, and no non-European parent or shareholder.
Finnish filings, as reported by the credit-data service Asiakastieto, show UpCloud Oy's revenue rising from EUR 16.1 million in 2021 to EUR 27.3 million in 2025, an operating loss in each of those years (EUR 1.3 million in 2025) and an equity ratio of 14% in 2025.
The platform is UpCloud's own (KVM and a storage network, not OpenStack); the buildings are not. Every zone is in a colocation facility: Equinix in Helsinki 1, Warsaw, Stockholm, Singapore and Sydney; Digital Realty in Frankfurt, Copenhagen, Madrid and Amsterdam; Telia in Helsinki 2, Verne in London, Green Mountain in Stavanger and CoreSite in the three US zones.
Leaving
Egress on the way out is free within the Fair Transfer allowance (21 TB a month for workload B), and Object Storage charges no request fees. The terms let you leave on 5 days' notice and contain an EU Data Act switching clause: a 30-day transitional period, extendable once up to 6 months, plus up to 30 days to retrieve data, with fees payable while resources run. At the end UpCloud deletes "all Customer Data" and "shall not be responsible for the integrity of the exportable data".
Most of what you build is standard: KVM images, upstream Kubernetes with an open-source CSI driver, the S3 API with the gaps above, and open-source PostgreSQL, MySQL, Valkey and OpenSearch. Lock-in sits in the private network setup, the Kubernetes defaults (UpCloud "strongly" advises against replacing its Cilium network plugin) and the UpCloud-specific Terraform resources; there is no proprietary serverless or database API, because none is offered.
What users report
We read every Hacker News comment mentioning UpCloud from October 2023 to October 2026 (106 hits) and one customer's write-up; Reddit could not be searched. UpCloud is discussed far less than Hetzner or OVHcloud.
Support is what users praise most. Users report "top customer service" (48542691, June 2026), that "their support team is very responsive" (39128618, January 2024) and that "the people there are super nice" (45137168, September 2025), among five accounts in that period (46437952, 47675387). The same searches found no complaint about support.
Dearer than Hetzner for raw compute. Users report that "They aren't cheap compared to DigitalOcean/Linode" (43319119, March 2025) and compare 48 cores at EUR 1,364 a month on UpCloud with EUR 288.49 on Hetzner (43458843, March 2025); both predate Hetzner's 2026 price increases. At the small end, one writes "Upcloud has the cheapest I found, you can get a full VPS for less than $4" (50017783, October 2026).
Free egress is an allowance. Users point out that "the cap after 24 TB of free bandwidth/month is at 100 mbps" (46654024, January 2026) and that the limits "are comparable to other hosts like DigitalOcean" (43458846, March 2025; answered with the policy's "no excess fees", 43459396).
Object Storage slowed down for one customer. Healthchecks.io moved its object storage from OVHcloud to UpCloud in 2024 and wrote on 17 April 2026 that "over time, the performance of UpCloud object storage deteriorated as well. There were periods where all operations would become slow and hit our timeout limits" (post, discussion); it now runs its own storage. That is one customer's experience, not a measurement, and another user wrote "Very happy with them" the same month (47712388). The status page lists five Object Storage incidents in 2026.
Fast CPU and disk. Users report faster CPU and disk I/O than at the providers they benchmarked (43319119, March 2025) and volumes that "easily hits 10k IOPS" (47824138, April 2026). We measured nothing.
Thin access control. Users report "You will not find equivalent IAM or BigQuery in indie clouds" (46336813, December 2025) and tools that signed in with account credentials (44339109, June 2025).
Steady single servers. Long-time customers report that "the available features work very reliably (for half a decade now). There are a few rough edges" (43459317, March 2025) and single servers "more reliable than multiregion with redundancy setups with other providers like Fly" (47818591, April 2026).
Who it suits, and who should look elsewhere
UpCloud suits small and mid-sized teams that want Kubernetes, PostgreSQL and S3 storage under a Finnish contract; services that send a lot of traffic out within the fair share; teams that value support they can reach at any hour; and products that need the US or Asia from the same European provider. Its H100 and L40S prices are low for workloads that can sit in Helsinki.
Look elsewhere if you need zones in separate buildings inside one region, or a database standby in another building; if you build on functions, queues or event-driven services; if you need fine-grained IAM, a key management service or certification beyond ISO 27001 (C5, SOC 2, PCI DSS); if you need dedicated vCPUs without a Private Cloud contract, GPUs outside Helsinki or the lowest price per core; or if you need post-paid invoices rather than a prepaid balance.
Sources
Checked on 15 and 16 October 2026. Pages marked "archived" were read from the Internet Archive, because upcloud.com turns away automated readers.
Pricing
- Pricing page, archived 7 October 2026 and archived 19 May 2026
- Pricing calculator, with the embedded price catalogue
- Fair Transfer Policy, archived 4 October 2026, network transfer and network billing
- Block storage billing, CDN billing
- Account balance, payment methods
- Support tiers, archived 22 September 2026
- AWS and Azure figures: AWS Price List bulk API and Azure Retail Prices API, converted at the ECB rate of 15 October 2026
Terms and SLA
Documentation
- Locations, changelog
- Cloud Server plans, Cloud Servers FAQ, system architecture, storage system
- S3 compatibility, Object Storage availability, File Storage availability
- PostgreSQL high availability, PostgreSQL configurations
- UKS control plane, UKS supported versions, UKS shared responsibility
- GPU availability, GPU spot
- DNS, subaccounts, Workspaces, Private Cloud
- Account requirements, free trial, support, API 1.3 accounts reference
- Terraform registry, UpCloudLtd on GitHub, CNCF k8s-conformance
Status and incidents
- Status page and history; incident links in the table above
Compliance
- ISO 27001 certificate
- CISPE Code of Conduct public register
- Compliance and security, archived 4 May 2026 and data centres, archived 22 September 2026
Company and ownership
- Tesi portfolio entry
- Asiakastieto: UpCloud Oy financial figures
- Tech.eu on the EUR 18 million round, 21 January 2020
- webhosting.today on the founder's return as CEO, 3 July 2026
User reports
- Healthchecks.io blog on its move to self-hosted object storage, 17 April 2026, and its Hacker News discussion
- Hacker News comments, by item number at news.ycombinator.com: 39128618 (January 2024), 43319119 (March 2025), 43458843, 43458846, 43459317 and 43459396 (March 2025), 44339109 (June 2025), 45137168 (September 2025), 46336813 and 46437952 (December 2025), 46654024 (January 2026), 47675387, 47712388, 47818591 and 47824138 (April 2026), 48542691 (June 2026), 50017783 (October 2026)
Data & Cloud
New guides and rule changes by email
When a rule changes, a deadline moves or we publish a new guide, we send it by email.