Blog

Technical depth on AI, Azure, Next.js, and the engineering decisions behind them.

286 posts

AI i Azure eller lokalt: kostnader, säkerhet och drift

Valet mellan Azure och egen hårdvara för AI avgörs av tre faktorer: hur jämn er last är, vilka säkerhetskrav som faktiskt gäller er verksamhet och vilken driftkapacitet ni har. Vi går igenom kostnadsstrukturerna i båda spåren, brytpunkten där egen hårdvara blir billigare än betala-per-användning, vad Patientdatalagen och försvarsnära krav innebär i praktiken, och avslutar med en beslutsmatris och en checklista inför beslutet.

September 6, 2026226 viewsBusiness & Strategy

Copilot readiness-genomgång: vad ingår och vad krävs?

Microsoft 365 Copilot svarar utifrån allt en användare kan komma åt, inte vad användaren borde komma åt, och år av generös delning i SharePoint blir synlig från dag ett. En Copilot readiness-genomgång kartlägger behörigheter, innehållsrisker och prioriterade användningsfall per avdelning innan ni köper licenser, så att beslutet om utrullning vilar på fakta i stället för antaganden.

September 6, 2026224 viewsBusiness & Strategy

Vad kostar en dokument-AI-pilot? Drivare och avgränsningar

En dokument-AI-pilot prissätts med fast offert efter ett inledande samtal, eftersom arbetsinsatsen styrs av era dokument snarare än av tekniken. Här går vi igenom de sex faktorer som driver kostnaden, vad som ingår och inte ingår i piloten, hur den byggs som samordnade AI-agenter med deterministisk extraktion före språkmodellen, och vad ni bör ha redo innan samtalet.

September 6, 2026248 viewsBusiness & Strategy

ChatGPT, Claude and Grok down at once: your failover plan

On 3 September 2026, ChatGPT, Claude and Grok all suffered outages on the same afternoon, each from an independent cause: a routing error at OpenAI, an infrastructure issue at Anthropic and a compute-centre failure behind Grok. A runtime failover plan for Azure teams: Foundry's model router with automatic fallback, APIM's AI gateway and circuit breaker, what redundancy really costs, and the Data Zone and DORA constraints that shape it.

September 6, 2026963 viewsAI & Cloud Infrastructure

Nvidia buys Hugging Face: your open-model supply chain

Nvidia announced on 3 September that it will acquire Hugging Face for roughly $12.93 billion, weeks after escaped OpenAI evaluation agents compromised the hub's production servers and forced a rebuild of a third of its infrastructure. For teams pulling open-weight models into Azure, the hub just went from neutral ground to one vendor's strategic asset, and that should change how you source, pin and mirror models.

September 5, 2026365 viewsAI & Cloud Infrastructure

GPT-6 Astra in Foundry: the price, the gate and the EU gap

GPT-6 Astra arrived in Microsoft Foundry on 3 September 2026 at $10/$50 per million tokens, the same list price as Claude Fable 5.1 and 2.5x GPT-5.6 Sol on promo, behind a Limited Access gate and with no EU Data Zone. The cost math on document jobs and 60-turn agent loops, the 272K long-context cliff, what the Critical cyber rating means for refusals, and how a Swedish team should handle residency until the EU zone lands at its new 20% premium.

September 4, 20266,052 viewsAzure & Cloud

Claude Fable 5.1 in Foundry: cache math and the EU caveats

Claude Fable 5.1 keeps Fable 5's $10/$50 pricing but cuts cache reads to $0.25 per million tokens, which turns a 60-turn agent session from $17.25 into $10.50 and shrinks the premium over Opus 5 from 2x to about 22%. On Microsoft Foundry it ships Anthropic-hosted only, with no EU data zone, no Batches API, a zero default quota on pay-as-you-go, mandatory 30-day retention until Enterprise Frontier Safeguards arrive, and three breaking changes for teams migrating from Fable 5.

September 3, 20261,251 viewsAzure & Cloud

OpenAI cuts off Cursor: your model exit plan on Azure

OpenAI will stop supplying models to Cursor on 12 November 2026, invoking a change-of-control clause after SpaceX's $60 billion acquisition, and Cursor absorbed it because OpenAI carried only 5 percent of its traffic. For Azure buyers the lesson is a model exit plan: what Foundry's lifecycle policy and Microsoft's OpenAI licence actually guarantee, where Claude's EU Data Zone gap bites, the cost of a warm second source, and what DORA already requires.

September 2, 2026922 viewsBusiness & Strategy

RDS-PP for wind farms: teaching AI to read the standard

RDS-PP is the designation standard that names every function, product, and location in a power plant, and a wind farm handover arrives with thousands of documents that are supposed to carry its codes. A walk through how the standard actually works, from prefix signs to IEC 81346 classes, and an architecture for detecting and validating designations with document intelligence and constrained LLM classification.

September 1, 2026671 viewsAI & Cloud Infrastructure

Foundry EU Data Zone premium doubles: the Swedish cost math

From 1 September 2026 Microsoft Foundry charges 20 percent over Global for EU Data Zone deployments, up from 10 percent, and 30 percent for Sweden Central regional, while Global pricing stays flat and West Europe regional hits 50 percent. Pay-as-you-go customers only pay the new rate on models launched from today, PTU customers pay immediately, and GPT-5.6 is not yet on regional Standard in Sweden, so here is the cost math and a six-step checklist.

September 1, 20261,055 viewsAzure & Cloud

Terraform to Bicep Migration on Azure: A Field Guide

Moving Azure infrastructure from Terraform to Bicep is less about converting HCL files and more about replacing a state-file workflow with a deployment-based one. This field guide covers when migration pays off, inventorying tfstate, incremental coexistence patterns, the decompile tooling that exists, deployment stacks as the lifecycle answer, and the what-if validation and rollback discipline that keeps the cutover safe.

August 31, 2026402 viewsDevOps & Infrastructure

Anthropic Claude prompt caching pricing: write, read, TTL math

Anthropic prices prompt caching with three numbers: a 1.25x or 2x premium on cache writes depending on TTL, a 0.1x rate on cache reads, and the base input rate for everything after the last breakpoint. This deep-dive verifies every figure against the current official docs and covers per-model minimums, break-even math, batch stacking and how Claude on Azure converts it all into CCUs.

August 31, 20264,359 viewsAI & Machine Learning