Category
AI & Cloud Infrastructure
AI hardware, data-centre infrastructure, GPUs, HBM memory supply, and the cloud + on-prem architecture that powers enterprise AI at scale.
-
AMD EPYC Venice: 256 cores, 3.30x rack throughput and the 2027 refresh call
AMD put EPYC Venice into production on TSMC 2nm in May 2026, then spent July announcing where it lands first: two new Azure VM series and Anthropic's 2 gigawatt Helios deployment. Here is how to read the numbers before
Manu Shukla -
GPU cloud pricing in India 2026: H100, H200 and B200 rates compared
Renting an H100 in India starts near ₹217/hour in 2026; IndiaAI subsidises compute to about ₹67-92/GPU-hour. Compare Indian providers, neoclouds and hyperscalers on H100, H200 and B200, with the buy-versus-rent maths.
Manu Shukla -
Stateful AI agents on Kubernetes: the 2026 Agent Sandbox and GPU scheduling playbook
AI agents are long-running, stateful and GPU-hungry, and standard Deployments fight them. Here is how Agent Sandbox, warm pools and Kubernetes DRA change the design in 2026, with configs and comparison tables.
Manu Shukla -
3 enterprise agent runtimes compared in 2026: Alibaba Agent Native Cloud, AWS AgentCore, Azure Foundry
Alibaba Cloud unveiled Agent Native Cloud at WAIC on 18 July 2026, adding AgentLoop and AgentTeams to AgentRun. Here is how the three big agent runtimes compare on billing, isolation, observability and lock-in.
Manu Shukla -
AMD Helios vs NVIDIA Vera Rubin: 7 spec traps in 2027 AI rack quotes
AMD and NVIDIA both ship 72-accelerator AI racks in 2026, and almost none of their headline numbers are measured the same way. A line-by-line normalisation of the two vendor-published spec sheets, and the seven
Manu Shukla -
Bedrock Agents Classic locks out new accounts on 30 July 2026: the AgentCore migration map
AWS put Bedrock Agents Classic into maintenance mode on 30 July 2026. Existing agents keep running, but new accounts get a 403 and the model catalog freezes. Here is the capability-by-capability migration map
Manu Shukla -
Qwen3.8-Max: 2.4T parameters, 1.2TB of VRAM, and the number Alibaba has not published
Qwen3.8-Max carries 2.4 trillion total parameters and needs roughly 1.2TB of VRAM at 4-bit. The number that decides whether you can run it, the active-parameter count, is the one Alibaba has not published.
Manu Shukla -
Azure Cobalt 200: what the 50% claim actually measures, and the 3 numbers that decide an Arm migration
Microsoft, AWS and Google all publish Arm performance gains, and all three use a different baseline. Cobalt 200 is measured against Cobalt 100, Graviton4 against Graviton3, and Axion N4A against x86. Here is how to read
Manu Shukla -
Bedrock Managed Knowledge Base vs self-managed RAG in 2026: 6 connectors, 8 regions, no Mumbai
AWS made Amazon Bedrock Managed Knowledge Base generally available on 17 June 2026. It removes the vector store, the parsing pipeline and the retrieval orchestration. It also removes some control, and it is not in
Manu Shukla -
Karnataka's data centre policy in 2026: what was announced on 15 July vs what is already in force
Karnataka said on 15 July 2026 that a new data centre policy is coming. A 2022-2027 policy is already published, and the change that actually moved the economics this year came from the Union Budget.
Manu Shukla -
Amazon Kendra closes to new customers on 30 July 2026: 6 features Bedrock Knowledge Bases cannot replace
AWS put Amazon Kendra into maintenance mode on 30 June 2026 and closes it to new customers on 30 July. Migrating to Bedrock Managed Knowledge Base drops you from 32 connectors to 7.
Manu Shukla -
Azure HorizonDB vs Flexible Server in 2026: 8 limits that decide it for you
Microsoft now runs two managed Postgres services on Azure. HorizonDB's preview limitations list decides most of these choices before any benchmark does, and one of those limits rules it out for Indian workloads entirely.
Manu Shukla -
Kubernetes DRA hits GA in 1.35: migrating GPU workloads off the device plugin in 2026
Dynamic Resource Allocation is GA in Kubernetes v1.35, but NVIDIA's upstream driver still ships its GPU kubelet plugin disabled by default. Here is what actually works today, with manifests.
Manu Shukla -
OpenTelemetry GenAI conventions: zero stable gen_ai attributes in 2026, and what to ship anyway
Vendor posts say OpenTelemetry's GenAI client spans went stable in early 2026. The specification says Development on every page, which sits below Alpha. Here is what that means for your instrumentation.
Manu Shukla -
The 2026 AI compute crunch: a capacity-planning playbook for GPU and power limits
Gartner expects power to constrain 40% of AI data centers by 2027, and GPU lead times run 36 to 52 weeks. This 2026 playbook covers capacity forecasting, spot vs reserved, neo-clouds and procurement governance.
Manu Shukla -
NVIDIA B200 vs H100 in 2026: why the pricier GPU is cheaper per token
A 2026 cost-per-token comparison of the NVIDIA B200 and H100 for LLM inference, with MLPerf v6.0 throughput, real cloud hourly rates, the arithmetic that makes the pricier chip cheaper per token, and the cases where
Manu Shukla -
Kubernetes 1.35 is the last release supporting containerd 1.x: a 2026 upgrade guide
Kubernetes 1.35 is the last version to support containerd 1.x; from 1.36 containerd 2.0 is the floor. Here is what breaks in your node config, how to find at-risk nodes, and how GKE, EKS and AKS differ.
Manu Shukla -
Build a web-grounded AI agent on Amazon Bedrock AgentCore: 2026 guide with code
Amazon Bedrock AgentCore is AWS's platform for production AI agents. Its Web Search tool reached GA on 17 June 2026 at $7 per 1,000 queries. Build a cited, web-grounded agent over MCP, with code and pricing.
Manu Shukla -
Connect Claude Code, Cursor and Codex to Amazon Bedrock's new console (2026)
Amazon Bedrock's new Mantle console connects Claude Code, Codex, Cursor and more to GPT, Claude and open-weight models with IAM or a Bedrock API key. Here is how to wire each coding agent through Bedrock.
Manu Shukla -
Local LLMs in production (2026): vLLM vs Ollama vs LM Studio, benchmarked
vLLM, Ollama and LM Studio solve different problems. Red Hat measured vLLM at 793 tokens/s versus Ollama's 41, and local inference costs about $0.18-0.29 per million tokens. Throughput, GPU cost, and when each wins.
Manu Shukla -
4 RAG embedding models compared in 2026: Nemotron 3 Embed vs OpenAI, Cohere and open-weight
A senior-engineer comparison of RAG embedding models in 2026: NVIDIA Nemotron 3 Embed, OpenAI text-embedding-3, Cohere Embed v4 and open-weight Qwen3, with RTEB benchmarks, pricing and self-host break-even math.
Manu Shukla -
AKS on bare metal in 2026: an edge Kubernetes play, not the GPU story you read about
Microsoft shipped AKS on bare metal to public preview on 2 June 2026. The validated hardware list runs from a $2,762 rugged mini PC to an ASUS NUC, and nothing on it trains a model. Here is what the feature actually
Manu Shukla -
Inkling's 975B open weights: what self-hosting really costs vs the API in 2026
Thinking Machines released Inkling on 15 July 2026 under Apache 2.0: 975B parameters, 41B active. We priced the GPUs against the Tinker API using AWS and vendor figures, and the break-even is brutal.
Manu Shukla -
Agent session isolation compared: AWS microVM, Azure Foundry, Google sandbox and Anthropic Managed Agents
Four vendors rebuilt their agent runtimes around the session in 2026, and they agree on routing but disagree on the compute primitive underneath. A comparison of AWS AgentCore, Azure AI Foundry, Google Agent Engine
Manu Shukla -
Cloud Run cross-region failover went GA on 29 June 2026: the service health setup that survives a regional outage
Cloud Run now fails over between regions automatically using readiness probes and serverless NEGs. Here is the working gcloud setup, plus the two defaults that silently break failover.
Manu Shukla -
Ingress NGINX is retired: the Gateway API migration, 34 annotations mapped (2026)
Ingress NGINX maintenance ended in March 2026: no releases, no bugfixes, no security updates. ingress2gateway v1.1.0 maps 34 ingress-nginx annotations to Gateway API, refuses 7 more, and there are two NGINX providers
Manu Shukla -
Kubernetes gang scheduling in v1.36: the Workload API rewrite your v1.35 manifests will not survive
Kubernetes v1.35 introduced the Workload API and native gang scheduling. Four months later v1.36 replaced the API version entirely. Here is the current shape, the migration, and the GPU maths that justifies it.
Manu Shukla -
Kubernetes OCI image volumes hit stable in v1.36: ship model weights without rebuilding images
Kubernetes image volumes reached stable in v1.36 after starting as alpha in v1.31. You can mount model weights straight from an OCI registry, read-only, with no rebuild of your model-server image.
Manu Shukla -
Running AI on Kubernetes in 2026: fix 5% GPU use with a managed platform
Kubernetes is the default home for AI, yet GPUs often run at 5% utilization, wasting over $100,000 a year. How a managed platform fixes GPU scheduling and cost for Indian enterprises running AI.
Manu Shukla -
Google's Gemini Enterprise agent platform in 2026: build, govern, and the gap
At Google Cloud Next '26, Vertex AI became the Gemini Enterprise agent platform. With 96% of firms running AI agents but only 12% able to govern them, here is what it changes and how it compares to Microsoft and AWS.
Manu Shukla -
Meta entered the cloud market in 2026: what it means for your AI budget
Bloomberg reported on July 1, 2026 that Meta is building a cloud business to rent out AI compute and host Llama models. A fourth serious cloud vendor changes the math on price, negotiation, and lock-in for enterprise
Manu Shukla -
SharePoint RCE CVE-2026-45659: a 7-step patch-and-harden checklist for 2026
CISA added SharePoint RCE CVE-2026-45659 (CVSS 8.8) to its Known Exploited Vulnerabilities catalog with a July 4, 2026 federal deadline. Here is the defensive patch-and-harden checklist for on-premises SharePoint Server.
Manu Shukla -
Manage Azure VMs from AWS in 2026: Systems Manager multicloud, now free
AWS removed its advanced-instances tier on June 30, 2026, so you can manage Azure VMs from AWS Systems Manager at no per-node charge. A setup walkthrough, an honest SSM versus Azure Arc comparison, and the App Manager
Manu Shukla -
Claude hits GA on Microsoft Foundry: what Consumption Units mean for your Azure bill in 2026
Claude reached general availability in Microsoft Foundry in late June 2026, billed in Claude Consumption Units where 100 CCU equals $1. How CCU conversion works and what it changes for your Azure bill.
Manu Shukla -
GPT-5.6's 3 tiers reset enterprise AI inference costs in 2026
OpenAI's GPT-5.6 (Sol, Terra, Luna) went public on July 9, 2026, with Luna at $1/$6 per million tokens. We compare the new tiers to Claude Sonnet 5 and Gemini, and the levers that cut inference bills 60 to 80 percent.
Manu Shukla -
NVIDIA's Vera Rubin hits the cloud: what the new instances mean for AI workloads in 2026
NVIDIA's Vera Rubin platform entered full production on June 1, 2026 and ships to eight cloud partners this fall, promising 5x inference over Blackwell and up to 10x lower cost per token. Here is what infrastructure
Manu Shukla -
3 Chinese open models that cut enterprise AI bills 60-90% in 2026
US firms routed up to 46% of their OpenRouter tokens to Chinese open models in 2026 because they cost 60-90% less. Here is how DeepSeek, Qwen and GLM compare with OpenAI and Google on price, quality and risk.
Manu Shukla -
2026: AFM Cloud Pro vs Gemini frontier for your enterprise AI stack
Apple runs AFM 3 Cloud Pro on Nvidia Blackwell GPUs inside Google Cloud, refined by Gemini frontier models, while reportedly paying Google about $1 billion a year. What Apple's Google-Nvidia architecture means
Manu Shukla -
5 architecture decisions behind AFM 3 Cloud Pro on Nvidia GPUs (2026)
Apple extended Private Cloud Compute to Nvidia Blackwell GPUs in Google Cloud to run AFM 3 Cloud Pro, its ~1.2T-parameter server model. Here are 5 architecture decisions behind it, for private-cloud AI teams.
Manu Shukla -
AFM Cloud Pro on Nvidia GPUs: what it means for enterprise in 2026
Apple's AFM 3 Cloud Pro runs a frontier model on Nvidia GPUs in Google Cloud under Private Cloud Compute. What that architecture means for enterprise teams, and the observability gap it leaves.
Manu Shukla -
5 architecture decisions behind Apple Intelligence on Nvidia GPUs (2026)
At WWDC 2026 Apple extended Private Cloud Compute to Google Cloud on Nvidia Blackwell GPUs. Here are five architecture decisions CTOs should take from Apple's private-AI design.
Manu Shukla -
Apple's system orchestrator: how 3-tier AI routing works in 2026
At WWDC 2026 Apple detailed a system orchestrator that decides what runs on-device, what goes to Private Cloud Compute, and what reaches a cloud model. How the routing works, for engineers.
Manu Shukla -
Healthcare AI in India 2026: 6 steps to stay CDSCO and DPDP compliant
India is moving healthcare AI from pilots to regulated deployment under CDSCO and DPDP. Here are six steps to classify, license, and run an AI medical-software product compliantly in 2026.
Manu Shukla -
OpenAI Weighs Drastic Token Price Cuts as Anthropic Hits $47B (2026)
OpenAI is weighing drastic token price cuts to win users from Anthropic per WSJ June 11, 2026. The competitive context, numbers, and what enterprise buyers should expect.
Manu Shukla -
SK Hynix–Nvidia Multi-Year AI Factories Deal: What It Means (2026)
SK Hynix and Nvidia signed a multi-year technology partnership on June 7, 2026 to co-develop next-generation memory for Vera Rubin, Vera CPUs, RTX Spark PCs and Jetson Thor robotics.
Manu Shukla -
Nvidia's AI PC Push Bets on Unproven Demand Beyond Niche Users
Nvidia's RTX Spark superchip enters a skeptical AI PC market in 2026. What was announced, why demand is unproven, what enterprise buyers should know.
Manu Shukla