AI Infrastructure

PowerCenter 10.5 left standard support on March 31, 2026. Move warehouse loads to dbt with Fivetran or Airbyte, keep IDMC only if you also keep Informatica MDM, and use extended support as the bridge.

Dropbox will pay about $350M in cash for Tigris Data and launched DBX Agent Cloud on top of it. Neither company has said whether Tigris pricing, zero egress or current endpoints survive the Q4 close.

Confluent made Lightning Tables generally available on AWS at Current 2026. Apps and agents can query live Kafka state by SQL, API or MCP, but only one table at a time, with no joins or aggregates.

Kimi K3.1 Pro's 1.56 TB weights are public and free for internal use. An eight-GPU B300 node breaks even with the $2/$10 API near 13 billion input tokens a month, and legal and tax scores fell from K3.

Snowflake's Cortex Training claims up to 61% lower fine-tuning cost, based on a LoRA SFT benchmark. At $9.60 per H200 hour, your own throughput decides whether you see any of it.

Marvell's $115 billion all-stock deal for Lumentum leaves customer agreements in place only until close. Buyers have about 12 to 18 months to qualify second-source optics and secure post-close supply terms.

Ubuntu has no fixed kernel for CVE-2026-98432, a pre-auth Linux TCP security flaw with a claimed AI-written root exploit, while RHEL shipped fixes on November 1. Set patch order by vendor status and multipath TCP exposure.

Azure's NCCads H100 v5 is the default for attested GPU inference at $6.98 an hour with no list premium in East US 2. Google's confidential A3 needs Spot or flex-start capacity, AWS has no attested GPU, and most teams can use cheaper controls instead.

Aramco closed on 49% of HUMAIN, and the EU filing describes joint control with PIF. Anyone buying HUMAIN compute, cloud or ALLaM should settle data separation, capacity priority and exit terms before signing.

LLM API SLAs pay credits for failed requests and say little about slow ones. Compare OpenAI, Azure, Bedrock, Vertex and Anthropic, then negotiate definitions and build a fallback that triggers on latency.

Nine AI developer platforms disclosed breaches in October 2026, and Wiz traced two stolen enterprise Bedrock keys from them to about 800,000 and 1.7 million calls. Rotate what those platforms held, hunt CloudTrail, and narrow your keys.

For one support agent at 300,000 sessions a month, Google's Agent Runtime bills $510 of compute, AgentCore $846 and Foundry prompt agents nothing. Memory, web search and idle timeouts set the rest of the bill.

monday.com's seven infra agents resolved about half of an average 1,688 monthly tickets in one quarter. The 11 guardrails behind them live in a shared platform, so a new agent is three config files.

Naveen Rao's Unconventional AI is buying NeuReality for its silicon team. Neither company has said what happens to NR-NEXUS, the NR2 AI-SuperNIC or existing licences, so customers should get continuity terms in writing before the Q4 close.

Mistral Large 4's 1.05T-parameter weights are on Hugging Face. At FP8 they fit one eight-GPU DGX B200 node, internal use is free, and serving the model to outside customers needs a separate commercial licence from Mistral.

Baseten's A10G costs $1.21 an hour against SageMaker's $1.52, even with no scale-to-zero. Scale-to-zero saves more, but only on gaps longer than the idle window plus startup, and every such gap hands one user a cold start.

NVIDIA AI Enterprise lists at $4,500 per GPU per year. For bare-metal inference, vLLM or SGLang on your own stack, or Red Hat AI for a support contract, replaces it; vGPU and production NIM are where you still need it.

GLM-5.3 on Cloudflare Workers AI returned other customers' data in 155 long-context responses because an undersized KV pool tripped an SGLang disaggregation bug. Upgrade SGLang to v0.5.22 and ask your inference provider how KV pages are isolated.

ServiceNow says internal AI now produces more than $1 billion in annualized value. Sales operations shrank, quota-bearing sellers grew double digits and total headcount slightly declined.

Oracle's OpenAI launch at AI World 2026 is mostly undated: AI Workspace is coming soon, App Composer is in Limited Availability, and TPUs and Cerebras reach OCI in 2027. Fusion Claw's 25 agentic apps are the part you can buy now.

KServe and llm-d stack: KServe's LLMInferenceService runs llm-d underneath. Ray Serve replaces both, and it suits Ray-native Python teams more than Kubernetes platform teams.

vLLM's October advisories let one request crash or silently hang a shared inference engine, and two paths lead to code execution. Upgrade to 0.32.0, update Pillow, remove old Ghostscript and fix your health probes.

Azure Neural TTS lists $15 per million characters against $38 to $80 for Cartesia and ElevenLabs, and keeps no real-time text by default. The vendors' latency claims measure different things, so test p95 from your own region before you sign.

Anthropic's S-1 shows $16.3B of H1 2026 revenue, a $974M Q2 operating profit and 48% of sales through cloud marketplaces. For Claude buyers, it shows which contract terms to lock in: version pins, notice periods, capacity and a fallback model.

Claude runs on Bedrock, Vertex AI and Foundry at the same list price, so the committed-spend agreement you already signed decides. Bedrock is the default, Vertex is the only Gemini home, and Foundry is the pick for GPT but the weakest for Claude.

Under the Commission's GPAI guidelines, a fine-tune makes you a model provider only past one-third of the original training compute. LoRA never gets close; continued pretraining on a base with undisclosed compute can.

An H100 running gpt-oss-120b at Lambda's $4.29 an hour must sustain 5,958 tokens per second every hour of the month to match a $0.20 blended API rate. Business-hours traffic leaves it at about twice the API's cost.

Nvidia's DSX POD puts seven Vera Rubin racks in a truckable container, but partners ship it in early 2027 with no price. The open-source OpenShell agent runtime is the GTC Berlin announcement you can deploy this week.

Azure and Google now meter reserved model capacity at pay-as-you-go token weights, so a monthly term needs about 99% utilization to save money. Annual terms break even at 84% on Azure and 75% on Google; Bedrock publishes too little to compute.

Kimi K3 costs $3 / $15 at four of the five hosts compared that offer it and gpt-oss-120b $0.15 / $0.60 at four of six, but Moonshot's vendor audit found schema accuracy from 100% down to 71.96%. Fireworks is the default for agents; Together is the one to avoid for tool-heavy work.

A 50-call agent run is about 51 billable units on Temporal, Inngest, Restate and DBOS, so 100K runs costs $80 to $310 a month. Step caps, licences and ops load decide the pick.

Meta's 816B Muse Spark Open is free for internal use, but its license bans operating critical infrastructure and any nuclear or military use, the same clauses Llama 3 carried.

Prime Intellect bought Moondream for its Photon engine, and Moondream Cloud and Lens close November 13. Base weights self-host free, but local fine-tunes still need a Moondream API key.

GitHub Copilot and Microsoft Foundry move onto Microsoft's own Maia chips by year end, and Maia 300 brings MXFP4 numerics in 2027. Customers can't rent or choose Maia, so record an eval baseline before the switch.

Intel's air-cooled D-series GPU holds a 300B model on one PCIe card, but at 1.5 TB/s a large dense model streams single-digit tokens per second. Sort your workloads by capacity versus bandwidth before the quote arrives.

On short chat messages, Amazon Comprehend's PII meter costs about 80x what Claude Haiku 5.5 charges to read the same text. Here is every meter on one workload, and how to cut the bill.

Nvidia is reportedly in talks to buy, acquihire or take a bigger stake in Reflection AI, and Beam's promised Apache 2.0 weights have not been published yet. Mirror a released checkpoint before you standardise on Beam.

Amazon is in talks to buy Decart for about $7 billion, which would put the DOS layer sold as neutral across Nvidia, Trainium and TPU inside Trainium's maker. Customers should get named multi-chip support in writing before a deal signs.

Perplexity's MIT-licensed pplx-embed-v2-late models search page images with no OCR or chunking, and a 0.6B model can query a 9B-built index. The catch is a multi-vector index that published measurements put at 20x to 64x a dense one.

Hightouch bought Strata, whose agent builds and maintains custom API integrations from a plain-language description. The deal is silent on Strata's standalone customers, pricing, and how agent-made changes are reviewed before they write to production.

micro1 will spend $1 billion over 12 months buying company SOPs, CRM data and decision records for AI training. Its public terms name no buyer, exclusivity or liability, and much of the data it wants carries customer and vendor rights.

Run FP8 by default. For 4-bit, INT4-AWQ beat NVFP4 on speed on three of four models in an independent Blackwell benchmark, and NVIDIA's own NVFP4 Gemma file was nearly the size of FP8.

Anthropic's OSS Scanner sends unreviewed, model-written bug reports to open-source maintainers and publishes nothing. Its disclosure pipeline has 516 of 6,157 findings patched and 584 identifiers issued, so many fixes will reach your dependencies with no CVE for your SCA tool to flag.

Clockwork raised $31M with LinkedIn running its GPU failover and Together AI set to resell it, while AWS bundles checkpointless recovery into HyperPod for free. Every recovery figure on offer was measured by the seller, so price failover on your own interruption rate.

E2B and Daytona charge the same $0.0504 per vCPU-hour and Modal Sandboxes 41% more. On a month with long idle tails, the billing shape moves the invoice by 10x or more, so price your own alive-to-busy ratio first.

Mistral Large 4 beats Kimi K3 on Harvey's legal agent benchmark and trails it on the Vals Index. It is a discounted preview API for now, and the licence for its end-of-October weights is unpublished.

SAP is buying TechWolf, whose certified integration writes inferred skills into Workday Skills Cloud. The only promise to non-SAP customers is continued availability, with no end date or price term.

Reflection's Beam is a 501B Apache 2.0 open model from a US lab, and its own table has it losing to Kimi K3 on 13 of 14 shared benchmarks. Weights and pricing aren't out yet.

Self-hosted IBM Bob swaps SaaS Bob's Claude, Mistral and Granite routing for one local model, Nemotron or Laguna. IBM has published no evals, GPU sizing or price for it.

A one-year H100 reservation breaks even at 65-80% measured utilisation, depending on the provider. Commit only your 90-day floor for six months, run checkpointed training on preemptible capacity, and pay on-demand for the burst.

Nebius bought Inferize to cut GPU cold starts in Token Factory from minutes to seconds, but published no date, price change or capacity guarantee, and its own docs warn released capacity may not come back. Keep your minimum replicas until both are in the contract.

Inside one Snowflake account, Claude Sonnet 4.5 costs $3.60/$18.00 per million tokens via AI_COMPLETE but $3.00/$15.00 via the REST API, and AI Credits never take your warehouse discount. Fix the AI path first, then move batch compute onto Iceberg. Leave the catalog last.

Inception's Mercury Voice cuts the voice-agent LLM line to about $0.009 a minute, but its 320ms latency is model-only and its benchmark win arrives without numbers. Bake it off on your own calls before you swap.

NVIDIA's Kumo Tabular ships under OpenMDW-1.1 for commercial use, while TabPFN-3.5's open weights stay non-commercial. Its #1 TabArena Elo is NVIDIA's own run, and it was pretrained only to 60,000 rows and 100 columns.

NetApp is buying PEAK:AIO to put its pNFS metadata engine inside ONTAP. The announcement says nothing about the standalone software customers run on their own Dell, Supermicro or Gigabyte servers — get continuity terms before close.

AMD is buying World Labs for about $8.2B in stock, and neither company has said what happens to Marble or the World API. World Labs' terms already allow assignment and termination at any time — export your worlds before the year-end close.

For air-gapped LLMs, run Apache-2.0 weights like gpt-oss-120b on vLLM, via Red Hat AI if you need a vendor. NVIDIA AI Enterprise's per-GPU fee is the loser, and an 8-GPU air-gapped node costs 30-40x the hosted API at 1B tokens a month, and even two GPUs cost about 9x.

No FinOps platform can split a shared API key. Tag AI spend at the gateway and the provider first, reconcile against the cost API, then add Vantage for one view and OpenCost for GPUs. Only the gateway can cap a budget.

LiveKit bought Loophole Labs and its Substrate hypervisor to cut voice-agent startup below three seconds. The announcement says nothing about Architect, Loophole's paid Kubernetes spot-migration product, or its pre-GA AGX sandboxes.

AssemblyAI is the default for post-call transcription and Deepgram for live voice agents. Self-hosting pays past roughly 2-6M call minutes a month with a two-engineer team — and then the model to run is NVIDIA Parakeet, not Whisper.