AI ROI & Cost

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.

GPT-6 Astra, GPT-6.1 Sol and GPT-6 Luna now draw down Google Cloud commits, but Google's savings-plan terms give third-party models no discount, and the private preview has no published price, regions or ZDR yet.

Microsoft, Google and Salesforce raised base prices to pay for bundled AI. A bigger discount off the new list can still raise your bill, so negotiate from last term's per-user dollars and your own usage report.

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.

Zoomtopia 2026 made ZoomMate Agent Teams, MCP servers and the Search & Ask APIs available now, while Centralized AI Management, Compliance Manager for AI notes and AI Guardrails are dated December 2026 through Q1 2027. Here is the rollout order that follows.

Kapital is buying Woz for $30M in cash and stock and moving both founders into bank roles. The announcement says nothing about WOZCODE, so engineering teams should baseline its savings, pin the version and plan an exit.

Self-hosting LibreChat or Open WebUI on API keys costs about $3.72 a user a month on Sonnet for a typical mix, against $20 seats. A small heavy-user tail on a frontier model erases the saving, so budgets go in before launch.

Boomi bought Elementary Data for a reported $40-45M. Elementary says contracts and standalone purchase hold only 'at this time' and 'for now,' so Cloud customers should lock terms before renewal and keep the Apache 2.0 package ready as an exit.

Liquid AI's d1-ultra matches Jev's $0.042 per million input tokens, but it re-bills every image for every question, and its Decision Index lead of about eight points comes from text-only tests and more abstention.

Neptune's hosted users lost every run they had not exported in March 2026. Run MLflow on a database and bucket you own, keep W&B only with your own bucket, and schedule a weekly export whatever you pick.

Hex-Rays, maker of IDA Pro, is buying HexAI (formerly Ventris) for a reported $30M-$35M. Its new AI research tools are currently available only through HexAI, so IDA customers should settle model choice, air-gap support and pricing before they renew.

DeepSeek's new Search & Fetch API costs $0.75 per 1,000 calls, about a tenth of Tavily's basic search, with billing from October 31. Queries go to a service that stores data in China under PRC-governed terms.

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.

Workday Rising 2026 shipped Sana Home, Agent Harness, Developer Agent and Data Cloud on Flex Credits, while the payroll, accounting and deployment autopilots reach early adopters in 2027. Workday publishes credits per action, not the price of a credit.

Exa is the safest default web search API for AI agents, and Tavily is the one to skip. At frontier-model prices, the page text an API returns can cost 6 to 20 times the search fee.

Freshworks is buying Torii, a Gartner Leader in SaaS management, to put AI spend chargeback inside Freshservice. Torii customers on ServiceNow or Jira have no written continuity promise yet and should get one before the deal closes.

Cohere bills rerank per search and Voyage per token, so candidate count and chunk length decide the bill. Voyage rerank-3 is cheaper at most shapes; self-host BGE for residency or high volume.

Asurion kept its invoice capture tool and put a Copilot Studio agent on the exception queue, where 90% of its PO invoices were failing. Touchless AP rose from under 20% to over 70% in the Americas.

Wharton and GBK's 2026 survey of 1,416 US leaders and employees finds 35% of departments tracking a growth metric report strong AI ROI, against 13% tracking efficiency alone. Hours saved is still the most common metric.

Muse Wonder costs the same $4/$20 per million tokens as Claude Opus 5.5 and scores 52 to its 58 on Artificial Analysis, but runs about a quarter cheaper per task and leads on Harvey's legal agent benchmark. Wonder Full is reachable only through Muse Code subscriptions that train on your code by default.

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.

A batched small LLM translates a million characters for under $1 and Google NMT charges $20, but full post-editing of the same text runs $13,600 or more. On reviewed content, buy the engine your editors correct least.

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.

Work IQ's Dynamics 365 preview builds Copilot's semantic model from your existing views, forms and relationships, outdated ones included, and lets agents write back to records. It is on by default in some new production environments and bills in Copilot Credits.

For a laptop fleet, ship your own pinned small model on llama.cpp and add Foundry Local where the NPU pilot passes. OS-provided models change under you, and at Flash-Lite prices the token bill is about $48K a year for 20,000 seats.

OpenAI told investors its run rate was near $50B at the end of September, about $20B below the circulating figure. Anthropic books partner sales gross and OpenAI books them net, so compare the two on one basis.

Run GEPA against 20 to 100 labeled examples before you fund a fine-tune. It beat RL fine-tuning in its paper and matched supervised fine-tuning in Databricks' test, but the prompt it produces is pinned to one model version.

Microsoft's Decision-1 lists at $0.042 per million input tokens with free output, the same price as TypeSafe's Jev. Its 'highest accuracy' claim rests on 36 unnamed benchmarks with no published scores or calibration data.

For annual 10 m scoring, AlphaEarth embeddings plus a few hundred of your own labels match a bespoke imagery pipeline. Pin the version, validate with a geographic buffer, and keep the per-scene vendor only for daily or sub-metre work.

Deutsche Telekom's €2.5B AI savings target for 2030 is gross and excludes token costs, which it caps at a low double-digit share of savings. Its per-function structure is worth copying; its mixed baselines are not.

GraphRAG's index costs about 1,000x a vector index, but at 2,000 questions a day its global search path costs far more than the index. Start with hybrid search and add a graph only where your eval set shows global questions failing.

Workday's requisition data from about 550 employers shows demand for basic AI skills such as prompting peaked in January 2026 and fell 25%, while hands-on AI building skills rose 51%. Internal moves fell at 57% of employers.

Google's coworker agents get their own Workspace account, mailbox, Drive and directory listing. Google published no price, plan list or GA date, so settle sponsorship and licensing before teams start creating them.

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.

OpenAI's Decisions API bills only input tokens at $0.10 per 1M, which saves a little on short tickets. It also drops Luna's cached-input discount, so long shared rubrics can cost about four times more than the Responses API.

A vendor-run benchmark is a screening signal. Business cases should cite independent runs with per-task logs, led by Epoch AI and Artificial Analysis, plus a 50-task replication on your own data.

Claude Haiku 5.5 matches GPT-6 Luna at $0.10/$0.50 per million tokens and beats it on Anthropic's evals, but prompts over 100K tokens cost 5x and its tokenizer counts about 30% more tokens than Haiku 4.5. Short-prompt classification and routing get the big saving; compaction and long-context work get much less.

Decision models such as Jev cost about 25x less per call than Claude Haiku 4.5 at list price, but trail LLMs on accuracy until you add labels and a calibration fit. Fine-tune if you have labels; cascade Jev into an LLM if you don't.

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.

Chronos-2 is the forecasting foundation model to self-host: Apache 2.0, P1 to P99 quantiles and a Decathlon deployment behind it. TimesFM 3.0 tops Google's benchmarks, but its weights are licensed for production only inside BigQuery, and Moirai 2.0's are non-commercial.

OpenAI's Ultrafast tier runs GPT-6 Astra at $60 input and $300 output per million tokens, 6x Standard, on US or global processing only. It ships off for Enterprise; turn it on per named user with a spend limit.

Cohere North 2 gives agents memory that persists across sessions, but Cohere has published no scope, retention or deletion terms for it. The token caps in North Admin are usable now; memory should wait for written answers.

OpenAI's Agents API hosted browser asks for approval once per website, and its own docs say that does not enforce confirmation before purchases or deletes. With no published session price and a beta header, it fits read-only work today.

@ChatGPT in Slack and Teams lets staff without a ChatGPT seat act through an admin-configured service account and shared app connections. Scope channels, connections and credit limits before you enable it.

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.

OpenAI's Codex Security Cloud scans GitHub repositories as commits land, tries to reproduce flaws in a sandbox and drafts fix PRs. It bills by the token with no per-scan estimate, so pilot it with a cap beside your SAST.

Supabase is buying Turso to give every agent its own database. Both companies say Turso keeps running, but neither names a price, an SLA or a close date, so Turso customers should get their current terms in writing now.

Microsoft's MAI-Transcribe-2-Streaming claims a 2.5% streaming WER at $0.54 an audio hour, but it ships as a public preview with no SLA and no word-level timestamps, at 2.7 times Grok's streaming price.

Cloudflare's Apache 2.0 Clef beats TypeSafe's Jev on intent and tool-call benchmarks, trails it on When2Call, and costs about 5.7x more per input token hosted. Amazon's Strands Decider 2B runs on a consumer GPU but scores 0.505 on hard tasks.

Anthropic's new FDE residency is a build vs buy decision. A checklist for evaluating a nomination against hiring or an integrator, and the three terms to get in writing first.

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.

Microsoft's rebuilt Copilot keeps chat and Office features on the $30 seat and meters Autopilot, Code, Cowork and the Fable and Astra models in Copilot Credits. Set per-user limits and turn off Auto-apply before it reaches your tenant.

Manus 2.0 lets an incoming email or Slack message start an agent run, days after Salt Labs showed one email could hijack a Manus agent. Pilot the Team plan with read-only connectors and keep Cue off corporate accounts.

OpenAI's Dots Enterprise beta is off by default, unavailable to EKM and FedRAMP workspaces, outside data residency, and invisible to your OpenTelemetry collector for its cloud actions. Here is what to decide before users ask.

A $0.05 voice AI minute bills $0.09 to $0.14 once transcription, the model, the voice and the phone line are added. On a 40,000-minute inbound workload, ElevenLabs Agents was the cheapest managed platform and Bland AI the most expensive.

On a 30-turn agent loop, prompt caching saves 80-89% depending on the provider, not 90%. A timestamp in the system prompt drops Claude's saving to 6%, and a volatile tool definition makes caching cost more than not caching.

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.

Prompt caching cuts a 400K-token long-context answer to $0.085 on Claude Sonnet 5.5, but RAG still answers for $0.017. The crossover formula, the accuracy-at-depth evidence and the latency users notice.