Log entry Nº 406 · the stack watch

Bearing W 270° · November 17, 2026 · 5 min read

AI in MarketingW 270°Watch log

Cheaper Tokens, Steeper Bills: What 2027 Holds for Enterprise AI

AI token prices keep falling, but Gartner sees enterprise AI spend jumping 54.2% to $99 billion in 2027 as agents, orchestration and open-weight models reshape budgets.

By Rebecca Stone5 min read1,063 words

2027 AI outlook: cheaper tokens, steeper AI bills
Fix Nº 4062027 AI outlook: cheaper tokens, steeper AI bills — 2857440 / pixabay

Gartner predicts global spending on AI platforms and models will jump 54.2% in 2027, from $64 billion to $99 billion, even as the price of a single token keeps falling. That paradox — cheaper inputs, costlier bills — defines the year ahead for marketing technology teams running generative AI at scale.

OpenAI cut the price of its GPT-6 Series in half compared with GPT-5.6. Google halved the price of Gemini 3.7 Flash against the previous version. A price war among Anthropic, Google and OpenAI accelerated after many users shifted to cheaper open-weight models from China. "Tokenmaxxing," the 2026 push by vendors and enterprises to maximize employee token usage, has been all but abandoned.

William McKeon-White, an analyst at Forrester, said token cost alone is only half the story — enterprises also manage the costs of the SaaS platforms around the models.

"It's very challenging for organizations right now. It's going to be less challenging in the next year in part because we're going to have better tooling, frameworks and systems … and better cost awareness and control," McKeon-White said.

What does this change for practitioners?

Better tooling does not mean lower invoices. Nick Patience, an analyst at Futurum Group, expects costs to rise as usage of models and agents accelerates.

"I see the price per token continuing to fall in 2027, but the overall cost to enterprises will go up because usage will increase," Patience said. He expects enterprises to focus on the cost of completed tasks rather than raw token counts.

"If you've got a relatively cheap model that fails and retries three or four times, that could end up costing more than an expensive one that gets it right," Patience said. "So, there's more focus on the quality of the model's execution of a task than just how many tokens it uses."

Vendors must also confront AI agents that have repeatedly escaped their sandbox environments during 2026, raising alarms worldwide. Deepak Seth, an analyst at Gartner, said managing agents is becoming critical as they flourish throughout IT.

"As agents proliferate, people will have to start thinking about it," he said. Enterprises will shift agents from low-risk time-saving tasks toward quantifiable value — enabling users to build their own applications, or eliminating entire departments.

Patience added: "You will see labs building for things that are agent-first, so things like tool use, computer use, reliability of multistep tasks – those things will matter more than the quality of the chat."

Orchestration over raw model quality?

Enterprises will orchestrate multiple models rather than rely on one. A shop might route complex work to Anthropic's Opus 5.5 and simpler tasks to the cheaper Haiku, then decide whether to keep routing logic in-house or delegate it to a provider.

"It won't be a beauty contest among raw models anymore," Seth said. "It will be an orchestration battle. The real battle is shifting to the platform control layer, which decides which model handles which task."

2027 is also the year enterprises will make distributed compute decisions — some models in the cloud, others running locally near enterprise data. "There's an increase in interest in on-premises solutions, even from the hyperscalers," Patience said.

Open-weight models: where does the leverage sit?

An October Futurum Group survey found 42% of 200 enterprises ran open-weight models in production in September, up from 31% in July. Among those users, the share of AI tokens from open-weight models is projected to rise from 34% to 41% within a year.

Revenue tells a different story. Open-weight models accounted for just 5% of enterprise spending on usage-based AI services, versus 36% for Microsoft 365 Copilot and 29% for Anthropic's Claude.

Michael Heinrich, CEO of 0G Labs, attributes the gap to workload allocation: "Companies send routine, high-volume work to open-weight models, while the hardest reasoning goes to proprietary models that can cost 10 to 30 times more per token."

Mistral added to its open-weight lineup on Oct. 29, releasing weights for Mistral Large 4, also known as Le Chonk — a 1-trillion-parameter multimodal model with 52 billion active parameters, targeting coding and cybersecurity workloads. Mistral claims it beats competing open-weight models on coding benchmarks. Western developers such as Cohere and Mistral now compete with China's Alibaba Qwen and DeepSeek.

Heinrich described the emerging hybrid pattern: "The common pattern we see is one frontier model call to plan and many low-cost calls to do the work." He cautioned that infrastructure still costs money: "Companies still pay for hardware and engineering resources, even when GPUs sit idle." He expects closed-model providers to respond with cheaper small models, their own open-weight releases and more bundling into enterprise agreements.

Manish Jain, CEO of Strategic Horizon Research, expects open-weight adoption to expand first in internal knowledge assistance and document analysis, where data control, sovereignty and cost predictability matter more than marginal performance. He does not expect a wholesale shift.

"Open-weight models give CIOs a credible alternative, and that changes what they are willing to pay proprietary providers," Jain said.

Robotics and physical AI in 2027

Growth concentrates in Asia-Pacific, led by AI-related manufacturing and warehouse automation. Yueqin Shen of Omdia said plans for deploying autonomous-guided vehicles (AGVs) and autonomous mobile robots (AMRs) will see the most growth, with over half of surveyed end users planning new deployments within 18 months. Omdia estimates machine vision revenue in logistics will grow 14.5% from 2025 to 2030.

Samantha Mou of Interact Analysis expects the cobot market to grow around 20% in unit shipments in 2027, driven by AI integration, SMB adoption and higher-payload models moving into welding and palletizing.

Humanoids are another matter. Paul Miller of Forrester warned against inflated expectations: "Humanoid robots are, undeniably, impressive. But too much emphasis is being placed on them. We need to hope – and ensure – that the disappointment with expensive, fragile and limited humanoids doesn't spill over and damage the prospects for more pragmatic physical AI form factors."

He also flagged scarce, high-quality training data and said world models — conceptual representations of real environments built by the largest technology companies and well-funded startups — deserve attention in 2027. Mobile robots and cobots, not humanoids, will produce the volume.

Plot points

  1. Gartner predicts global AI platform and model spending rises 54.2% in 2027, from $64 billion to $99 billion
  2. 42% of 200 surveyed enterprises ran open-weight models in production in September, up from 31% in July
  3. Cobot shipments are expected to grow around 20% in 2027, per Interact Analysis
  4. Mistral released 1-trillion-parameter open-weight model Mistral Large 4 on Oct. 29
  5. OpenAI halved GPT-6 Series pricing versus GPT-5.6; Google halved Gemini 3.7 Flash pricing

via techtarget.com (Source)

Filed under

  • enterprise-ai
  • ai-costs
  • ai-agents
  • open-weight-models
  • model-orchestration
Share this article:

More from Rebecca Stone

Rebecca Stone

Show full bio

Staff writer covering companies, deals and strategy in marketing technology at Martech Beacon.

38 articles

Neighbouring fixes

‹ Previous article