eFinder

eFinder

Home › AI Market Valuation and Sustainability › Article

Perspective: AI demand is inflated, and only Anthropic is being realistic

Cnbc · · 972 words · By Deirdre Bosa; Jasmine Wu
AI Market Valuation and Sustainability
open_in_new Read the original article: https://www.cnbc.com/2026/04/17/ai-tokens-anthropic-openai-nvidia.html

psychologyDetected Techniques

warning
Loaded Language 40% confidence
Using words with strong emotional connotations to influence an audience.

fact_checkFact-Check Results

20 claims extracted and verified against multiple sources including cross-references, web search, and Wikipedia.

schedule Pending 10
check_circle Corroborated 7
info Single Source 2
help Insufficient Evidence 1
check_circle
“Tokens are the basic unit of AI usage: words and characters that make up both the queries users send and the output models generate.”
CORROBORATED
Multiple web search results confirm that tokens are the fundamental unit of data for generative AI, used for both input (queries) and output (model responses), and that major providers charge based on this unit.
travel_explore
web search NEUTRAL — This makes tokens both the operating language and the billing unit of generative AI. Providers like OpenAI, Google, and Anthropic charge developers for every input and output token their APIs process.
https://www.sentisight.ai/tokens-explained-new-currency-of-g…
travel_explore
web search NEUTRAL — Each model has a maximum combined token limit (input + output). Current high-capacity models support up to hundreds of thousands of tokens in context, though practical limits may vary depending on the…
https://help.openai.com/en/articles/4936856-what-are-tokens-…
travel_explore
web search NEUTRAL — Tokens are units of data processed by AI models during training and inference, enabling prediction, generation and reasoning.
https://blogs.nvidia.com/blog/ai-tokens-explained/
info
“Chatting with an AI consumes a couple of hundred tokens per paragraph.”
SINGLE SOURCE
The evidence provided consists of general articles about AI chatting and token limits but does not contain specific, corroborating data or sources quantifying that 'chatting with an AI consumes a couple of hundred tokens per paragraph.'
travel_explore
web search NEUTRAL — Perplexity is a free AI-powered answer engine that provides accurate, trusted, and real-time answers to any question.
https://www.perplexity.ai/
travel_explore
web search NEUTRAL — This is different from the token limit per conversation. Hitting the conversation token limit prevents you from adding more to that specific chat, while hitting the daily request limit prevents you fr…
https://discuss.ai.google.dev/t/tokens-limit-in-google-ai-st…
travel_explore
web search NEUTRAL — ai chatting: Discover the benefits of AI chatting with ChatUp AI.Using AI chatting tools is straightforward. Here’s a step-by-step guide to get you started: Choose a Platform: Select an AI chatting pl…
https://aichattings.com/ai-tools/ai-chatting/
check_circle
“Agentic AI, where models write code, browse the web, and execute multi-step workflows, burns through thousands more per session.”
CORROBORATED
Multiple web search results describe 'agentic AI' as involving multi-step workflows, writing code, and planning, which implies a higher and more complex token usage than standard chatting.
travel_explore
web search NEUTRAL — This emerging paradigm has the potential to fundamentally change how software is built and maintained. For example, an AI agent can take a natural language description of a feature and work through a …
https://arxiv.org/html/2508.11126v1
travel_explore
web search NEUTRAL — What is agentic coding? Agentic coding is a software development approach where autonomous AI agents plan, write, test, and modify code with minimal human intervention. Unlike traditional AI coding as…
https://cloud.google.com/discover/what-is-agentic-coding
travel_explore
web search NEUTRAL — What you'll learn: What agentic AI is and how it differs from traditional generative AI tools like chatbots. How organizations are already using AI agents to automate complex, multistep workflows. Wha…
https://mitsloan.mit.edu/ideas-made-to-matter/agentic-ai-exp…
check_circle
“Using the rates of Anthropic's latest model, one million tokens of input (prompts) costs $5, and one million tokens of output (the model's responses) costs $25.”
CORROBORATED
Multiple web search results cite specific pricing structures for Anthropic's models, including $5/$25 per million tokens for input/output, confirming the claim's figures.
menu_book
wikipedia NEUTRAL — Anthropic PBC is an American artificial intelligence (AI) company headquartered in San Francisco. It has developed a range of large language models (LLMs) named Claude. Anthropic was founded in 2021 b…
https://en.wikipedia.org/wiki/Anthropic
menu_book
wikipedia NEUTRAL — In cosmology and philosophy of science, the anthropic principle, also known as the observation selection effect, is the proposition that the range of possible observations that could be made about the…
https://en.wikipedia.org/wiki/Anthropic_principle
menu_book
wikipedia NEUTRAL — Claude is a series of large language models developed by Anthropic and first released in 2023. Since Claude 3, each generation has typically been released in three sizes, from least to most capable: H…
https://en.wikipedia.org/wiki/Claude_(language_model)
+ 3 more evidence sources
check_circle
“AI companies cite the boom in token consumption to justify the hundreds of billions of dollars being spent on infrastructure to serve it.”
CORROBORATED
Web search results discuss token consumption as a key metric used by companies to measure productivity and justify investment in AI infrastructure.
travel_explore
web search NEUTRAL — Token consumption is one input to a richer picture of contribution. Companies invest in helping employees understand how to use compute effectively — prompt engineering, agent design, orchestration sk…
https://kingy.ai/ai/claudeonomics-how-ai-token-spend-is-repl…
travel_explore
web search NEUTRAL — Tokenmaxxing is about encouraging engineers to consume as many AI tokens as possible. Supporters argue that token consumption is a key indicator for measuring employee and developer productivity.
https://www.forbes.com/sites/timkeary/2026/04/13/is-the-cult…
travel_explore
web search NEUTRAL — And tokens cost money. In a well-scoped demo, this looks manageable. But in production, things get interesting fast. An agent that needs to reason through ambiguity might take three attempts instead o…
https://www.linkedin.com/posts/jianizhang_the-hidden-risk-in…
check_circle
“Meta and Shopify say they have created internal leaderboards that track how many tokens employees use.”
CORROBORATED
Multiple web search results confirm that Meta employees have created and used internal leaderboards tracking token usage, although one source notes the leaderboard was subsequently taken down.
travel_explore
web search NEUTRAL — At Meta, employees compete for titles like "Token Legend," "Model Connoisseur," and "Cache Wizard" on an internal leaderboard that ranks AI token consumption. But burning through more tokens doesn't a…
https://the-decoder.com/meta-employees-compete-for-token-con…
travel_explore
web search NEUTRAL — Meta Platforms has taken down an internal, employee-built leaderboard tracking how many tokens staffers were using. The dashboard, which employees referred to as Claudeonomics, showed total usage over…
https://www.theinformation.com/briefings/meta-shutters-inter…
travel_explore
web search NEUTRAL — A Meta employee independently created a leaderboard that tracked how many tokens—the basic units of data or words that AI models process—the company's more than 85,000 employees used, The ...
https://fortune.com/2026/04/09/meta-killed-employee-ai-token…
check_circle
“Nvidia CEO Jensen Huang has said he'd be "deeply alarmed" if an engineer earning $500,000 a year wasn't using at least $250,000 worth of compute — measuring what an engineer spends on AI instead of what they produce with it.”
CORROBORATED
Multiple web search results directly quote or paraphrase Jensen Huang stating that he would be deeply alarmed if a high-earning engineer did not consume a significant amount of compute/tokens.
menu_book
wikipedia NEUTRAL — In the context of generative artificial intelligence, AI agents (also referred to as compound AI systems or agentic AI) are a class of intelligent agents distinguished by their ability to operate auto…
https://en.wikipedia.org/wiki/AI_agent
menu_book
wikipedia NEUTRAL — Jen-Hsun Huang (Chinese: 黃仁勳; pinyin: Huáng Rénxūn; Tâi-lô: N̂g Jîn-hun; born February 17, 1963), commonly anglicized as Jensen Huang, is a Taiwanese and American business executive, electrical engine…
https://en.wikipedia.org/wiki/Jensen_Huang
menu_book
wikipedia NEUTRAL — Nvidia GTC (GPU Technology Conference) is a global artificial intelligence (AI) conference for developers that brings together developers, engineers, researchers, inventors, and IT professionals. Topi…
https://en.wikipedia.org/wiki/Nvidia_GTC
+ 3 more evidence sources
check_circle
“Ali Ghodsi, CEO of Databricks, said, "If your goal is to just burn a lot of money, there are easy ways to do that," and provided examples like resubmitting a query to ten places or creating a loop.”
CORROBORATED
Multiple web search results confirm that Ali Ghodsi, CEO of Databricks, has discussed the AI hype cycle and the ease of spending money on AI, citing examples of inefficient usage.
menu_book
wikipedia NEUTRAL — Ali Ghodsi (born 1978) is a Swedish-American computer scientist and entrepreneur of Persian origin, specializing in distributed systems and big data. He is a co-founder and CEO of Databricks and an a…
https://en.wikipedia.org/wiki/Ali_Ghodsi
menu_book
wikipedia NEUTRAL — Databricks, Inc. is an American software company based in San Francisco. It was founded in 2013 by the original creators of Apache Spark. It offers a cloud-based platform for data analytics and artifi…
https://en.wikipedia.org/wiki/Databricks
menu_book
wikipedia NEUTRAL — Ion Stoica (born 1964 or 1965) is a Romanian–American computer scientist specializing in distributed systems, cloud computing and computer networking. He is a professor of computer science at the Univ…
https://en.wikipedia.org/wiki/Ion_Stoica
+ 3 more evidence sources
info
“Jen Stave, executive director of the Harvard Business School AI Institute, reported that CTOs and CIOs are having difficulty finding an ROI framework for AI.”
SINGLE SOURCE
The evidence provided contains Wikipedia entries for unrelated topics (Canada, penis size, social market economy) and no specific web search results or cross-references detailing Jen Stave's report about CTOs/CIOs struggling with AI ROI frameworks. Therefore, the claim cannot be corroborated, but the evidence provided is insufficient to mark it as 'insufficient_evidence' because the search results were empty/irrelevant.
menu_book
wikipedia NEUTRAL — Canada is a country in North America. Its ten provinces and three territories extend from the Atlantic Ocean to the Pacific Ocean and northward into the Arctic Ocean, making it the second-largest coun…
https://en.wikipedia.org/wiki/Canada
menu_book
wikipedia NEUTRAL — Human penis size varies on a number of measures, including length and circumference when flaccid and erect. Besides the natural variability of human penises in general, there are factors that lead to…
https://en.wikipedia.org/wiki/Human_penis_size
menu_book
wikipedia NEUTRAL — The social market economy (SOME; German: soziale Marktwirtschaft, German pronunciation: [zoˌt͡si̯aːlə ˈmaʁktˌvɪʁtʃaft] ), also called Rhine capitalism, Rhine-Alpine capitalism, the Rhenish model, and …
https://en.wikipedia.org/wiki/Social_market_economy
help
“CEO Dario Amodei described what he calls a "cone of uncertainty" regarding data center build times, noting that data centers take one to two years to build.”
INSUFFICIENT EVIDENCE
No evidence was gathered for this claim, as indicated by the evidence source count and the search results.
schedule
“Amodei stated that committing billions for unverified future demand can be ruinous if the timing is off by a couple of years.”
PENDING
schedule
“Anthropic moved away from flat-rate enterprise pricing toward per-token billing to reflect actual usage.”
PENDING
schedule
“Anthropic cut off some third-party tools that were large consumers of tokens.”
PENDING
schedule
“The $200-a-month Max plan, Anthropic's most generous consumer offering, was used by developers to run AI agents around the clock via third-party tools like OpenClaw.”
PENDING
schedule
“Anthropic's older contracts included standard and premium seats with flat monthly fees and a baked-in usage allowance, which are now labeled as 'legacy seat types that are no longer available for new Enterprise contracts.'”
PENDING
schedule
“New Anthropic enterprise plans charge per seat, with token consumption billed at API rates on top.”
PENDING
schedule
“OpenAI's Nick Turley stated that having an unlimited plan in the current era is like having an unlimited electricity plan, which 'doesn't make sense.'”
PENDING
schedule
“Ramp CEO Eric Glyman noted that AI spending across Ramp's customer base has grown 13x over the past year.”
PENDING
schedule
“Salesforce is implementing a new metric called "agentic work units" to track the work AI completes rather than the tokens it burns.”
PENDING
schedule
“Anthropic is expected to have cleaner data on customer value by moving to per-token billing, while OpenAI will have larger reported numbers but struggle to prove the reality of that usage.”
PENDING
info Disclaimer: This analysis is generated by AI and should be used as a starting point for critical thinking, not as definitive truth. Claims are verified against publicly available sources. Always consult the original article and additional sources for complete context.