Model

AlphaSense measures the AI price war, and its own bill is falling

The market-intelligence firm's event data became the gauge for companies trading down from frontier AI models. In the same fortnight, its own engineers showed the same trade running inside its stack, weeks after a $7.5 billion round priced it at 12.5 times ARR.

Vincent JiangVincent Jiangthree-minute read
Share
Cerebras chief executive Andrew Feldman, in glasses and a mustard sweater, listens across a table at a panel discussion with two other people
1 / 7Slide 1 of 7
Cerebras chief executive Andrew Feldman, left, at a Village Global panel. His company published the AlphaSense case study on 30 September 2026, and its stock closed below its $185 IPO price the same day.

The meter read sixfold

Executive mentions of open-weight AI models on corporate calls and events ran six times higher in August and September 2026 than a year earlier 1. The count comes from AlphaSense's~$4B — AlphaSense, private, latest valuation ~$4B own archive of filings and transcripts, and the Financial Times reported the figures as evidence of the corporate retreat from frontier-model pricing 1. AT&T attached a harder number to the same move: about 40% of its AI workloads already run on open models, with 70% in sight within a year 1.

The $7.5 billion mark, on the record

Samantha Greenberg, two months into the CFO chair after 18 years as a technology investor, has put the pre-IPO numbers on record: $350 million raised at a $7.5 billion valuation, nearly double the prior round, on more than $600 million in annual recurring revenue after the first quarter, growing well over 40% 2.

Token consumption on the platform is up 25-fold year on year, a 20 trillion-token run-rate, and international business already makes up 21% of ARR 2. Accenture, a customer for years, was one of the lead investors in the round 2. At that mark, AlphaSense is priced at 12.5 times ARR.

ARR has tripled since 2023; the new round pays 12.5 times it

$0M$200M$400M$600MEnd-2023Oct 2025Q1 2026$7.5B round = 12.5x ARR
Data
ARR
End-2023$200M
Oct 2025$500M
Q1 2026$600M
Annual recurring revenue at each company disclosure, in millions of US dollars. Multiple computed against the $7.5 billion valuation Greenberg put on record. Sources: Fortune via Yahoo Finance (April 2024), Forbes (October 2025), CFO interview (September 2026).2,5,6

The trade runs one layer down

Days after the meter reading, AlphaSense's engineers published their own bill. A case study co-signed by seven of them, dated 30 September 2026, shows Generative Search routing its latency-critical calls, intent, tools and evidence scoring, to Cerebras: p90 time to first token down 88%, from 19.5 seconds to 2.3, with rival providers running up to twice as slow 3.

The same week, OpenAI$1.18T — OpenAI, private, latest valuation $1.18T made the cuts permanent, halving GPT-6 Sol and Luna API prices to $2 and $0.10 per million input tokens, while Anthropic's$350B — Anthropic, private, latest valuation $350B Claude Opus 5.5 runs about 40% cheaper than its predecessor 41. AlphaSense has said it builds on Anthropic's models and fine-tuned open ones 5; no source discloses what it pays for inference. Even so, the company that sells the map of deflation is buying the terrain at the new prices.

Both sides carry the same risk

The AlphaSense figures are company-claimed and unaudited, with no filing behind them yet 2. The trade-down its data measures points at seat-priced software too: AlphaSense charged roughly $10,000 to $20,000 a seat per year as of 2024 5, and Greenberg herself describes pricing sliding toward consumption and outcome-based models 2.

Cerebras holds the mirror risk. It closed 30 September 2026 below its $185 IPO price even though second-quarter core revenue rose 103% to $209.9 million, after an unconfirmed report that its biggest customer, OpenAI, had moved its newest ultrafast model to Nvidia chips 78. For a stock built on one anchor customer, the AlphaSense deployment is a named enterprise logo; the case study names an architecture, not a revenue figure 37.

Two readings will test the trade

Cerebras reports third-quarter core revenue against guidance of $214 million to $216 million 7. The AlphaSense tell is simpler: hold ARR growth above 40% while token prices keep falling, and deflation pays for the multiple. When the trade-down finally reaches the meter itself, the meter will read that too.

Get Model by email, free

The AI labs, their models, and the business behind them. In our daily email, every weekday.

Reader comments

0 comments

    Model

    Nvidia shows how to run Z.AI's GLM-5.3 on $190,000 of Nvidia hardware, with no per-token bill to Z.AI

    The tutorial walks enterprises through serving GLM-5.3 on two DGX Stations, roughly $190,000 at listed reseller prices, with no per-token bill to Z.AI. For a stock carrying 118.6 times sales, that reads as a price list for everything Z.AI sells.

    Model

    Anthropic's Claude models tie at 58 on a coding index Cognition partly built, days before its Nasdaq IPO

    A third of the refreshed index runs on FrontierCode, a benchmark built by Cognition, whose Devin agent announced one of the two tied winners, Claude Fable 5.2, on the day of its release. Anthropic's model lists on Nasdaq on November 19; Cognition's Devin has offers from $70 billion on the table.

    Model

    Anthropic's Fable 5.2 passes 41.3% of real freelance jobs before its Nasdaq debut

    The one independent test of finished freelance work puts Anthropic's Fable 5.2 first at 41.3%, and the model still fails most of it, days before a Nasdaq debut priced at up to $1.4 trillion.

    Model

    Hedra cut Krea 3's edit time from 22.9 to 3.3 seconds on AMD, and the app doesn't own the speed

    The eight-second edit Krea advertises is manufactured in the serving stack, and a partner's measurement shows the app does not own the speed. Inference efficiency accrues to the optimizer and the chip vendor, which caps a per-image app's margin and hands AMD a demand channel beyond hyperscalers.

    Model

    xAI ships AI teammates at $30-$200 a month, ten days after Musk said serving them loses money

    The published card sets a chatbot price against an always-on machine the CEO says loses money on every subscriber; SpaceX shareholders fund both sides of that trade in one stock.

    Model

    Google ships Gemini 4 Hydrogen at $1.50 a million tokens and gives it away in Search

    Gemini 4 Hydrogen scores 51 on the Artificial Analysis index against 50 for the flagship Argon, at $1.50 per million input tokens with no long-context surcharge, and it is free in Search. That strips both halves of a model lab's premium case the same week Anthropic prices an IPO on one.

    Subscribe to TheInference for free