--- title: "AlphaSense's agents burn 54 trillion tokens a year; Cerebras posted 14.2% gross margins" description: "The market-intelligence platform's agents now do 99% of its research and burn 54 trillion tokens a year, served in part by Cerebras hardware, the supplier it named in a joint post. Cerebras reports a 14% gross margin and just hit an all-time low days before earnings." publisher: "The Inference" section: "Silicon" published: 2026-10-18T14:06:47.075Z modified: 2026-10-18T14:06:47.075Z canonical: https://theinference.org/article/alphasense-s-agents-burn-54-trillion-tokens-a-year-cerebras-posted-14-2-gross-margins language: en keywords: "AI Agents, Inference, AI Chips, Earnings, Enterprise AI" --- # AlphaSense's agents burn 54 trillion tokens a year; Cerebras posted 14.2% gross margins > The market-intelligence platform's agents now do 99% of its research and burn 54 trillion tokens a year, served in part by Cerebras hardware, the supplier it named in a joint post. Cerebras reports a 14% gross margin and just hit an all-time low days before earnings. On October 6 [AlphaSense](https://theinference.org/markets/companies/alphasense), a private market-intelligence platform, put out numbers no human had asked for: its AI agents now perform 99% of all research activity on the platform, roughly 2 billion agent actions a year against 20 million human ones, and token consumption has reached an annualized run rate of 54 trillion, five times the level of early 2026 [1][6]. The company sells subscriptions to banks, and banks keep buying: [more than 10,000 investment-banking users were added in the third quarter](https://www.streetinsider.com/Globe+Newswire/AlphaSense+Adds+Over+10%2C000+Investment+Banking+Users+as+an+Additional+6+of+the+World%E2%80%99s+10+Largest+Investment+Banks+Choose+AlphaSense+as+Their+Deal+Infrastructure/27153957.html), segment ARR grew 75% year over year, and 18 of the world's 20 largest investment banks now use the platform [1][2]. Read those two lines together and the tension is visible: the company sells seats to people, 10,000 more of them in a single quarter, while its own disclosure says agents, not people, do almost all the work, and every one of those agent actions consumes tokens someone must serve [1][6]. ## Every token is an inference call someone must serve On September 30 AlphaSense and [Cerebras Systems](https://theinference.org/markets/companies/cerebras-systems) (CBRS) [published a joint post](https://www.cerebras.ai/blog/how-alphasense-uses-fast-inference-to-make-agentic-research-interactive) naming the supplier: Cerebras hardware runs the same model 8.5 times faster, cuts the p90 time to first token by 88% from 19.5 seconds to 2.3 seconds, and lets the platform review three times as much evidence at no added latency [7]. That is a real customer, at real scale, in a workload category (fast agentic inference) that Cerebras has built its pitch around. The joint post does not say what share of the 54-trillion-token volume Cerebras carries, only that it runs the agentic research. ## The market priced the opposite The listed supplier looks like it is losing money on the era it enables. Cerebras [reported a 14.2% GAAP gross margin for the June quarter](https://financefeeds.com/cerebras-cbrs-stock-falling-below-185-ipo-price/), down from 31.1% a year earlier, against roughly 75% claimed for [Nvidia](https://theinference.org/markets/companies/nvidia) [9][4]. Its core gross margin, which excludes some costs, fell to 40.6%, and management guided the September quarter to 38-40%, calling it the trough [8]. On October 16 the stock [fell 10.5% to a new all-time low](https://x.com/LaMonicaBuzz/status/2111192184099655868), more than 30% below its $185 IPO price and 67% below its May peak of $386 [5]. *Chart: **The inference supplier earns a 14% gross margin against about 75% at Nvidia** Columns comparing gross margin in percent for the quarter ended June 30, 2026: Cerebras GAAP 14.2, Cerebras core 40.6, Nvidia roughly 75.* *Gross margin in percent, quarter ended June 30, 2026. Cerebras GAAP margin from the June 30 10-Q as reported by FinanceFeeds; core margin is the company measure cited by Zacks. Nvidia figure is a market estimate cited by Convequity and drawn as an estimate. [4][8][9]* ## What the two disclosures allow, together The arithmetic nobody has done: one enterprise customer's agentic workload, at 54 trillion tokens annualized and 2 billion agent actions, is precisely the demand a 14% gross-margin inference vendor needs but has not yet converted into profit [1][6][9]. Size the customer against the supplier and the ceiling is low: AlphaSense's whole business, [above $700 million of ARR by one tracker's estimate](https://www.arr.club/signal/alphasense-added-10000-investment-banking-users-q3), a private figure [2], sits against a supplier with $25.4 billion of remaining performance obligations at June 30, much of it the [OpenAI](https://theinference.org/markets/companies/openai) agreement, which recognized $56.8 million of revenue that quarter, though OpenAI could buy compute elsewhere [9]. Even the entire AlphaSense workload on Cerebras hardware would be a few percent of near-term revenue at most, so the token run rate is a demand signal, not a revenue driver. ## The verdict, and the dated test The bear case says a 14% gross margin means inference demand does not pay, and investors have cut Cerebras accordingly [4][5]. The bull case says the margin is a trough artifact: the company rented back systems to meet demand, and [expects recovery as owned capacity replaces them](https://finance.yahoo.com/markets/stocks/articles/hold-cerebras-stock-despite-25-170200599.html) [8]. Neither camp has touched the AlphaSense token figures. The September-quarter earnings print, days away, reports the guided margin and settles whether the trough was as billed [8]. Adoption evidence and margin doubt are both on the record; the print prices one of them. ## Takeaway AlphaSense's 54-trillion-token agent workload is real demand for fast inference, but its named supplier Cerebras is serving it at a 14.2% GAAP gross margin, and the token run rate is a demand signal, not yet a revenue driver: the September-quarter print settles the trough question. ## Sources 1. [StreetInsider (Globe Newswire), AlphaSense Adds Over 10,000 Investment Banking Users, 6 October 2026](https://www.streetinsider.com/Globe+Newswire/AlphaSense+Adds+Over+10%2C000+Investment+Banking+Users+as+an+Additional+6+of+the+World%E2%80%99s+10+Largest+Investment+Banks+Choose+AlphaSense+as+Their+Deal+Infrastructure/27153957.html) 2. [ARR Club, AlphaSense added 10,000+ investment banking users in Q3, 9 October 2026](https://www.arr.club/signal/alphasense-added-10000-investment-banking-users-q3) 4. [Convequity on X, gross margin gap between Cerebras and Nvidia, 17 October 2026](https://x.com/convequity/status/2111443518581326186) 5. [LaMonicaBuzz on X, CBRS new all-time low, 16 October 2026](https://x.com/LaMonicaBuzz/status/2111192184099655868) 6. [Tanay Jaithalia on X, agent v human usage at AlphaSense, 16 October 2026](https://x.com/tanayj/status/2110439242883342749) 7. [Cerebras blog, How AlphaSense Uses Fast Inference to Make Agentic Research Interactive, 30 September 2026](https://www.cerebras.ai/blog/how-alphasense-uses-fast-inference-to-make-agentic-research-interactive) 8. [Yahoo Finance (Zacks), Should You Hold on to Cerebras Stock Despite Its 25% One-Month Dip?, 16 October 2026](https://finance.yahoo.com/markets/stocks/articles/hold-cerebras-stock-despite-25-170200599.html) 9. [FinanceFeeds, Why Cerebras Stock Is Falling Below Its $185 IPO Price in 2026, 2 October 2026](https://financefeeds.com/cerebras-cbrs-stock-falling-below-185-ipo-price/) --- AlphaSense's agents burn 54 trillion tokens a year; Cerebras posted 14.2% gross margins — The Inference. Canonical: https://theinference.org/article/alphasense-s-agents-burn-54-trillion-tokens-a-year-cerebras-posted-14-2-gross-margins