---
title: "Cerebras doubles its chip's clock speed as OpenAI's fast tier runs at 300 tokens a second, down from 750"
description: "OpenAI's new speed lane tops out at 300 tokens a second, less than half the 750 its last model pulled from Cerebras silicon, and one research firm says Nvidia GPUs are doing the work. Cerebras is doubling its flagship wafer's clock to defend a contract worth more than $20 billion, and the stock swung 9 percent on a two-sentence Altman post."
publisher: "The Inference"
section: "Silicon"
published: 2026-10-16T14:33:46.426Z
modified: 2026-10-16T14:33:46.426Z
canonical: https://theinference.org/article/cerebras-doubles-its-chip-s-clock-speed-as-openai-s-fast-tier-runs-at-300-tokens-a-second-down-from-750
language: en
keywords: "Cerebras, OpenAI, Nvidia, Semiconductors, Stocks"
---

# Cerebras doubles its chip's clock speed as OpenAI's fast tier runs at 300 tokens a second, down from 750

> OpenAI's new speed lane tops out at 300 tokens a second, less than half the 750 its last model pulled from Cerebras silicon, and one research firm says Nvidia GPUs are doing the work. Cerebras is doubling its flagship wafer's clock to defend a contract worth more than $20 billion, and the stock swung 9 percent on a two-sentence Altman post.

## The fast tier's ceiling fell from 750 tokens to 300

[OpenAI](https://theinference.org/markets/companies/openai) [introduced its Ultrafast speed lane on September 29](https://venturebeat.com/technology/openais-gpt-6-1-sol-offers-astra-like-performance-at-1-5th-price-a-new-ultrafast-tier-clocks-at-300-tokens-per-second): up to 300 tokens a second, with the version for its newest model, GPT-6.1 Sol, due within days [1]. The previous model's fast lane, previewed in August, had run on Cerebras chips at up to 750 tokens a second [1]. The next day, research firm [SemiAnalysis](https://theinference.org/markets/companies/semianalysis) posted that the new tier runs on [Nvidia](https://theinference.org/markets/companies/nvidia) GPUs. [Cerebras](https://theinference.org/markets/companies/cerebras-systems) fell 7 percent, and neither OpenAI nor Cerebras has confirmed the hardware [2][3].

## Altman said partner; Microsoft said Maia

OpenAI chief executive [Sam Altman](https://theinference.org/markets/people/sam-altman) [called Cerebras "a close partner"](https://theinference.org/article/altman-calls-cerebras-a-close-partner-as-openai-s-fastest-tier-runs-on-nvidia) on October 2, naming no chip, and [the stock rose about 9 percent](https://theinference.org/article/cerebras-rallied-9-on-altman-s-praise-but-openai-s-20-billion-revenue-gap-sets-a-november-test) [3]. On October 12, [Microsoft](https://theinference.org/markets/companies/microsoft) [said its own Maia chips become "the primary inference system for Microsoft products and services" from 2027](https://blogs.microsoft.com/blog/2026/10/12/microsoft-to-expand-azure-maia-system-to-multi-gigawatt-deployments/), on silicon that already runs OpenAI models; Cerebras fell 10.69 percent [4][5]. The relationship under question is a multiyear agreement worth more than $20 billion for 750 megawatts of capacity [6].

## Nvidia is serving the lane Cerebras was built for

Inference, running a trained model to produce answers, is Cerebras's whole business. Its Wafer-Scale Engine, more than 50 times the size of a conventional processor, is built to answer first, and the company claims output 10 to 15 times faster than GPU platforms [7]. Batch size counts how many requests a chip handles at once. GPUs are most efficient with thousands; Cerebras wins with one. SemiAnalysis [reported the new tier runs at low batch size](https://finance.yahoo.com/technology/ai/articles/cerebras-stock-slides-nvidia-reportedly-142404033.html), meaning Nvidia is answering the fast lane a single request at a time, the game Cerebras's hardware was designed to win [2].

## The reply: double the clock, keep the wafer

Cerebras's answer, [detailed on October 12](https://chipdispatch.com/articles/cerebras-doubles-wse-3-clocks-to-2-8-ghz-for-cs-4-nexus-systems-05e78a1c), is not a new chip. CS-4 Nexus ships this quarter with the Wafer-Scale Engine 3's clock doubled to roughly 2.8 GHz from 1.4, on the same 900,000 cores, in a rack built to host three generations of machines through 2029 [8]. The bet is that agentic software multiplies the demand for fast single answers: one question now spawns "a family of daughter agents" coordinating together, says chief strategy officer Andy Hock, and that is "where speed really makes the difference" [7].

## Insiders sold the unlock; Ark bought the dip

About 19.4 million shares became eligible to trade on October 14 [6], into a stock that opened at $142.17 on October 16, just above its 52-week low of $133.88 and far below the $185 price of its May listing [5][9]. Insiders sold 1,864,452 shares for $387.7 million over three months; chief operating officer [Dhiraj Mallick](https://theinference.org/markets/people/dhiraj-mallick) halved his holding at $204.04 on September 29 under a pre-arranged trading plan [9]. Cathie Wood's Ark Invest took the other side, buying about $16.3 million across October 13 and 14 [6].

*Chart: **From $210 insider sales to below the $185 IPO price in six weeks** Line chart of Cerebras share price milestones from the May 2026 IPO through October 16, 2026, falling from $210 insider sale prices to the $142 area, below the $185 IPO price.*

*Milestone CBRS prices in US dollars. The May point is the IPO offer price; September 4 and 29 are insider sale execution prices; October 2, 12 and 14 are session closes; October 5 is an intraday quote captured at publication time; October 16 is the opening price. Sources: Benzinga via TradingView, TokenPost, MarketBeat. [3][5][6][9]*

## A $25.4 billion backlog, and only a fifth converts by 2028

The $25.4 billion backlog stands against [a full-year core revenue guide of $880 to $890 million](https://dyax.io/en/news/yh-702d6cd6ab722265), after second-quarter core revenue more than doubled to about $210 million, and the stock carries a 217-times forward earnings multiple [10]. Cerebras [expects only about 22 percent of that backlog, near $5.6 billion, to become revenue by June 2028](https://www.fool.com/investing/2026/09/26/prediction-cerebras-revenue-will-more-than-double-in-2027/), on timing a customer can shift [11]. Three readings settle it: whether OpenAI or Cerebras names the silicon behind Ultrafast, whether CS-4 ships this quarter, and the third-quarter report against its $214 to $216 million revenue guide [8][11].

*Chart: **Core revenue more than doubled in Q2; the year is guided to $880 to $890 million** Column chart of Cerebras core revenue: about $210 million in Q2 2026, guided to $214 to $216 million in Q3 and $880 to $890 million for the full year.*

*Core revenue in millions of US dollars; the Q3 and full-year figures are company guidance, shown at midpoint with the guided range marked. Q2 2026 is the reported quarter, more than double a year earlier. Sources: DYAX citing Yahoo Finance; The Motley Fool. [10][11]*

## Takeaway

OpenAI's Ultrafast tier runs at 300 tokens a second, less than half the 750 the last model pulled from Cerebras, and SemiAnalysis says Nvidia is serving it. Cerebras is doubling its wafer's clock to 2.8 GHz to defend a contract worth more than $20 billion, with the Q3 print against a $214 to $216 million guide as the next test.

## Sources

1. [VentureBeat, Carl Franzen, "OpenAI's GPT-6.1 Sol offers Astra-like performance at 1/5th price. A new Ultrafast tier clocks at 300 tokens per second", 29 September 2026](https://venturebeat.com/technology/openais-gpt-6-1-sol-offers-astra-like-performance-at-1-5th-price-a-new-ultrafast-tier-clocks-at-300-tokens-per-second)
2. [Investing.com via Yahoo Finance, Louis Juricic, "Cerebras stock slides as Nvidia reportedly powers OpenAI's 'Ultrafast' tier", 30 September 2026](https://finance.yahoo.com/technology/ai/articles/cerebras-stock-slides-nvidia-reportedly-142404033.html)
3. [Benzinga via TradingView, "Cerebras Stock Fell on OpenAI Chip Report, Then Rose 9% After Altman Weighed In", 5 October 2026](https://www.tradingview.com/news/benzinga:e63007e08094b:0-cerebras-stock-fell-on-openai-chip-report-then-rose-9-after-altman-weighed-in/)
4. [Microsoft Official Blog, Scott Guthrie, "Microsoft to expand Azure Maia system to multi-gigawatt deployments", 12 October 2026](https://blogs.microsoft.com/blog/2026/10/12/microsoft-to-expand-azure-maia-system-to-multi-gigawatt-deployments/)
5. [TokenPost, Enna Lee, "Cerebras Shares Hit 52-Week Low After Microsoft Plans Maia Expansion", 12 October 2026](https://www.tokenpost.com/news/investing/30995)
6. [Benzinga via TradingView, "Cathie Wood Buys $12.4 Million of CoreWeave Stock as Analyst Warns of Equity Wipeout, Ark Sells Palantir Shares", 16 October 2026](https://www.tradingview.com/news/benzinga:c1f9df714094b:0-cathie-wood-buys-12-4-million-of-coreweave-stock-as-analyst-warns-of-equity-wipeout-ark-sells-palantir-shares/)
7. [SiliconANGLE, Jonathan Anthony, "Cerebras positions inference speed as a competitive edge in agentic AI", 15 October 2026](https://siliconangle.com/2026/10/15/agentic-workflows-make-speed-requirement-says-cerebras-aidatapipelineforum/)
8. [Chip Dispatch, Nathan Brooks, "Cerebras Doubles WSE-3 Clocks to 2.8 GHz for CS-4 Nexus Systems", 12 October 2026](https://chipdispatch.com/articles/cerebras-doubles-wse-3-clocks-to-2-8-ghz-for-cs-4-nexus-systems-05e78a1c)
9. [MarketBeat, "Cerebras Systems Inc. $CBRS Stock: NFSG Corp Grows Stake", 16 October 2026](https://www.marketbeat.com/instant-alerts/filing-cerebras-systems-inc-cbrs-stock-nfsg-corp-grows-stake-2026-10-16/)
10. [DYAX citing Yahoo Finance, "Can Cerebras Convert Its $25.4B Backlog Into High-Margin Growth?", 13 October 2026](https://dyax.io/en/news/yh-702d6cd6ab722265)
11. [The Motley Fool, Daniel Sparks, "Prediction: Cerebras' Revenue Will More Than Double in 2027", 26 September 2026](https://www.fool.com/investing/2026/09/26/prediction-cerebras-revenue-will-more-than-double-in-2027/)


---

Cerebras doubles its chip's clock speed as OpenAI's fast tier runs at 300 tokens a second, down from 750 — The Inference. Canonical: https://theinference.org/article/cerebras-doubles-its-chip-s-clock-speed-as-openai-s-fast-tier-runs-at-300-tokens-a-second-down-from-750