---
title: "Exa says its agents wrote kernels that cut indexing cost 47% on Amazon's chips, not Nvidia's"
description: "Exa's first public number on the cost of running its index was earned on Amazon's AI chips rather than Nvidia's, and it lands two weeks after DeepSeek, the Chinese AI lab, set a 75-cent floor under the price of search."
publisher: "The Inference"
section: "Platforms"
published: 2026-11-19T15:11:12.587Z
modified: 2026-11-19T15:11:12.587Z
canonical: https://theinference.org/article/exa-says-its-agents-wrote-kernels-that-cut-indexing-cost-47-on-amazon-s-chips-not-nvidia-s
language: en
keywords: "Exa, AI Search, Amazon Web Services, AI Infrastructure, Chips"
---

# Exa says its agents wrote kernels that cut indexing cost 47% on Amazon's chips, not Nvidia's

> Exa's first public number on the cost of running its index was earned on Amazon's AI chips rather than Nvidia's, and it lands two weeks after DeepSeek, the Chinese AI lab, set a 75-cent floor under the price of search.

[Exa](https://theinference.org/markets/companies/exa) has published the first number on what its search index costs to run, and AI agents wrote the custom kernels behind it. On November 18, [Exa](https://exa.ai/pricing), whose search API is what an AI agent calls for the live web, [posted the details on AWS's machine-learning blog](https://aws.amazon.com/blogs/machine-learning/exa-boosts-embedding-throughput-by-50-to-index-billions-of-documents-on-aws-ai-chips/) [1]. It runs the embedding step of its web index on [Amazon](https://theinference.org/markets/companies/amazon)'s Inferentia2 and Trainium2 chips rather than [Nvidia](https://theinference.org/markets/companies/nvidia) GPUs, and reports, in its own words, "a 47% improvement in performance and cost per 1 million documents embedded" [1]. The figure is self-benchmarked, on a page co-published by Exa's cloud vendor [1].

## A model change forced the rewrite

A better document model required re-embedding 30 billion high-priority documents, which temporarily doubled the embedding load, and Exa used that window to re-pick its silicon [1]. Off-the-shelf kernels from Amazon's Neuron SDK came first, then two custom pieces [1]. One fused kernel stopped the chip's memory stalls; the other rebuilt attention around the exact shapes of Exa's model, so no compute went to padding [1]. Together they took throughput to roughly 1.5 times the naive compile, and the same code now carries to Trainium2 [1].

## The agents did the porting work

Every optimization came out of a harness built on Flyte, a machine-learning operations platform [1]. A candidate kernel is compiled, checked against the reference model to 0.999 cosine similarity, profiled, then run through a half-hour saturation benchmark on real hardware [1]. Lightweight models grid-search compiler settings, reasoning agents test hypotheses and find bottlenecks, and the strongest code generators write the kernels that fix them [1]. AWS's own Neuron Science team published [a September playbook](https://www.techtimes.com/articles/328113/20260928/aws-trainium-runs-real-time-ai-video-generation-without-nvidia-nki-kernel-playbook-published.htm) in which an agentic tool wrote a working Trainium kernel on its first attempt [6].

## The cost side was the missing number

Exa's revenue is public: $100 million annualized, [disclosed on October 13](https://theinference.org/article/exa-discloses-100-million-of-revenue-at-a-22x-multiple-four-days-before-deepseek-set-a-0-75-floor), on 2 billion search requests a month [2]. Its May valuation prices the company at roughly 22 times that run rate [2].

The floor is public too: [DeepSeek](https://theinference.org/markets/companies/deepseek), the Chinese AI lab backed by the High-Flyer quant fund, charges 75 cents per 1,000 search requests, against $4 for Exa's cheapest tier [2][3][4]. The cost side had no number until this week, and the only document for it is Exa's own post, tuned so far for Inferentia2 in Oregon [1].

The floor moved again on publication day: from November 18, DeepSeek [stopped billing search calls made inside its own models separately](https://api-docs.deepseek.com/quick_start/pricing/) and counts them as token usage instead [4].

*Chart: **DeepSeek's floor sits far beneath every listed search API** Bar chart of search API pay-as-you-go list prices per 1,000 requests, from DeepSeek at $0.75 up to Tavily at $8, with Exa's Lite and Auto tiers and DeepSeek highlighted.*

*Pay-as-you-go list prices per 1,000 search requests as posted, October to November 2026; Perplexity adds token fees on top. Sources: The Inference, Exa and DeepSeek pricing pages. [2][3][4]*

## Amazon wins a reference workload, Nvidia does not

For Amazon, the post is a production inference workload from a company whose entire business is inference, running on its own silicon, and AWS already counts [Anthropic](https://theinference.org/markets/companies/anthropic), [Databricks](https://theinference.org/markets/companies/databricks) and [Uber](https://theinference.org/markets/companies/uber) among Trainium customers [1][5]. For Nvidia, the same workload is inference revenue it did not win [1].

Exa can cut prices only as fast as its cost per million documents falls. So the two things to watch are plain: whether Exa holds its $4 Lite price now that its unit-cost curve is public [3], and whether tuning done for one chip in one region carries to Trainium2 and other regions at the same gain [1].

## Takeaway

Exa's first public cost number for its web index was earned on Amazon's Inferentia2 and Trainium2 chips, not Nvidia GPUs, and the 47% gain per million documents is self-benchmarked. Its $4 Lite tier still sits more than five times above DeepSeek's 75-cent floor.

## Sources

1. [AWS Machine Learning Blog, "Exa boosts embedding throughput by 50% to index billions of documents on AWS AI chips," 18 November 2026](https://aws.amazon.com/blogs/machine-learning/exa-boosts-embedding-throughput-by-50-to-index-billions-of-documents-on-aws-ai-chips/)
2. [The Inference, "Exa discloses $100 million of revenue at a 22x multiple, four days before DeepSeek set a $0.75 floor," 8 November 2026](https://theinference.org/article/exa-discloses-100-million-of-revenue-at-a-22x-multiple-four-days-before-deepseek-set-a-0-75-floor)
3. [Exa, "Pricing," endpoint pricing table, accessed 19 November 2026](https://exa.ai/pricing)
4. [DeepSeek API Docs, "Models & Pricing," accessed 19 November 2026](https://api-docs.deepseek.com/quick_start/pricing/)
5. [AWS, "AWS Trainium" product page, accessed 19 November 2026](https://aws.amazon.com/ai/machine-learning/trainium/)
6. [TechTimes, "AWS Trainium Runs Real-Time AI Video Generation Without NVIDIA: NKI Kernel Playbook Published," 28 September 2026](https://www.techtimes.com/articles/328113/20260928/aws-trainium-runs-real-time-ai-video-generation-without-nvidia-nki-kernel-playbook-published.htm)


---

Exa says its agents wrote kernels that cut indexing cost 47% on Amazon's chips, not Nvidia's — The Inference. Canonical: https://theinference.org/article/exa-says-its-agents-wrote-kernels-that-cut-indexing-cost-47-on-amazon-s-chips-not-nvidia-s
