---
title: "Amazon edits Nvidia's lock-in out of PyTorch and points its $25 billion chip business at CoreWeave and Nebius"
description: "Amazon's TorchNeuron runs PyTorch apps on its Trainium chips after one configuration change, and Amazon is donating code upstream to strip PyTorch of its chip-specific ties. The same day, Citi's Tyler Radke told CNBC that bond yields, not portability, are what price neoclouds like CoreWeave and Nebius."
publisher: "The Inference"
section: "Silicon"
published: 2026-10-19T23:29:00.232Z
modified: 2026-10-20T01:19:09.007Z
canonical: https://theinference.org/article/amazon-edits-nvidia-s-lock-in-out-of-pytorch-and-points-its-25-billion-chip-business-at-coreweave-and-nebius
language: en
keywords: "Amazon, Nvidia, PyTorch, Neoclouds, AI Chips"
---

# Amazon edits Nvidia's lock-in out of PyTorch and points its $25 billion chip business at CoreWeave and Nebius

> Amazon's TorchNeuron runs PyTorch apps on its Trainium chips after one configuration change, and Amazon is donating code upstream to strip PyTorch of its chip-specific ties. The same day, Citi's Tyler Radke told CNBC that bond yields, not portability, are what price neoclouds like CoreWeave and Nebius.

[Amazon](https://theinference.org/markets/companies/amazon) announced TorchNeuron on Monday October 19 and opens the pip-installable public beta on Tuesday October 20: applications written in PyTorch, the open-source framework most AI code is built with, then run on Amazon's Trainium AI chips after one configuration change, model.to('neuron'), no rewrite required [1][5]. Hours later on [CNBC](https://www.cnbc.com/2026/10/15/whats-behind-the-sell-off-in-ai-neocloud-stocks-like-coreweave-and-nebius-.html), [Citi](https://theinference.org/markets/companies/citigroup)'s co-head of US software research Tyler Radke argued the reverse: bond yields, not portability, set prices for the neoclouds, the specialized AI-computing renters [CoreWeave](https://theinference.org/markets/companies/coreweave) and [Nebius](https://theinference.org/markets/companies/nebius) [3][4].

## The port is routine; the upstream edits are the attack

Config-level ports to [AMD](https://theinference.org/markets/companies/amd) or [Apple](https://theinference.org/markets/companies/apple) chips already existed, so the easy port itself is routine [5]. The new move is inside the framework. AWS has [filed pull requests to PyTorch itself](https://pytorch.org/blog/introducing-torchneuron-native-pytorch-backend-for-aws-trainium/) that [strip out hardcoded assumptions favoring CUDA](https://theinference.org/article/deepseek-just-gave-away-the-software-layer-nvidia-s-china-moat-stands-on), the [Nvidia](https://theinference.org/markets/companies/nvidia) software layer that made its chips the default, and it registered Trainium through PrivateUse1, the extension point the PyTorch community built for custom hardware [1].

Colin Brace, VP of technology software at Annapurna Labs, Amazon's chip-design unit, named the market he is hunting: founders betting on "single-purpose compute providers" like CoreWeave and Nebius [5]. Amazon counts 85% of AI researchers and developers writing their code in PyTorch [2].

## A $25 billion run rate, and OpenAI's 2 gigawatts

The prize is already sized. Amazon's custom-silicon business clears a $25 billion annual revenue run rate, and [OpenAI](https://theinference.org/markets/companies/openai) signed for 2 gigawatts of Trainium in February inside [a deal in which Amazon is investing $50 billion in the lab](https://www.aboutamazon.com/news/innovation-at-amazon/torchneuron-pytorch-aws-trainium-chips-ai) [2][6]. Deepgram, a speech-AI startup, measured about 60% more audio seconds per dollar on Trainium than its GPU baseline, per its CTO; Brace cites 70% [1][5]. "The real story is removing the lock-in that kept developers from having a real choice at all," Brace said [5].

*Chart: **Nvidia's quarterly revenue has risen more than fivefold, the default now under attack** Line chart of Nvidia's reported quarterly revenue over twelve quarters, rising from $18.1B to $96.2B*

*Reported quarterly revenue, US$B, calendar-labeled quarters covering Nvidia's fiscal periods as filed. Source: Nvidia SEC filings. [11]*

## Citi's desk: the scarcity is real, the debt is the problem

Radke's arithmetic runs through rates. [The 10-year Treasury yield printed 5.09% on October 16](https://fred.stlouisfed.org/series/DGS10), which he puts 30% above its March lows [3][7]. That repricing, not weak demand, cut CoreWeave more than 25% last week to $61.39; its 2032 bond yields 15.77%, and five-year credit default swaps, debt insurance, trade near 1,140 basis points [8][9].

*Chart: **CoreWeave debt nearly quintupled in seven quarters while revenue more than tripled** Line chart of CoreWeave's reported debt and revenue by quarter: debt rises from $10.6 billion in Q4 2024 to $51.6 billion in Q2 2026 while revenue rises from $0.7 billion to $2.6 billion*

*Total debt and revenue as reported, US$B, by calendar quarter. Wells Fargo models net debt above $200B by 2030. Source: CoreWeave SEC filings. [12][9]*

## Demand is strong, and posted prices keep rising

Demand on his tape is strong: H100 spot rentals at record highs, labs locking in contracts at the highest prices they have seen [3]. [Anthropic](https://theinference.org/markets/companies/anthropic), the AI lab behind the Claude models in which Amazon has invested billions, posted $16.8 billion of quarterly revenue, up from $11.6 billion [8].

Posted price lists tell the same story. Nebius lifted [its posted on-demand H100 rate 16.9% to $4.50 an hour](https://financefeeds.com/nebius-coreweave-nvidia-gpu-prices-up-21-percent/) effective October 1, and CoreWeave is signing three-to-six-month contracts near $40 million per megawatt of annualized revenue [10]. [Wells Fargo](https://theinference.org/markets/companies/wells-fargo) models CoreWeave net debt above $200 billion by 2030 and still rates the stock Overweight [9].

*Chart: **Nebius raised posted Nvidia rental rates up to 21% on October 1** Bar chart of Nebius posted on-demand rate increases by GPU: H100 16.9%, H200 20.0%, B200 18.9%, B300 21.0%*

*Increase in the posted on-demand GPU-hour rate effective October 1, 2026, versus the prior column on nebius.com/prices; posted card rates, not receipts. Source: FinanceFeeds, October 2, 2026. [10]*

## No document prices the switching-cost premium

No document prices how much of a neocloud's market value, or Nvidia's, is a switching-cost premium, and no source yet shows a customer moving chips because of TorchNeuron; the record shows code changes, not migrations [1][5]. [Gartner](https://theinference.org/markets/companies/gartner) sets the bear ceiling, predicting more than 80% of neocloud providers cease independent operations by 2030 [8]. Radke keeps his own conviction in software, [Microsoft](https://theinference.org/markets/companies/microsoft) and [Palantir](https://theinference.org/markets/companies/palantir), the data-analytics maker, because their value does not require a view on [GPU depreciation or financing costs](https://theinference.org/article/nvidia-prices-its-gpus-at-ten-years-wall-street-lends-against-three) [3].

## A free port names the winners and losers

A free port lowers the cost of leaving, and that names the winners and losers: Amazon's Trainium and cloud business gain the exit, while neocloud equity and Nvidia's software lock-in carry the premium at risk. Three readings settle it: whether the PyTorch maintainers accept Amazon's CUDA-stripping changes after Tuesday's open-source launch [1]; whether Nebius's posted $4.50 hour actually bills [10]; and CoreWeave's November report, where a $200 billion debt path meets a $40 million megawatt [9].

## Takeaway

Amazon's one-line port to Trainium lowers the cost of leaving Nvidia, but Citi's desk says bond yields, not portability, are what price CoreWeave and Nebius.

## Sources

1. [PyTorch blog (AWS and PyTorch contributors), "Introducing TorchNeuron: Native PyTorch backend for AWS Trainium", 19 October 2026](https://pytorch.org/blog/introducing-torchneuron-native-pytorch-backend-for-aws-trainium/)
2. [Amazon (About Amazon), "Amazon opens its AWS Trainium AI chips to every PyTorch developer, no rewrite needed", 19 October 2026](https://www.aboutamazon.com/news/innovation-at-amazon/torchneuron-pytorch-aws-trainium-chips-ai)
3. [Seeking Alpha, "Bonds have taken control of the markets' narrative - analyst" (Citi's Tyler Radke on CNBC), 19 October 2026](https://seekingalpha.com/news/4654432-bonds-have-taken-control-of-the-markets-narrative-analyst)
4. [CNBC via MSN, "Bond market is in control of the narrative, says Citi's Tyler Radke" (Closing Bell Overtime), 19 October 2026](https://www.msn.com/en-us/money/investment/bond-market-is-in-control-of-the-narrative-says-citi-s-tyler-radke/vi-AA2fBJh5)
5. [TechCrunch, "Amazon is taking its biggest shot at Nvidia yet with new TorchNeuron dev tool", 19 October 2026](https://techcrunch.com/2026/10/19/amazon-is-taking-its-biggest-shot-at-nvidia-yet-with-new-torchneuron-dev-tool/)
6. [GeekWire, "Amazon challenges Nvidia's software advantage with new tool for its own AI chips", 19 October 2026](https://www.geekwire.com/2026/amazon-challenges-nvidias-software-advantage-with-new-tool-for-its-ai-chips/)
7. [FRED, Federal Reserve Bank of St. Louis, 10-Year Treasury Constant Maturity (DGS10), observation 16 October 2026, retrieved 19 October 2026](https://fred.stlouisfed.org/series/DGS10)
8. [CNBC, "What's behind the sell-off in AI neocloud stocks like CoreWeave and Nebius", 15 October 2026](https://www.cnbc.com/2026/10/15/whats-behind-the-sell-off-in-ai-neocloud-stocks-like-coreweave-and-nebius-.html)
9. [Yahoo Finance, "Wells Fargo sees CoreWeave net debt topping $200 billion by 2030, keeps Overweight", 19 October 2026](https://finance.yahoo.com/markets/stocks/articles/wells-fargo-sees-coreweave-net-142157361.html)
10. [FinanceFeeds, "Nebius and CoreWeave Lift Nvidia GPU Prices by Up to 21%", 2 October 2026](https://financefeeds.com/nebius-coreweave-nvidia-gpu-prices-up-21-percent/)
11. [Nvidia, quarterly figures from its SEC filings, retrieved 19 October 2026](https://www.sec.gov/cgi-bin/browse-edgar?action=getcompany&CIK=0001045810)
12. [CoreWeave, quarterly figures from its SEC filings, retrieved 19 October 2026](https://www.sec.gov/cgi-bin/browse-edgar?action=getcompany&CIK=0001769628)


---

Amazon edits Nvidia's lock-in out of PyTorch and points its $25 billion chip business at CoreWeave and Nebius — The Inference. Canonical: https://theinference.org/article/amazon-edits-nvidia-s-lock-in-out-of-pytorch-and-points-its-25-billion-chip-business-at-coreweave-and-nebius