Anthropic baut Gigawatt-TPU-Kapazität mit Google und Broadcom aus

Neue Rechenkapazität ab 2027 wird Claude und andere Frontier-Modelle unterstützen – Anzeichen für extremen Bedarf bei KI-Infrastruktur.

https://agentenlog.de/posts/2026-04-09-anthropic-expandiert-partnerschaft-mit-google-und-broadcom-fuer-gigawatt-tpu-kapazitaet

#Anthropic #TPU

Anthropic baut Gigawatt-TPU-Kapazität mit Google und Broadcom aus – Agentenlog

Neue Rechenkapazität ab 2027 wird Claude und andere Frontier-Modelle unterstützen – Anzeichen für extremen Bedarf bei KI-Infrastruktur.

Agentenlog

Anthropic (@AnthropicAI)

Anthropic의 Claude 관련 수요가 빠르게 증가하며 연간 매출 실행률(run-rate revenue)이 300억 달러를 넘어섰다는 내용이다. 2025년 말 90억 달러에서 크게 늘어난 수치로, 앞서 발표된 대규모 TPU 협력이 Claude 확장에 필요한 컴퓨팅을 뒷받침한다.

https://x.com/AnthropicAI/status/2041275563466502560

#anthropic #claude #revenue #tpu #ai

Anthropic (@AnthropicAI) on X

Our run-rate revenue has surpassed $30 billion, up from $9 billion at the end of 2025, as demand for Claude continues to accelerate. This partnership gives us the compute to keep pace. Read more: https://t.co/XgSjL0And7

X (formerly Twitter)

Anthropic (@AnthropicAI)

Anthropic이 Google, Broadcom과 차세대 TPU 수 기가와트 규모의 공급 계약을 체결했다. 2027년부터 가동이 시작되며, 이를 통해 Claude 모델을 학습하고 서비스할 컴퓨팅 인프라를 확보해 프런티어 모델 경쟁력을 강화하려는 움직임이다.

https://x.com/AnthropicAI/status/2041275561704931636

#anthropic #claude #google #broadcom #tpu

Anthropic (@AnthropicAI) on X

We've signed an agreement with Google and Broadcom for multiple gigawatts of next-generation TPU capacity, coming online starting in 2027, to train and serve frontier Claude models.

X (formerly Twitter)

ITmedia NEWS (@itmedia_news)

Anthropic이 Google, Broadcom과의 협력을 확대해 차세대 TPU 기반으로 3.5GW 규모의 AI 인프라를 확보한다고 발표했다. 대규모 컴퓨팅 자원 확보를 통해 향후 모델 학습·추론 역량을 크게 강화할 것으로 보이는 중요한 AI 인프라 뉴스다.

https://x.com/itmedia_news/status/2041325098578383332

#anthropic #google #broadcom #tpu #aiinfrastructure

ITmedia NEWS (@itmedia_news) on X

Anthropic、GoogleとBroadcomとの提携拡大 次世代「TPU」で3.5GWのAIインフラ確保へ https://t.co/fLjpTbzIzT

X (formerly Twitter)

Googleとの$210億TPU契約:Anthropicが“コンピュート単一統合”という賭けに出た

AI企業が計算資源を争う2026年、Anthropicは$210億相当のTPU(Tensor Processing Unit、AI処理専用半導体)チップを確保するため、GoogleとBroadcomとの大型提携を締結した。金額だけが問題ではない。OpenAIが複数ベンダーへ計算負荷を分散させる方向に動く中、Anthropicはあえて「Googleとの深い単一統合」を選択した。Claude […]

https://xenospectrum.com/anthropic-google-broadcom-tpu-compute/

Wes Roth (@WesRoth)

Anthropic가 Google과 Broadcom과 새 계약을 맺어 2027년부터 가동될 차세대 TPU 컴퓨트 용량 수 기가와트를 확보했다. Google과의 거래에도 불구하고 AWS가 여전히 Anthropic의 주요 클라우드 파트너라고 재확인했다.

https://x.com/WesRoth/status/2041320057200898188

#anthropic #google #broadcom #tpu #aws

Wes Roth (@WesRoth) on X

Anthropic has signed a new agreement with Google and Broadcom to secure multiple gigawatts of next-generation TPU compute capacity, which is expected to come online starting in 2027. Despite the Google deal, Anthropic explicitly reaffirmed that Amazon (AWS) remains its primary

X (formerly Twitter)

AI Leaks and News (@AILeaksAndNews)

Anthropic이 Google 및 Broadcom과 차세대 TPU 수기가와트 규모 공급 계약을 체결했다. 2027년부터 추가 컴퓨팅 자원이 공급되며, Claude 모델의 학습과 추론을 위한 대규모 AI 인프라 확보라는 점에서 매우 중요한 발표다.

https://x.com/AILeaksAndNews/status/2041277524596888036

#anthropic #google #broadcom #tpu #claude

AI Leaks and News (@AILeaksAndNews) on X

Anthropic have signed a deal for multiple gigawatts of next generation TPUs The new AI compute deal with Google and Broadcom will bring much needed additional capacity online starting in 2027 A new round of compute to train and inference Claude models

X (formerly Twitter)
Google stellt TPU v7 'Ironwood' vor und meldet eine Reduktion der Compute Carbon Intensity auf 79 gCO2e/EFLOP. Der Drop resultiert primär aus verfünffachter Rechenleistung gegenüber v5p. Gleichzeitig sinkt der Energie-Overhead durch optimiertes Workload-Scheduling und FP8-Quantisierung auch in älteren Modellreihen. #Google #TPU #Hardware #AI #News
https://www.all-ai.de/news/news26/google-ironwood-emmisionen
Googles Ironwood-Chip senkt Emissionen um Faktor 3,7

Die siebte TPU-Generation fängt den Energiehunger neuer KI-Modelle durch verfünffachte Rechenleistung auf.

All-AI.de
🌗 介紹 nanocode:你能用 200 美元買到的最強 Claude Code
➤ 以憲法 AI 與 TPU 算力打造專屬的 AI 程式設計夥伴
https://github.com/salmanmohammadi/nanocode/discussions/1
「nanocode」是一個開源庫,旨在引導開發者從零開始訓練專屬的「代理型(Agentic)」程式設計模型。該專案啟發自 Andrej Karpathy 的 nanochat,採用與 Anthropic 類似的「憲法 AI(Constitutional AI)」訓練策略。開發者透過定義核心指導方針(SOUL.md)、建構代理介面、生成合成數據,並運用偏好優化(Preference Optimization)來對齊模型行為。該專案不僅高度優化了 TPU 上的訓練效能,還調整了 tokenizer 以提升程式碼的處理效率,讓個人開發者能以極具成本效益的方式(約 200 美元),在數小時內訓練出專用的高階程式設計 AI。
+ 這太驚人了!僅僅 200 美元就能訓練一個
#人工智慧 #程式開發 #模型訓練 #JAX #TPU
Introducing nanocode: The best Claude Code that $200 can buy. · salmanmohammadi nanocode · Discussion #1

I'm so excited to share nanocode. This is a library showing you how to train your own Claude Code end-to-end. To a first approximation, we will follow the simplest possible approach for training us...

GitHub