AI Chips & Data Centres Daily

Patrick · Nerra Network

The silicon, the buildings and the power behind AI. Every day: the chips and systems that shipped, the data centres that broke ground or switched on, the grid and cooling constraints that decide where they go, and the supply chain and export rules underneath it all — with the numbers. Engineering and capex first, never hype. AI Disclosure: This podcast is curated by Patrick but uses AI-generated voice synthesis for audio production.

  1. 1d ago

    Ep 47: TSMC is studying two paths into 3D DRAM Nvidia wants made in-house.

    AI Chips & Data Centres Daily TSMC is studying two paths into 3D DRAM Nvidia wants made in-house. What You Need to Know: TSMC is studying two paths into 3D DRAM after a request from Nvidia: build a new fab, or buy a plant and convert it. The Information’s November 6 report, carried by several outlets, describes cell-stacked memory, not an extension of the HBM4 base-die work TSMC already does with SK hynix. ... Sources: xenospectrum.com · semiconductor-digest.com · datacenterknowledge.com · utilitydive.com · datacenterdynamics.com · hpcwire.com · techpowerup.com · fourweekmba.com AI Disclosure: This podcast is curated by Patrick but uses AI-generated voice synthesis for audio production. 🎬 Watch on YouTube (public from 8:30 AM ET): https://www.youtube.com/watch?v=HEARXpnL08w If this episode was useful, a rating or a short review on Apple Podcasts or Spotify is how the next listener finds the show — and following it in your app means the next episode is there when you are. 📝 Full show notes, transcript & sources: read the episode page 📬 Get the sources and notes by email, free: join the newsletter 🌐 Part of the Nerra Network — explore every show at nerranetwork.com.

  2. 2d ago

    Ep 46: Reported China orders of six million RTX PRO 5500s exceed a reported three-million 2027…

    AI Chips & Data Centres Daily Reported China orders of six million RTX PRO 5500s exceed a reported three-million 2027 build. What You Need to Know: Reported orders from China for Nvidia’s RTX PRO 5500 top six million cards, against a source’s ceiling of three million units that can be built in 2027, while a US listing prices the cut-down card above the flagship RTX PRO 6000. ... Sources: techpowerup.com · hpcwire.com · datacenterdynamics.com · utilitydive.com · semiengineering.com · eetimes.com · tomshardware.com · semiconductor-digest.com AI Disclosure: This podcast is curated by Patrick but uses AI-generated voice synthesis for audio production. 🎬 Watch on YouTube (public from 8:30 AM ET): https://www.youtube.com/watch?v=Csp5_v7OuM4 If this episode was useful, a rating or a short review on Apple Podcasts or Spotify is how the next listener finds the show — and following it in your app means the next episode is there when you are. 📝 Full show notes, transcript & sources: read the episode page 📬 Get the sources and notes by email, free: join the newsletter 🌐 Part of the Nerra Network — explore every show at nerranetwork.com.

  3. 3d ago

    Ep 45: Nvidia ships CUDA 13.5, taking Vera Rubin from early access into production deployments.

    AI Chips & Data Centres Daily Nvidia ships CUDA 13.5, taking Vera Rubin from early access into production deployments. What You Need to Know: Nvidia has shipped CUDA 13.5, moving Vera Rubin into production on a platform it has described as up to 144 GPUs per rack. Samsung’s second Taylor, Texas fab is reported fully booked, mostly by AMD, for 2 nm HBM base dies. ... Sources: finance.biggo.com · techpowerup.com · hpcwire.com · datacenterknowledge.com · utilitydive.com · datacenterdynamics.com · ad-hoc-news.de · tomshardware.com AI Disclosure: This podcast is curated by Patrick but uses AI-generated voice synthesis for audio production. 🎬 Watch on YouTube (public from 8:30 AM ET): https://www.youtube.com/watch?v=CE07IgpEtjY If this episode was useful, a rating or a short review on Apple Podcasts or Spotify is how the next listener finds the show — and following it in your app means the next episode is there when you are. 📝 Full show notes, transcript & sources: read the episode page 📬 Get the sources and notes by email, free: join the newsletter 🌐 Part of the Nerra Network — explore every show at nerranetwork.com.

  4. 4d ago

    Ep 44: TSMC expects A16 volume production in 2027, a year ahead of Intel's 14A.

    AI Chips & Data Centres Daily TSMC expects A16 volume production in 2027, a year ahead of Intel's 14A. What You Need to Know: TSMC expects A16 volume production in 2027, a year ahead of Intel's planned 14A high-volume ramp. The node is the company's first with Super Power Rail backside power delivery, and it is not using High-NA EUV to get there. ... Sources: startupfortune.com · ir.amd.com · techpowerup.com · hpcwire.com · datacenterdynamics.com · semiconductor-digest.com · eetimes.com · digitimes.com AI Disclosure: This podcast is curated by Patrick but uses AI-generated voice synthesis for audio production. If this episode was useful, a rating or a short review on Apple Podcasts or Spotify is how the next listener finds the show — and following it in your app means the next episode is there when you are. 📝 Full show notes, transcript & sources: read the episode page 📬 Get the sources and notes by email, free: join the newsletter 🌐 Part of the Nerra Network — explore every show at nerranetwork.com.

  5. 5d ago

    Ep 43: Twenty billion pounds is lined up for up to one gigawatt beside a Yorkshire power station.

    AI Chips & Data Centres Daily Twenty billion pounds is lined up for up to one gigawatt beside a Yorkshire power station. What You Need to Know: Brookfield’s estimate for a campus of up to one gigawatt at Drax’s North Yorkshire station is more than twenty billion pounds, with a first phase of up to one hundred megawatts aimed at the first quarter of 2028. ... Sources: datacenterdynamics.com · eetimes.com · hpcwire.com · datacenterknowledge.com · tomshardware.com · businesskorea.co.kr · electronicsforyou.biz · tradingview.com AI Disclosure: This podcast is curated by Patrick but uses AI-generated voice synthesis for audio production. If this episode was useful, a rating or a short review on Apple Podcasts or Spotify is how the next listener finds the show — and following it in your app means the next episode is there when you are. 📝 Full show notes, transcript & sources: read the episode page 📬 Get the sources and notes by email, free: join the newsletter 🌐 Part of the Nerra Network — explore every show at nerranetwork.com.

  6. 6d ago

    Ep 42: A 72-GPU Rubin rack now carries 20.7 terabytes of HBM4 for inference.

    AI Chips & Data Centres Daily A 72-GPU Rubin rack now carries 20.7 terabytes of HBM4 for inference. What You Need to Know: ASRock Rack showed a Vera Rubin NVL72 at OCP with 20.7 terabytes of HBM4 across 72 GPUs. OpenAI left its Seoul meetings without a new memory contract, while TrendForce sees HBM's blended price more than doubling in 2027. ... Sources: servethehome.com · t.co · eetimes.com · wccftech.com · newsletter.semianalysis.com · cio.economictimes.indiatimes.com · finance.biggo.com · finance.yahoo.com AI Disclosure: This podcast is curated by Patrick but uses AI-generated voice synthesis for audio production. If this episode was useful, a rating or a short review on Apple Podcasts or Spotify is how the next listener finds the show — and following it in your app means the next episode is there when you are. 📝 Full show notes, transcript & sources: read the episode page 📬 Get the sources and notes by email, free: join the newsletter 🌐 Part of the Nerra Network — explore every show at nerranetwork.com.

  7. Nov 1

    Ep 41: Seven point three billion dollars of quarterly data-center revenue still leaves Intel short…

    AI Chips & Data Centres Daily Seven point three billion dollars of quarterly data-center revenue still leaves Intel short of Xeon demand. What You Need to Know: Intel's data-center and AI business booked $7.3 billion in the third quarter of 2026, up 17% from the prior quarter and nearly 80% from a year earlier, and the company is still short of Xeon supply. ... Sources: tomshardware.com · koreatimes.co.kr · servethehome.com · arxiv.org · info.siteselectiongroup.com · techpowerup.com AI Disclosure: This podcast is curated by Patrick but uses AI-generated voice synthesis for audio production. If this episode was useful, a rating or a short review on Apple Podcasts or Spotify is how the next listener finds the show — and following it in your app means the next episode is there when you are. 📝 Full show notes, transcript & sources: read the episode page 📬 Get the sources and notes by email, free: join the newsletter 🌐 Part of the Nerra Network — explore every show at nerranetwork.com.

  8. Oct 31

    Ep 40: Six hundred million dollars is what a two-week Fab 28 halt costs Intel in server-chip…

    AI Chips & Data Centres Daily Six hundred million dollars is what a two-week Fab 28 halt costs Intel in server-chip revenue. What You Need to Know: Six hundred million dollars is what Intel expects to lose across two quarters after a two-week shutdown at Fab 28 near Kiryat Gat, almost all of it in 10-nanometer-class server chips. ... Sources: techpowerup.com · servethehome.com · tomshardware.com · datacenterdynamics.com · digitalinformationworld.com · bisnow.com · asiae.co.kr AI Disclosure: This podcast is curated by Patrick but uses AI-generated voice synthesis for audio production. If this episode was useful, a rating or a short review on Apple Podcasts or Spotify is how the next listener finds the show — and following it in your app means the next episode is there when you are. 📝 Full show notes, transcript & sources: read the episode page 📬 Get the sources and notes by email, free: join the newsletter 🌐 Part of the Nerra Network — explore every show at nerranetwork.com.

See All (47)

About

The silicon, the buildings and the power behind AI. Every day: the chips and systems that shipped, the data centres that broke ground or switched on, the grid and cooling constraints that decide where they go, and the supply chain and export rules underneath it all — with the numbers. Engineering and capex first, never hype. AI Disclosure: This podcast is curated by Patrick but uses AI-generated voice synthesis for audio production.

Information