AI 行业热点
🎙️ 播客精选
Who Feeds the GPUs? Inside AI’s Hidden $30B Layer | Renen Hallak, VAST Data
The MAD Podcast with Matt Turck · 2026-09-24
Speaker 1 | 00:00 - 00:17
Sometimes it scares me. We had a customer, one of these AI clouds, they said we’re probably going to need about 500 petabytes over the next three years. Last week, came back to us and said, we’re gonna need an extra two exabytes on top of that 500 petabytes. I think in the next ten years, we’ll see more difference than we did in the last thousand years.
Speaker 2 | 00:17 - 00:30
Hi. I’m Matt Turk. Welcome to the MAD podcast. We talk a lot about AI models on this show, and we talk a lot about compute. But today is about the layer in between, the crucial software infra…
🐦 X/Twitter 热点
Swyx (@swyx)
- In Jan this year I called my content strategy shot: “Scaling without Slop”.
It’s finally starting to work. It took us 3 years to reach our first 100k on youtube. It only took 1.2 months for the next 100k. Similar other metrics on AEO/SEO/subscriber traction and have a lot of New Media ideas that I’m excited to pursue.
officially giving notice of the next phase of Latent Space, AINews, and what the rest of swyx inc has been cooking below [23 ❤️]
- more conferences could implement this. so much high value time wasted without thought [504 ❤️ 29 🔄]
Josh Woodward (@joshwoodward)
- [5 ❤️]
- Dreambeans is one of our newer experiments in @GoogleLabs. It has a growing cult following.
It’s simple: a fixed number of “beans” brew every morning. They point you at the real world, with real people, doing things you care about together.
Give it a try! [149 ❤️ 5 🔄]
Peter Yang (@petergyang)
- Astra blew up 3D models
Then Opus blew up videos
I don’t even know what’s next anymore [12 ❤️] - 😂 this is love [39 ❤️]
- This is called stroking the AI’s ego and it works [17 ❤️]
Thariq (@trq212)
- incredibly thoughtful piece on AI and creativity
hopefully we can make our producst better at enabling and amplifying creatives [518 ❤️ 12 🔄]
- this should let you customize the plan mode prompt, create + share your own modes or just ignore it and rebind shift+tab to something else all together [137 ❤️]
- lots feedback here, many of you are planning yourself & don’t need plan mode
others prefer the UX of entering a mode where Claude is just thinking & brainstorming with you
my plan is to:
- make plan mode into a built-in mod
- allow mods to add new modes or override shift+tab [1654 ❤️ 29 🔄]
Amjad Masad (@amasad)
- Muse can now make apps on Replit [1674 ❤️ 50 🔄]
Guillermo Rauch (@rauchg)
- Spend of OpenAI vs Anthropic vs Open
(Vercel AI Gateway, last 2 months)
• Anthropic still #1 in spend, but went 69% → 40%
• OpenAI: 10% → 24% in spend
• GPT-6 Astra + GPT 5.6 Sol are ripping
• OpenAI now leads in tokens #
• Kimi K3 + DeepSeek took ~half of Anthropic’s loss
• Opus 5.5 is up to 10% of spend in 2 days
• OpenAI is 62% of image generations
Watch here: [635 ❤️ 34 🔄]
Matt Turck (@mattturck)
- This conversation with the excellent Renen Hallak of @VAST_Data is also available on Spotify, Apple Podcasts and here on YouTube (like and subcribe!): [18 ❤️ 1 🔄]
- In Jensen’s 5 layer cake analogy for AI (energy, chips, infrastructure, models, applications), the middle layer (esp the software infra part) is much less understood - and thats where @VAST_Data became a $30B company few people know about, powering @SpaceXAI, @CoreWeave, @nebiusai, @MistralAI, @nscale etc
My conversation with Renen Hallak, CEO:
00:00 Intro
00:51 The hidden software layer in NVIDIA’s AI stack 02:28 What actually makes an “AI factory”?
05:13 Should Walmart and Goldman build their own AI?
06:20 “We infer during the day, fine-tune at night”
13:16 Announcement: frontier models & sensitive data
15:32 From P vs. NP to founding VAST Data
17:32 OpenAI, Navier–Stokes and 10,000 agents
20:25 The pre-transformer insight behind VAST
21:55 DASE: VAST’s “shared everything” architecture
25:18 “Storage was where startups go to die”
27:39 Trillions of vectors: why old databases break
29:01 Are S3, Snowflake and Databricks ready for AI
31:29 Data gravity, vendor lock-in and zero churn
33:13 Training vs. inference: why the infrastructure changes
34:46 Model routing, KV caches, RAG & agent memory
36:59 Identity, permissions and security for AI agents 40:27 Can multi-agent systems unlock scientific discovery?
41:55 DataEnclave: how confidential AI protects data and weights
45:16 Who should be AI’s trust layer?
46:37 “Sometimes it scares me”: 500 petabytes to 2 exabytes
50:08 Is circular AI financing creating systemic risk
51:32 Why VAST is profitable when AI infra isn’t
53:24 What separates the winning neoclouds?
55:06 “Their lunch is being eaten”: why hyperscalers lag
59:10 Where will the trillions accrue across the AI stack
1:01:18 NVIDIA: “There’s no legal document between us”
1:03:59 What VAST learned from xAI and @elonmusk
1:05:36 “Bad things loudly and often”: building at AI speed
1:06:53 More change in 10 years than the previous 1,000?
1:08:39 VAST’s endgame: all the data in the world [19 ❤️ 8 🔄]
Nikunj Kothari (@nikunj)
- “we believe that every small business owner, specifically for us in the trades, will have bespoke software,
and that last mile is actually where differentiation can become, where you can provide a unique experience to your customers, to your employees” [4 ❤️]
- This is the way.. [31 ❤️ 4 🔄]
- We are in that stage of the “revenue” lifecycle.. [9 ❤️]
Peter Steinberger (@steipete)
- I put Daybreak on it and found 8 more long-standing leaks. Care for your oss dependencies! [19 ❤️]
- Afghanistan is more excited about AI than we are? [317 ❤️ 14 🔄]
- If you just tell the agent to clean up, it will stop far too early. Give it an ambitious goal. Try “remove 20% of the least useful tests while maintaining code coverage within 2%” [316 ❤️ 10 🔄]
Dan Shipper (@danshipper)
- so freaking cool [44 ❤️ 1 🔄]
- LIVE: Write-along [10 ❤️]
Aditya Agarwal (@adityaag)
- It was great speaking with @SteveHiltonx at @spc for a conversation on California’s future.
We have more work to do. Onwards. [7 ❤️ 2 🔄]
📝 博客文章
由 Follow Builders 自动生成 · 2026-09-25