Acquired live from Google Cloud Next

Year: 2026 · ▶ Watch on YouTube

Jason Davenport (Technical Lead for Developer Experience) · Ben Gilbert (Co-Founder, Host Acquired) · David Rosenthal (Co-Founder, Host Acquired)

Switch language → zh

Segments (7)

  • 00:00:00 · Introduction — Jason Davenport
    • The host introduces the show and his guests, Ben Gilbert and David Rosenthal from the Acquired podcast.
  • 00:31:00 · Reflecting on the Google Saga — Ben Gilbert
    • The guests discuss their extensive podcast series on Google and whether the latest announcements necessitate a fourth part.
  • 01:56:00 · The Shift from Training to Inference — Ben Gilbert
    • The guests analyze the keynote’s focus on the growing importance of inference workloads over training, highlighted by the new TPU v8 architecture.
  • 03:28:00 · Barriers to AI Adoption — Jason Davenport
    • The conversation explores the challenges of scaling AI from proof-of-concept to widespread use and the barriers to adopting AI agents.
  • 06:36:00 · AI Adoption: The Internet but Faster — Ben Gilbert
    • The guests compare the rapid adoption of AI to the internet, noting the shift from a one-time ‘crawl’ (training) to continuous ‘serving’ (inference).
  • 09:08:00 · Google Cloud’s Enterprise Transformation — David Rosenthal
    • The guests describe Google Cloud’s evolution from a developer-focused platform to a major enterprise player, driven by a massive expansion of its sales force under Thomas Kurian.
  • 11:48:00 · Looking Ahead: The Impact of Instant Inference — Ben Gilbert
    • The guests express excitement for how the 5x speedup in TPU v8’s inference will make AI tasks feel synchronous, unlocking new, unforeseen applications.

Products Announced (1)

  • 02:01:00 · TPU v8 (New)
    • Split architecture with dedicated training and inference chips · 5x speedup in inference performance · Significant generational improvement over TPU v7
    • Announced at keynote

Customer Stories (2)

  • 12:04:00 · Snap — Early, large-scale user of Google App Engine, becoming a major customer by accident before Google Cloud had a formal enterprise sales motion.
  • 08:10:00 · YouTube TV — A demo from the keynote showed an AI agent handling a multilingual customer support scenario for YouTube TV.

Benchmarks Shown (1)

  • 14:02:00 · TPU v8 Inference Performance: 5x
    • Compared to previous generation TPU

Demos (2)

  • 08:10:00 ✓ · YouTube TV AI Support Agent — Referenced from keynote
    • An AI agent handling a multilingual customer support request for YouTube TV, demonstrating real-time translation and problem-solving.
  • 17:43:00 ✓ · Slide Deck Generation in Gemini — Referenced from keynote
    • Generating a slide deck directly within Gemini for Enterprise.

Notable Quotes (4)

  • 00:59:00 — Ben Gilbert:

    So the question is, do we need a part four here at some point?

  • 06:47:00 — Ben Gilbert:

    The internet but faster.

  • 11:03:00 — David Rosenthal:

    It’s not even like a transformation, it’s like a, I don’t know, refounding or something. Like, Google Cloud today versus Google Cloud 10 years ago are like completely different species.

  • 11:43:00 — Ben Gilbert:

    We need 1,000 times more salespeople in order to run an effective, customer-focused enterprise organization.

Visual Signals

On-screen (6)

  • 00:06:00 · Jason Davenport, Technical Lead for Developer Experience, Google Cloud
    • Identifies the host and his role at Google.
  • 00:11:00 · Google Cloud Next Live from Vegas
    • Establishes the context of the event and location.
  • 01:10:00 · Ben Gilbert, Co-Founder, Host Acquired
    • Identifies one of the guests and his affiliation.
  • 02:31:00 · David Rosenthal, Co-Founder, Host Acquired
    • Identifies the other guest and his affiliation.
  • 20:13:00 · Google Cloud Next '26 logo
    • Branding for the event.
  • 20:17:00 · Gemini, summarize this break: 'Be right back.'
    • A creative, branded interstitial graphic for the live stream break.

Stage (1)

  • 00:00:00 · Three speakers are seated at a desk in a live stream studio setup, with a convention floor visible in the background.

Key Topics

Artificial Intelligence · Google Cloud · Acquired Podcast · TPU v8 · Inference vs. Training · AI Adoption · Enterprise Cloud · Google History · Gemini · AI Agents · Cloud Transformation · Jeff Dean · Thomas Kurian · Cloud Infrastructure

Takeaways

  • The AI industry is experiencing a significant shift from focusing on training workloads to optimizing for inference, which is becoming the dominant compute use case.
  • Google’s new TPU v8, with its specialized training and inference chips, represents a major step in this direction, promising a 5x speedup in inference that could enable more real-time, synchronous AI applications.
  • The primary barrier to widespread AI adoption is shifting from technological capability to user behavior and the lack of intuitive, widely adopted user interfaces for AI agents.
  • Google Cloud has fundamentally transformed from a developer-centric PaaS into a formidable enterprise-first cloud, a change catalyzed by leadership decisions to massively scale its customer-facing sales organization.
  • Google’s vertical integration of its own chips (TPU), cloud platform, and frontier models (Gemini) provides a unique structural advantage over other hyperscalers.
  • The analogy for the current AI era is ‘the internet, but faster,’ suggesting a more compressed timeline for both technological advancement and market disruption compared to the early web.
  • As AI becomes more efficient and cost-effective, the next wave of innovation will likely come from true business transformation and the creation of new, AI-native business models, rather than just efficiency gains.