⚡ Signal Brief — August 29, 2025

#99: This Week Tech — Microsoft’s first in-house AI, Nvidia’s blowout, Snowflake pops, Meta races Llama 4.x, and the Fed leans dovish

---

This Week in AI: model launches shift from labs to living rooms (Samsung TVs), chips and data platforms reprice the AI trade, and policymakers all but tee up a September cut.

📌 TL;DR: This Week’s Headlines

• Microsoft debuts its first homegrown AI models (MAI-Voice-1, MAI-1-preview) alongside OpenAI partnerships. The Verge

• Samsung + Microsoft: Copilot is coming to 2025 TVs and Smart Monitors. The Times of India

• Nvidia smashes Q2 FY26: $46.7B revenue; Data Center up, zero H20 sales to China noted. NVIDIA Newsroom

• Snowflake surges ~14% on raised outlook as AI workloads accelerate. Reuters

• Meta rushes next Llama (4.X/4.5) for release by year-end, per new reports. Yahoo News Canada

• Fed tilt: Powell’s Jackson Hole tone + Waller’s remarks push markets toward September cuts. ReutersMarketWatch

• Europe sovereignty moves: new EU-funded trapped-ion quantum cloud service goes live; sovereign AI cloud expansion announced. HPCwire

• Apple sets Sept 9 event with AI in focus. Reuters

Over the past week, AI went mainstream in devices and platforms: Microsoft shipped its first in-house models and Samsung will put Copilot on TVs, even as chips and data platforms (Nvidia, Snowflake) set the market tone. Meta’s race to its next Llama underscores intensifying model competition, while Europe’s compute/sovereignty push adds new quantum and “sovereign AI cloud” capacity. Macro-wise, Fed signals kept risk appetite buoyant going into September.

🔮 AI & Big Tech

• Microsoft’s first in-house models (MAI-Voice-1, MAI-1-preview)
  Low-latency speech (1 minute of audio in <1s on a single GPU) and a new LLM trained on ~15k H100s, with early Copilot uses and public testing on LMArena.
  Why it matters: Microsoft is adding its own foundation blocks next to OpenAI—expanding optionality, enabling faster vertical features, and tightening cost/control over its AI stack.

• Meta’s next Llama by year-end (reports)
  Meta is reportedly targeting “Llama 4.X/4.5” before 2026.
  Why it matters: The cadence of big-model releases is accelerating—raising pressure on inference costs, eval transparency, and enterprise migration timelines.

• JetBrains Koog 0.4.0 adds GPT-5 support & structured output
  Agent framework update lands with iOS targets and OpenTelemetry.
  Why it matters: Enterprise AI pilots move faster into production when frameworks offer observability, reliability, and structured outputs—Koog adds those missing pieces.

🎤 Tech & Organizational Moves

• Samsung to ship Copilot on 2025 TVs/Monitors
  Copilot becomes a living-room assistant for content discovery and queries.
  Why it matters: Multimodal assistants are entering ambient screens, creating new daily touchpoints and consumer data streams that feed back into ecosystem AI models.

• OpenAI talent magnet: senior move from Meta
  Meta director Chaya Nayak joins OpenAI Special Initiatives.
  Why it matters: Top operators continue to migrate toward frontier labs, strengthening OpenAI’s institutional heft in governance, policy, and scaling efforts.

📈 Market & Platform News

• Nvidia (Q2 FY26): $46.7B revenue; Blackwell DC +17% QoQ; no H20 sales to China
  Guides the AI stack; supply mix and export-control workarounds in focus.
  Why it matters: Nvidia remains the heartbeat of the AI build-out—its results steer capex cycles, while export restrictions directly reshape product allocation

• Snowflake pops ~14% on AI demand; raises product-revenue forecast
  Analysts lift targets; Databricks’ >$100B raise chatter looms in the backdrop.
  Why it matters: AI is re-rating the value of data platforms; Snowflake proves retrieval and governance layers are indispensable to AI adoption.

• Macro tailwind: Powell tone + Waller path to cuts
  Markets price higher odds of a Sept 17 cut; rotation into small caps gains steam.
  Why it matters: A cheaper cost of capital extends the AI build-out runway—making infra investments and late-stage rounds more attractive despite near-term froth.

• Apple sets Sept 9 event with AI front-and-center
  Investors watching Apple Intelligence roadmap and new hardware.
  Why it matters: Apple’s AI framing will reset consumer expectations for assistants—pushing rivals to match seamless device+cloud integration.

💸 Venture Capital & Startups

• Lovable (Stockholm) attracts bids at ~$4B valuation (report)
  Weeks after a $200M round, offers now value the “vibe coding” startup >$4B.
  Why it matters: The European AI app layer is heating up, with novel UX in coding assistants commanding venture multiples usually reserved for infra.

• Who’s raising $100M+ in 2025?
  Fresh TechCrunch tally lists 33 US AI startups with $100M+ rounds this year.
  Why it matters: Late-stage capital is consolidating in a handful of winners—distribution power and proprietary data access decide who survives.

• EU dealflow:

  • Moxso (DK) raises €4.7M seed for adaptive, AI-powered security awareness.
    Why it matters: Security startups embedding AI + human oversight are carving a sticky niche in SME and mid-market IT.
    EU-Startups

  • Interhuman AI (DK) lands €2M pre-seed for “social AI” interfaces.
    Why it matters: Interfaces that parse tone and non-verbal signals preview the next wave of human–AI interaction.
    Tech Startups

• Deal tape (PE/VC lens)
  Axios Pro Rata highlights strong late-stage/crossover activity even amid macro jitters.
  Why it matters: If September cuts land, liquidity could reopen for IPOs and mega-rounds into Q4.
  Axios

🌍 Geopolitics & Tech Power

• EU boosts sovereignty with cloud-based trapped-ion quantum access (QCDC)
  First EU cloud service for trapped-ion quantum computing goes live.
  Why it matters: Europe is expanding compute sovereignty beyond GPUs—quantum joins AI as a strategic digital capability.

• Sovereign AI cloud expansion in Europe
  VAST Data to power BUZZ HPC’s expansion for sovereign workloads.
  Why it matters: Regionalized compute and data residency give governments and enterprises confidence to run sensitive AI locally.

• Chip policy spillovers in earnings
  Nvidia disclosed no H20 sales to China in Q2.
  Why it matters: Export controls are no longer theoretical—they’re visibly shaping vendor roadmaps, SKU strategy, and global supply chains.

---

Missed last week’s update?

---

Hit subscribe to get it in your inbox. And if this spoke to you:

➡️ Forward this to a strategy peer who’s feeling the same shift. We’re building a smarter, tech-equipped strategy community—one layer at a time.

Let’s stack it up.

A. Pawlowski | The Strategy Stack