Up North AIUp North
Back to news

Analog Devices Buys Its Way Into Edge AI for $1.55B

OpenAI Makes ChatGPT Sites a Team Sport. Google Locks In 400MW of Geothermal for AI Data Centers.

Share

Analog Devices Buys Its Way Into Edge AI for $1.55B

ADI is acquiring Alif Semiconductor, maker of AI-native fusion processors and low-power microcontrollers, in a deal worth up to $1.55B including earnouts [4][5][6]. Alif's chips already ship in real consumer and industrial products — this isn't a research bet, it's revenue-generating silicon.

The strategic logic: pair Alif's edge compute with ADI's analog, sensing, and power expertise to build what they're calling "Physical Intelligence" — systems that sense, reason, and act locally, without round-tripping to the cloud. That's the quiet story under every AI headline this year: intelligence is moving to the edge because latency, cost, and reliability all demand it.

This is ADI's second AI silicon acquisition in 2026, and analysts on X are reading it as confirmation that the real growth market isn't another chatbot — it's robotics, industrial automation, and data center infrastructure that needs to make decisions in milliseconds, not seconds.

OpenAI Makes ChatGPT Sites a Team Sport

Three months after launch, ChatGPT Sites has already produced over 5 million sites — and OpenAI just shipped the features that turn it from a solo prototyping tool into something teams can actually use [7][8]. "Build Together" adds multi-user editing and publishing, there's private sharing before you go live, custom domains, database inspection, and deployment that's roughly twice as fast.

Colleagues collaborating around a table covered with printed AI chat logs

This is the incremental-but-important kind of update. The gap between "I can vibe-code a site alone" and "my team can build and ship one together" is exactly the gap between a toy and a tool. Developers on X are already sharing MVPs built collaboratively in hours, not sprints.

Worth watching: this is OpenAI encroaching further into no-code/low-code platform territory that used to belong to Webflow, Replit, and Vercel. The build layer is consolidating fast.

Google Locks In 400MW of Geothermal for AI Data Centers

Google signed a power purchase agreement with Fervo Energy for roughly 400MW of round-the-clock enhanced geothermal power from the Cape Station project in Utah, coming online in 2028 [9][10][11]. This builds on a 3GW framework the two companies announced back in March — Cape Station itself is permitted for up to 2GW total.

The subtext is the real story: AI's energy appetite is now large enough to single-handedly justify next-generation power infrastructure that wouldn't have penciled out otherwise. Geothermal has been "promising" for a decade; AI demand is what's actually getting it built at scale.

Sustainability voices on X are calling this a genuine win — 24/7 clean baseload power is the hard problem solar and wind haven't solved, and data centers need exactly that.

What This Means For Your Business

Three of today's four stories are really the same story wearing different clothes: the bottleneck in AI is no longer "can we build the model" — it's power, silicon, and trust. Google is buying gigawatts before they're built. ADI is buying its way into physical-world compute. Anthropic is spending real operational overhead to prove its systems can be trusted. None of that is about writing better code. It's about judgment: what to build, what to trust, what to commit capital and reputation to years in advance.

The ChatGPT Sites update is the small story that matters more than it looks. When a team can go from prompt to deployed, shared web app in half the time, with less friction than before — the scarce skill isn't "can you code the thing," it's "do you know what thing is worth building, and can you direct a team (human or AI) toward it fast enough to matter." That's the orchestration shift in miniature, playing out in a product feature.

If you're making AI decisions for your company right now, the lesson from today isn't "wait for the frontier to slow down" or "wait for the chips to arrive." It's that the companies moving fastest are the ones locking in physical and trust infrastructure now, while treating the software layer as increasingly commoditized and fast to spin up. Code is free. The scarce resource is knowing what to commit to before everyone else does.

Key takeaway: The AI race is bifurcating into two games — infrastructure and trust, which take years and capital to build, and software, which takes hours and prompts. Bet on the former, move fast on the latter, and don't confuse the two.

See what we're exploring →

Sources

  1. https://darioamodei.com/post/we-must-pace-the-frontier
  2. https://www.cnn.com/2026/09/12/tech/anthropic-ceo-essay-ai
  3. https://www.axios.com/2026/09/12/anthropic-ai-amodei-pacing
  4. https://investor.analog.com/news-releases/news-release-details/analog-devices-acquire-alif-semiconductor-adding-ai-native/
  5. https://www.reuters.com/technology/analog-devices-buy-alif-semiconductor-135-billion-2026-09-09/
  6. https://www.wsj.com/tech/analog-devices-to-acquire-alif-semiconductor-for-1-35-billion-1b31d6fe
  7. https://www.androidauthority.com/chatgpt-sites-new-features-3710506/
  8. https://www.ithome.com/1/001/485.htm
  9. https://techcrunch.com/2026/09/02/enhanced-geothermal-notches-another-win-as-google-buys-400-mw-from-fervo/
  10. https://www.bloomberg.com/news/articles/2026/09-01/fervo-signs-geothermal-power-deal-with-google-for-2028-launch
  11. https://www.canarymedia.com/articles/geothermal/fervo-google-deal-next-gen-geothermal

Stay ahead of AI

No spam. Unsubscribe anytime.

Want to go deeper?

Reading the news is one thing. Exploring the frontier is another. See what we're building.