Peter Thiel has just backed a bold new vision for AI infrastructure. His latest $140M investment is going into Panthalassa - a startup developing floating AI compute platforms powered by ocean waves.

As energy grids struggle to keep up and permits for new facilities become harder to secure, the next generation of AI data centers may not be built on land at all. Silicon Valley is now looking offshore for scalable AI power.

Here’s what we’re covering today 👇

• 🌊 Peter Thiel’s Biggest AI Bet Is Heading Out to Sea
• 💼 Coinbase Cuts 14% of Jobs as CEO Brian Armstrong Pushes AI-First Strategy

AI infrastructure is heading into uncharted territory 🌊

Panthalassa has raised $140M to build floating AI compute nodes powered entirely by ocean waves- no traditional power grid required.

The funding round was led by Peter Thiel, with backing from John Doerr, Marc Benioff’s TIME Ventures, and Founders Fund.

Here’s the wild part 👇

These 85-meter offshore steel platforms are designed to generate energy directly from ocean waves while running AI inference workloads onboard. Instead of transmitting electricity back to shore, the system sends processed AI outputs to customers through low-Earth-orbit satellites.

Their next major deployment, Ocean-3, is scheduled for the northern Pacific in 2026, with commercial fleets expected to follow in 2027.

The company has now raised around $210M total, building on earlier Ocean-1 and Ocean-2 prototypes tested in 2021 and 2024.

And honestly, this idea starts making a lot more sense when you look at what’s happening on land.

AI companies are running into massive energy bottlenecks:
• Nearly half of planned global data center projects for 2026 are reportedly facing delays due to power shortages
• Morgan Stanley estimates the US could face a 49 GW energy shortfall by 2028
• Microsoft has already started signing direct gas agreements with Chevron to keep AI projects moving

Panthalassa’s bet is simple:

Building AI infrastructure in the ocean may end up being faster than waiting for land-based permits and grid expansion.

The future of AI may not just be smarter models.
It may also depend on who can generate enough power to run them.

Coinbase is reportedly laying off around 693 employees as CEO Brian Armstrong pushes the company toward an AI-first operating model.

The vision is aggressive:

• Armstrong wants AI to generate 50% of Coinbase’s internal code
• Tools like Copilot and Cursor are becoming mandatory across teams
• Traditional org structures are being flattened to a maximum of five layers
• Small “AI-native pods” are replacing larger teams, sometimes with just one person handling engineering, product, and design together

According to Armstrong, engineers can now ship in days what entire teams once took weeks to build.

That shift is also changing leadership expectations inside the company. Pure management roles are reportedly being phased out, with leaders expected to remain hands-on contributors instead of only overseeing teams.

Coinbase expects to take around $50M-$60M in restructuring charges during Q2, while affected US employees will receive severance packages based on tenure.

What’s fascinating is how the market reacted.

Despite posting a reported $667M quarterly loss and facing continued pressure from declining crypto prices, Coinbase shares still climbed after the announcement. Investors appear to be rewarding the AI productivity narrative even as the broader crypto market remains under pressure.

And that’s the bigger story here.

AI is increasingly becoming both:
• a genuine productivity multiplier
• and a strategic narrative companies can lean on during difficult market cycles

The question now is whether these AI-first restructures actually create long-term efficiency or simply reduce headcount faster than organizations can adapt.